Merge branch 'master' into lto

.github/workflows/backport.yml

@@ -8,7 +8,7 @@ jobs:
     if: github.repository_owner == 'NixOS' && github.event.pull_request.merged == true && (github.event_name != 'labeled' || startsWith('backport', github.event.label.name))
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v3
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           # required to find all branches

.github/workflows/ci.yml

@@ -14,10 +14,10 @@ jobs:
     runs-on: ${{ matrix.os }}
     timeout-minutes: 60
     steps:
-    - uses: actions/checkout@v2.4.0
+    - uses: actions/checkout@v3
       with:
         fetch-depth: 0
-    - uses: cachix/install-nix-action@v16
+    - uses: cachix/install-nix-action@v17
     - run: echo CACHIX_NAME="$(echo $GITHUB_REPOSITORY-install-tests | tr "[A-Z]/" "[a-z]-")" >> $GITHUB_ENV
     - uses: cachix/cachix-action@v10
       if: needs.check_cachix.outputs.secret == 'true'
@@ -46,11 +46,11 @@ jobs:
     outputs:
       installerURL: ${{ steps.prepare-installer.outputs.installerURL }}
     steps:
-    - uses: actions/checkout@v2.4.0
+    - uses: actions/checkout@v3
       with:
         fetch-depth: 0
     - run: echo CACHIX_NAME="$(echo $GITHUB_REPOSITORY-install-tests | tr "[A-Z]/" "[a-z]-")" >> $GITHUB_ENV
-    - uses: cachix/install-nix-action@v16
+    - uses: cachix/install-nix-action@v17
     - uses: cachix/cachix-action@v10
       with:
         name: '${{ env.CACHIX_NAME }}'
@@ -67,9 +67,9 @@ jobs:
         os: [ubuntu-latest, macos-latest]
     runs-on: ${{ matrix.os }}
     steps:
-    - uses: actions/checkout@v2.4.0
+    - uses: actions/checkout@v3
     - run: echo CACHIX_NAME="$(echo $GITHUB_REPOSITORY-install-tests | tr "[A-Z]/" "[a-z]-")" >> $GITHUB_ENV
-    - uses: cachix/install-nix-action@v16
+    - uses: cachix/install-nix-action@v17
       with:
         install_url: '${{needs.installer.outputs.installerURL}}'
         install_options: "--tarball-url-prefix https://${{ env.CACHIX_NAME }}.cachix.org/serve"
@@ -83,10 +83,10 @@ jobs:
       needs.check_cachix.outputs.secret == 'true'
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v2.4.0
+    - uses: actions/checkout@v3
       with:
         fetch-depth: 0
-    - uses: cachix/install-nix-action@v16
+    - uses: cachix/install-nix-action@v17
     - run: echo CACHIX_NAME="$(echo $GITHUB_REPOSITORY-install-tests | tr "[A-Z]/" "[a-z]-")" >> $GITHUB_ENV
     - run: echo NIX_VERSION="$(nix-instantiate --eval -E '(import ./default.nix).defaultPackage.${builtins.currentSystem}.version' | tr -d \")" >> $GITHUB_ENV
     - uses: cachix/cachix-action@v10

.github/workflows/hydra_status.yml

@@ -9,7 +9,7 @@ jobs:
     if: github.repository_owner == 'NixOS'
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v2.4.0
+    - uses: actions/checkout@v3
       with:
         fetch-depth: 0
     - run: bash scripts/check-hydra-status.sh

.gitignore

@@ -79,6 +79,7 @@ perl/Makefile.config
 /tests/shell.drv
 /tests/config.nix
 /tests/ca/config.nix
+/tests/repl-result-out
 
 # /tests/lang/
 /tests/lang/*.out
@@ -90,6 +91,7 @@ perl/Makefile.config
 
 /misc/systemd/nix-daemon.service
 /misc/systemd/nix-daemon.socket
+/misc/systemd/nix-daemon.conf
 /misc/upstart/nix-daemon.conf
 
 /src/resolve-system-dependencies/resolve-system-dependencies

.version

@@ -1 +1 @@
-2.8.0
+2.9.0

doc/manual/generate-options.nix

@@ -6,7 +6,8 @@ options:
 concatStrings (map
   (name:
     let option = options.${name}; in
-    "  - `${name}`  \n\n"
+    "  - [`${name}`](#conf-${name})"
+    + "<p id=\"conf-${name}\"></p>\n\n"
     + concatStrings (map (s: "    ${s}\n") (splitLines option.description)) + "\n\n"
     + (if option.documentDefault
        then "    **Default:** " + (

doc/manual/src/SUMMARY.md.in

@@ -72,6 +72,7 @@
   - [CLI guideline](contributing/cli-guideline.md)
 - [Release Notes](release-notes/release-notes.md)
   - [Release X.Y (202?-??-??)](release-notes/rl-next.md)
+  - [Release 2.8 (2022-04-19)](release-notes/rl-2.8.md)
   - [Release 2.7 (2022-03-07)](release-notes/rl-2.7.md)
   - [Release 2.6 (2022-01-24)](release-notes/rl-2.6.md)
   - [Release 2.5 (2021-12-13)](release-notes/rl-2.5.md)

doc/manual/src/advanced-topics/distributed-builds.md

@@ -110,7 +110,7 @@ default, set it to `-`.
 7.  A comma-separated list of *mandatory features*. A machine will only
     be used to build a derivation if all of the machine’s mandatory
     features appear in the derivation’s `requiredSystemFeatures`
-    attribute..
+    attribute.
 
 8.  The (base64-encoded) public host key of the remote machine. If omitted, SSH
     will use its regular known-hosts file. Specifically, the field is calculated

doc/manual/src/installation/installing-binary.md

@@ -84,7 +84,9 @@ The installer will modify `/etc/bashrc`, and `/etc/zshrc` if they exist.
 The installer will first back up these files with a `.backup-before-nix`
 extension. The installer will also create `/etc/profile.d/nix.sh`.
 
-You can uninstall Nix with the following commands:
+## Uninstalling
+
+### Linux
 
 ```console
 sudo rm -rf /etc/profile/nix.sh /etc/nix /nix ~root/.nix-profile ~root/.nix-defexpr ~root/.nix-channels ~/.nix-profile ~/.nix-defexpr ~/.nix-channels
@@ -95,15 +97,95 @@ sudo systemctl stop nix-daemon.service
 sudo systemctl disable nix-daemon.socket
 sudo systemctl disable nix-daemon.service
 sudo systemctl daemon-reload
-
-# If you are on macOS, you will need to run:
-sudo launchctl unload /Library/LaunchDaemons/org.nixos.nix-daemon.plist
-sudo rm /Library/LaunchDaemons/org.nixos.nix-daemon.plist
 ```
 
 There may also be references to Nix in `/etc/profile`, `/etc/bashrc`,
 and `/etc/zshrc` which you may remove.
 
+### macOS
+
+1. Edit `/etc/zshrc` and `/etc/bashrc` to remove the lines sourcing
+   `nix-daemon.sh`, which should look like this:
+
+   ```bash
+   # Nix
+   if [ -e '/nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh' ]; then
+     . '/nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh'
+   fi
+   # End Nix
+   ```
+
+   If these files haven't been altered since installing Nix you can simply put
+   the backups back in place:
+
+   ```console
+   sudo mv /etc/zshrc.backup-before-nix /etc/zshrc
+   sudo mv /etc/bashrc.backup-before-nix /etc/bashrc
+   ```
+
+   This will stop shells from sourcing the file and bringing everything you
+   installed using Nix in scope.
+
+2. Stop and remove the Nix daemon services:
+
+   ```console
+   sudo launchctl unload /Library/LaunchDaemons/org.nixos.nix-daemon.plist
+   sudo rm /Library/LaunchDaemons/org.nixos.nix-daemon.plist
+   sudo launchctl unload /Library/LaunchDaemons/org.nixos.darwin-store.plist
+   sudo rm /Library/LaunchDaemons/org.nixos.darwin-store.plist
+   ```
+
+   This stops the Nix daemon and prevents it from being started next time you
+   boot the system.
+
+3. Remove the `nixbld` group and the `_nixbuildN` users:
+
+   ```console
+   sudo dscl . -delete /Groups/nixbld
+   for u in $(sudo dscl . -list /Users | grep _nixbld); do sudo dscl . -delete /Users/$u; done
+   ```
+
+   This will remove all the build users that no longer serve a purpose.
+
+4. Edit fstab using `sudo vifs` to remove the line mounting the Nix Store
+   volume on `/nix`, which looks like this,
+   `LABEL=Nix\040Store /nix apfs rw,nobrowse`. This will prevent automatic
+   mounting of the Nix Store volume.
+
+5. Edit `/etc/synthetic.conf` to remove the `nix` line. If this is the only
+   line in the file you can remove it entirely, `sudo rm /etc/synthetic.conf`.
+   This will prevent the creation of the empty `/nix` directory to provide a
+   mountpoint for the Nix Store volume.
+
+6. Remove the files Nix added to your system:
+
+   ```console
+   sudo rm -rf /etc/nix /var/root/.nix-profile /var/root/.nix-defexpr /var/root/.nix-channels ~/.nix-profile ~/.nix-defexpr ~/.nix-channels
+   ```
+
+   This gets rid of any data Nix may have created except for the store which is
+   removed next.
+
+7. Remove the Nix Store volume:
+   
+   ```console
+   sudo diskutil apfs deleteVolume /nix
+   ```
+
+   This will remove the Nix Store volume and everything that was added to the
+   store.
+
+> **Note**
+> 
+> After you complete the steps here, you will still have an empty `/nix`
+> directory. This is an expected sign of a successful uninstall. The empty
+> `/nix` directory will disappear the next time you reboot.
+>
+> You do not have to reboot to finish uninstalling Nix. The uninstall is
+> complete. macOS (Catalina+) directly controls root directories and its
+> read-only root will prevent you from manually deleting the empty `/nix`
+> mountpoint.
+
 # macOS Installation <a name="sect-macos-installation-change-store-prefix"></a><a name="sect-macos-installation-encrypted-volume"></a><a name="sect-macos-installation-symlink"></a><a name="sect-macos-installation-recommended-notes"></a>
 <!-- Note: anchors above to catch permalinks to old explanations -->
 

doc/manual/src/release-notes/rl-2.7.md

@@ -1,4 +1,4 @@
-# Release X.Y (2022-03-07)
+# Release 2.7 (2022-03-07)
 
 * Nix will now make some helpful suggestions when you mistype
   something on the command line. For instance, if you type `nix build

doc/manual/src/release-notes/rl-2.8.md

@@ -0,0 +1,53 @@
+# Release 2.8 (2022-04-19)
+
+* New experimental command: `nix fmt`, which applies a formatter
+  defined by the `formatter.<system>` flake output to the Nix
+  expressions in a flake.
+
+* Various Nix commands can now read expressions from standard input
+  using `--file -`.
+
+* New experimental builtin function `builtins.fetchClosure` that
+  copies a closure from a binary cache at evaluation time and rewrites
+  it to content-addressed form (if it isn't already). Like
+  `builtins.storePath`, this allows importing pre-built store paths;
+  the difference is that it doesn't require the user to configure
+  binary caches and trusted public keys.
+
+  This function is only available if you enable the experimental
+  feature `fetch-closure`.
+
+* New experimental feature: *impure derivations*. These are
+  derivations that can produce a different result every time they're
+  built. Here is an example:
+
+  ```nix
+  stdenv.mkDerivation {
+    name = "impure";
+    __impure = true; # marks this derivation as impure
+    buildCommand = "date > $out";
+  }
+  ```
+
+  Running `nix build` twice on this expression will build the
+  derivation twice, producing two different content-addressed store
+  paths. Like fixed-output derivations, impure derivations have access
+  to the network. Only fixed-output derivations and impure derivations
+  can depend on an impure derivation.
+
+* `nix store make-content-addressable` has been renamed to `nix store
+  make-content-addressed`.
+
+* The `nixosModule` flake output attribute has been renamed consistent
+  with the `.default` renames in Nix 2.7.
+
+  * `nixosModule` → `nixosModules.default`
+
+  As before, the old output will continue to work, but `nix flake check` will
+  issue a warning about it.
+
+* `nix run` is now stricter in what it accepts: members of the `apps`
+  flake output are now required to be apps (as defined in [the
+  manual](https://nixos.org/manual/nix/stable/command-ref/new-cli/nix3-run.html#apps)),
+  and members of `packages` or `legacyPackages` must be derivations
+  (not apps).

doc/manual/src/release-notes/rl-next.md

@@ -1,4 +1,14 @@
 # Release X.Y (202?-??-??)
 
+* `nix repl` has a new build-'n-link (`:bl`) command that builds a derivation
+  while creating GC root symlinks.
+
+* The path produced by `builtins.toFile` is now allowed to be imported or read
+  even with restricted evaluation. Note that this will not work with a
+  read-only store.
+
+* `nix build` has a new `--print-out-paths` flag to print the resulting output paths.
+  This matches the default behaviour of `nix-build`.
+
 * Nix can now be built with LTO by passing `--enable-lto` to `configure`.
-  LTO is currently only supported when building with GCC.
+  LTO is currently only supported when building with GCC.

docker.nix

@@ -22,6 +22,7 @@ let
     findutils
     iana-etc
     git
+    openssh
   ];
 
   users = {

flake.lock

@@ -18,11 +18,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1632864508,
+        "lastModified": 1645296114,
-        "narHash": "sha256-d127FIvGR41XbVRDPVvozUPQ/uRHbHwvfyKHwEt5xFM=",
+        "narHash": "sha256-y53N7TyIkXsjMpOG7RhvqJFGDacLs9HlyHeSTBioqYU=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "82891b5e2c2359d7e58d08849e4c89511ab94234",
+        "rev": "530a53dcbc9437363471167a5e4762c5fcfa34a1",
         "type": "github"
       },
       "original": {

misc/bash/completion.sh

@@ -15,7 +15,7 @@ function _complete_nix {
         else
             COMPREPLY+=("$completion")
         fi
-    done < <(NIX_GET_COMPLETIONS=$cword "${words[@]/#\~/$HOME}" 2>/dev/null)
+    done < <(NIX_GET_COMPLETIONS=$cword "${words[@]}" 2>/dev/null)
     __ltrim_colon_completions "$cur"
 }
 

misc/systemd/local.mk

@@ -1,7 +1,8 @@
 ifdef HOST_LINUX
 
   $(foreach n, nix-daemon.socket nix-daemon.service, $(eval $(call install-file-in, $(d)/$(n), $(prefix)/lib/systemd/system, 0644)))
+  $(foreach n, nix-daemon.conf, $(eval $(call install-file-in, $(d)/$(n), $(prefix)/lib/tmpfiles.d, 0644)))
 
-  clean-files += $(d)/nix-daemon.socket $(d)/nix-daemon.service
+  clean-files += $(d)/nix-daemon.socket $(d)/nix-daemon.service $(d)/nix-daemon.conf
 
 endif

misc/systemd/nix-daemon.conf.in

@@ -0,0 +1 @@
+d @localstatedir@/nix/daemon-socket 0755 root root - -

misc/systemd/nix-daemon.service.in

@@ -1,7 +1,9 @@
 [Unit]
 Description=Nix Daemon
+Documentation=man:nix-daemon https://nixos.org/manual
 RequiresMountsFor=@storedir@
 RequiresMountsFor=@localstatedir@
+RequiresMountsFor=@localstatedir@/nix/db
 ConditionPathIsReadWrite=@localstatedir@/nix/daemon-socket
 
 [Service]

scripts/check-hydra-status.sh

@@ -14,7 +14,7 @@ curl -sS -H 'Accept: application/json' https://hydra.nixos.org/jobset/nix/master
 someBuildFailed=0
 
 for buildId in $BUILDS_FOR_LATEST_EVAL; do
-  buildInfo=$(curl -sS -H 'Accept: application/json' "https://hydra.nixos.org/build/$buildId")
+  buildInfo=$(curl --fail -sS -H 'Accept: application/json' "https://hydra.nixos.org/build/$buildId")
 
   finished=$(echo "$buildInfo" | jq -r '.finished')
 

scripts/install-multi-user.sh

@@ -423,6 +423,18 @@ EOF
         fi
     done
 
+    if [ "$(uname -s)" = "Linux" ] && [ ! -e /run/systemd/system ]; then
+        warning <<EOF
+We did not detect systemd on your system. With a multi-user install
+without systemd you will have to manually configure your init system to
+launch the Nix daemon after installation.
+EOF
+        if ! ui_confirm "Do you want to proceed with a multi-user installation?"; then
+            failure <<EOF
+You have aborted the installation.
+EOF
+        fi
+    fi
 }
 
 setup_report() {
@@ -739,7 +751,7 @@ install_from_extracted_nix() {
         cd "$EXTRACTED_NIX_PATH"
 
         _sudo "to copy the basic Nix files to the new store at $NIX_ROOT/store" \
-              cp -RLp ./store/* "$NIX_ROOT/store/"
+              cp -RPp ./store/* "$NIX_ROOT/store/"
 
         _sudo "to make the new store non-writable at $NIX_ROOT/store" \
               chmod -R ugo-w "$NIX_ROOT/store/"

scripts/install-systemd-multi-user.sh

@@ -9,6 +9,8 @@ readonly SERVICE_DEST=/etc/systemd/system/nix-daemon.service
 readonly SOCKET_SRC=/lib/systemd/system/nix-daemon.socket
 readonly SOCKET_DEST=/etc/systemd/system/nix-daemon.socket
 
+readonly TMPFILES_SRC=/lib/tmpfiles.d/nix-daemon.conf
+readonly TMPFILES_DEST=/etc/tmpfiles.d/nix-daemon.conf
 
 # Path for the systemd override unit file to contain the proxy settings
 readonly SERVICE_OVERRIDE=${SERVICE_DEST}.d/override.conf
@@ -83,6 +85,13 @@ EOF
 poly_configure_nix_daemon_service() {
     if [ -e /run/systemd/system ]; then
         task "Setting up the nix-daemon systemd service"
+
+        _sudo "to create the nix-daemon tmpfiles config" \
+              ln -sfn /nix/var/nix/profiles/default/$TMPFILES_SRC $TMPFILES_DEST
+
+        _sudo "to run systemd-tmpfiles once to pick that path up" \
+             systemd-tmpfiles --create --prefix=/nix/var/nix
+
         _sudo "to set up the nix-daemon service" \
               systemctl link "/nix/var/nix/profiles/default$SERVICE_SRC"
 

scripts/install.in

@@ -82,7 +82,7 @@ if [ "$(uname -s)" != "Darwin" ]; then
 fi
 
 if command -v curl > /dev/null 2>&1; then
-    fetch() { curl -L "$1" -o "$2"; }
+    fetch() { curl --fail -L "$1" -o "$2"; }
 elif command -v wget > /dev/null 2>&1; then
     fetch() { wget "$1" -O "$2"; }
 else

src/build-remote/build-remote.cc

@@ -300,7 +300,7 @@ connected:
 
         std::set<Realisation> missingRealisations;
         StorePathSet missingPaths;
-        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations) && !derivationHasKnownOutputPaths(drv.type())) {
+        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations) && !drv.type().hasKnownOutputPaths()) {
             for (auto & outputName : wantedOutputs) {
                 auto thisOutputHash = outputHashes.at(outputName);
                 auto thisOutputId = DrvOutput{ thisOutputHash, outputName };

src/libcmd/command.cc

@@ -197,16 +197,17 @@ void StorePathCommand::run(ref<Store> store, std::vector<StorePath> && storePath
     run(store, *storePaths.begin());
 }
 
-Strings editorFor(const Pos & pos)
+Strings editorFor(const Path & file, uint32_t line)
 {
     auto editor = getEnv("EDITOR").value_or("cat");
     auto args = tokenizeString<Strings>(editor);
-    if (pos.line > 0 && (
+    if (line > 0 && (
         editor.find("emacs") != std::string::npos ||
         editor.find("nano") != std::string::npos ||
-        editor.find("vim") != std::string::npos))
+        editor.find("vim") != std::string::npos ||
-        args.push_back(fmt("+%d", pos.line));
+        editor.find("kak") != std::string::npos))
-    args.push_back(pos.file);
+        args.push_back(fmt("+%d", line));
+    args.push_back(file);
     return args;
 }
 

src/libcmd/command.hh

@@ -85,11 +85,12 @@ struct SourceExprCommand : virtual Args, MixFlakeOptions
 {
     std::optional<Path> file;
     std::optional<std::string> expr;
+    bool readOnlyMode = false;
 
     // FIXME: move this; not all commands (e.g. 'nix run') use it.
     OperateOn operateOn = OperateOn::Output;
 
-    SourceExprCommand();
+    SourceExprCommand(bool supportReadOnlyMode = false);
 
     std::vector<std::shared_ptr<Installable>> parseInstallables(
         ref<Store> store, std::vector<std::string> ss);
@@ -128,13 +129,13 @@ struct InstallableCommand : virtual Args, SourceExprCommand
 {
     std::shared_ptr<Installable> installable;
 
-    InstallableCommand();
+    InstallableCommand(bool supportReadOnlyMode = false);
 
     void prepare() override;
 
     std::optional<FlakeRef> getFlakeRefForCompletion() override
     {
-        return parseFlakeRef(_installable, absPath("."));
+        return parseFlakeRefWithFragment(_installable, absPath(".")).first;
     }
 
 private:
@@ -218,7 +219,7 @@ static RegisterCommand registerCommand2(std::vector<std::string> && name)
 
 /* Helper function to generate args that invoke $EDITOR on
    filename:lineno. */
-Strings editorFor(const Pos & pos);
+Strings editorFor(const Path & file, uint32_t line);
 
 struct MixProfile : virtual StoreCommand
 {

src/libcmd/common-eval-args.cc

@@ -7,6 +7,7 @@
 #include "registry.hh"
 #include "flake/flakeref.hh"
 #include "store-api.hh"
+#include "command.hh"
 
 namespace nix {
 
@@ -59,6 +60,9 @@ MixEvalArgs::MixEvalArgs()
             fetchers::Attrs extraAttrs;
             if (to.subdir != "") extraAttrs["dir"] = to.subdir;
             fetchers::overrideRegistry(from.input, to.input, extraAttrs);
+        }},
+        .completer = {[&](size_t, std::string_view prefix) {
+            completeFlakeRef(openStore(), prefix);
         }}
     });
 

src/libcmd/installables.cc

@@ -1,4 +1,6 @@
+#include "globals.hh"
 #include "installables.hh"
+#include "util.hh"
 #include "command.hh"
 #include "attr-path.hh"
 #include "common-eval-args.hh"
@@ -99,6 +101,14 @@ MixFlakeOptions::MixFlakeOptions()
             lockFlags.inputOverrides.insert_or_assign(
                 flake::parseInputPath(inputPath),
                 parseFlakeRef(flakeRef, absPath("."), true));
+        }},
+        .completer = {[&](size_t n, std::string_view prefix) {
+            if (n == 0) {
+                if (auto flakeRef = getFlakeRefForCompletion())
+                    completeFlakeInputPath(getEvalState(), *flakeRef, prefix);
+            } else if (n == 1) {
+                completeFlakeRef(getEvalState()->store, prefix);
+            }
         }}
     });
 
@@ -129,12 +139,14 @@ MixFlakeOptions::MixFlakeOptions()
     });
 }
 
-SourceExprCommand::SourceExprCommand()
+SourceExprCommand::SourceExprCommand(bool supportReadOnlyMode)
 {
     addFlag({
         .longName = "file",
         .shortName = 'f',
-        .description = "Interpret installables as attribute paths relative to the Nix expression stored in *file*.",
+        .description =
+            "Interpret installables as attribute paths relative to the Nix expression stored in *file*. "
+            "If *file* is the character -, then a Nix expression will be read from standard input.",
         .category = installablesCategory,
         .labels = {"file"},
         .handler = {&file},
@@ -155,6 +167,17 @@ SourceExprCommand::SourceExprCommand()
         .category = installablesCategory,
         .handler = {&operateOn, OperateOn::Derivation},
     });
+
+    if (supportReadOnlyMode) {
+        addFlag({
+            .longName = "read-only",
+            .description =
+                "Do not instantiate each evaluated derivation. "
+                "This improves performance, but can cause errors when accessing "
+                "store paths of derivations during evaluation.",
+            .handler = {&readOnlyMode, true},
+        });
+    }
 }
 
 Strings SourceExprCommand::getDefaultFlakeAttrPaths()
@@ -180,6 +203,8 @@ Strings SourceExprCommand::getDefaultFlakeAttrPathPrefixes()
 void SourceExprCommand::completeInstallable(std::string_view prefix)
 {
     if (file) {
+        completionType = ctAttrs;
+
         evalSettings.pureEval = false;
         auto state = getEvalState();
         Expr *e = state->parseExprFromFile(
@@ -208,13 +233,14 @@ void SourceExprCommand::completeInstallable(std::string_view prefix)
         Value v2;
         state->autoCallFunction(*autoArgs, v1, v2);
 
-        completionType = ctAttrs;
-
         if (v2.type() == nAttrs) {
             for (auto & i : *v2.attrs) {
-                std::string name = i.name;
+                std::string name = state->symbols[i.name];
                 if (name.find(searchWord) == 0) {
-                    completions->add(i.name);
+                    if (prefix_ == "")
+                        completions->add(name);
+                    else
+                        completions->add(prefix_ + "." + name);
                 }
             }
         }
@@ -242,10 +268,11 @@ void completeFlakeRefWithFragment(
         if (hash == std::string::npos) {
             completeFlakeRef(evalState->store, prefix);
         } else {
+            completionType = ctAttrs;
+
             auto fragment = prefix.substr(hash + 1);
             auto flakeRefS = std::string(prefix.substr(0, hash));
-            // FIXME: do tilde expansion.
+            auto flakeRef = parseFlakeRef(expandTilde(flakeRefS), absPath("."));
-            auto flakeRef = parseFlakeRef(flakeRefS, absPath("."));
 
             auto evalCache = openEvalCache(*evalState,
                 std::make_shared<flake::LockedFlake>(lockFlake(*evalState, flakeRef, lockFlags)));
@@ -257,8 +284,6 @@ void completeFlakeRefWithFragment(
                flake. */
             attrPathPrefixes.push_back("");
 
-            completionType = ctAttrs;
-
             for (auto & attrPathPrefixS : attrPathPrefixes) {
                 auto attrPathPrefix = parseAttrPath(*evalState, attrPathPrefixS);
                 auto attrPathS = attrPathPrefixS + std::string(fragment);
@@ -332,16 +357,16 @@ DerivedPath Installable::toDerivedPath()
     return std::move(buildables[0]);
 }
 
-std::vector<std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>>
+std::vector<ref<eval_cache::AttrCursor>>
 Installable::getCursors(EvalState & state)
 {
     auto evalCache =
         std::make_shared<nix::eval_cache::EvalCache>(std::nullopt, state,
             [&]() { return toValue(state).first; });
-    return {{evalCache->getRoot(), ""}};
+    return {evalCache->getRoot()};
 }
 
-std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>
+ref<eval_cache::AttrCursor>
 Installable::getCursor(EvalState & state)
 {
     auto cursors = getCursors(state);
@@ -448,7 +473,7 @@ struct InstallableAttrPath : InstallableValue
 
     std::string what() const override { return attrPath; }
 
-    std::pair<Value *, Pos> toValue(EvalState & state) override
+    std::pair<Value *, PosIdx> toValue(EvalState & state) override
     {
         auto [vRes, pos] = findAlongAttrPath(state, attrPath, *cmd.getAutoArgs(state), **v);
         state.forceValue(*vRes, pos);
@@ -564,43 +589,21 @@ InstallableFlake::InstallableFlake(
 
 std::tuple<std::string, FlakeRef, InstallableValue::DerivationInfo> InstallableFlake::toDerivation()
 {
-    auto lockedFlake = getLockedFlake();
+    auto attr = getCursor(*state);
 
-    auto cache = openEvalCache(*state, lockedFlake);
+    auto attrPath = attr->getAttrPathStr();
-    auto root = cache->getRoot();
 
-    Suggestions suggestions;
+    if (!attr->isDerivation())
+        throw Error("flake output attribute '%s' is not a derivation", attrPath);
 
-    for (auto & attrPath : getActualAttrPaths()) {
+    auto drvPath = attr->forceDerivation();
-        debug("trying flake output attribute '%s'", attrPath);
 
-        auto attrOrSuggestions = root->findAlongAttrPath(
+    auto drvInfo = DerivationInfo {
-            parseAttrPath(*state, attrPath),
+        std::move(drvPath),
-            true
+        attr->getAttr("outputName")->getString()
-        );
+    };
 
-        if (!attrOrSuggestions) {
+    return {attrPath, getLockedFlake()->flake.lockedRef, std::move(drvInfo)};
-            suggestions += attrOrSuggestions.getSuggestions();
-            continue;
-        }
-
-        auto attr = *attrOrSuggestions;
-
-        if (!attr->isDerivation())
-            throw Error("flake output attribute '%s' is not a derivation", attrPath);
-
-        auto drvPath = attr->forceDerivation();
-
-        auto drvInfo = DerivationInfo {
-            std::move(drvPath),
-            attr->getAttr(state->sOutputName)->getString()
-        };
-
-        return {attrPath, lockedFlake->flake.lockedRef, std::move(drvInfo)};
-    }
-
-    throw Error(suggestions, "flake '%s' does not provide attribute %s",
-        flakeRef, showAttrPaths(getActualAttrPaths()));
 }
 
 std::vector<InstallableValue::DerivationInfo> InstallableFlake::toDerivations()
@@ -610,35 +613,12 @@ std::vector<InstallableValue::DerivationInfo> InstallableFlake::toDerivations()
     return res;
 }
 
-std::pair<Value *, Pos> InstallableFlake::toValue(EvalState & state)
+std::pair<Value *, PosIdx> InstallableFlake::toValue(EvalState & state)
 {
-    auto lockedFlake = getLockedFlake();
+    return {&getCursor(state)->forceValue(), noPos};
-
-    auto vOutputs = getFlakeOutputs(state, *lockedFlake);
-
-    auto emptyArgs = state.allocBindings(0);
-
-    Suggestions suggestions;
-
-    for (auto & attrPath : getActualAttrPaths()) {
-        try {
-            auto [v, pos] = findAlongAttrPath(state, attrPath, *emptyArgs, *vOutputs);
-            state.forceValue(*v, pos);
-            return {v, pos};
-        } catch (AttrPathNotFound & e) {
-            suggestions += e.info().suggestions;
-        }
-    }
-
-    throw Error(
-        suggestions,
-        "flake '%s' does not provide attribute %s",
-        flakeRef,
-        showAttrPaths(getActualAttrPaths())
-    );
 }
 
-std::vector<std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>>
+std::vector<ref<eval_cache::AttrCursor>>
 InstallableFlake::getCursors(EvalState & state)
 {
     auto evalCache = openEvalCache(state,
@@ -646,21 +626,55 @@ InstallableFlake::getCursors(EvalState & state)
 
     auto root = evalCache->getRoot();
 
-    std::vector<std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>> res;
+    std::vector<ref<eval_cache::AttrCursor>> res;
 
     for (auto & attrPath : getActualAttrPaths()) {
         auto attr = root->findAlongAttrPath(parseAttrPath(state, attrPath));
-        if (attr) res.push_back({*attr, attrPath});
+        if (attr) res.push_back(ref(*attr));
     }
 
     return res;
 }
 
+ref<eval_cache::AttrCursor> InstallableFlake::getCursor(EvalState & state)
+{
+    auto lockedFlake = getLockedFlake();
+
+    auto cache = openEvalCache(state, lockedFlake);
+    auto root = cache->getRoot();
+
+    Suggestions suggestions;
+
+    auto attrPaths = getActualAttrPaths();
+
+    for (auto & attrPath : attrPaths) {
+        debug("trying flake output attribute '%s'", attrPath);
+
+        auto attrOrSuggestions = root->findAlongAttrPath(
+            parseAttrPath(state, attrPath),
+            true
+        );
+
+        if (!attrOrSuggestions) {
+            suggestions += attrOrSuggestions.getSuggestions();
+            continue;
+        }
+
+        return *attrOrSuggestions;
+    }
+
+    throw Error(
+        suggestions,
+        "flake '%s' does not provide attribute %s",
+        flakeRef,
+        showAttrPaths(attrPaths));
+}
+
 std::shared_ptr<flake::LockedFlake> InstallableFlake::getLockedFlake() const
 {
-    flake::LockFlags lockFlagsApplyConfig = lockFlags;
-    lockFlagsApplyConfig.applyNixConfig = true;
     if (!_lockedFlake) {
+        flake::LockFlags lockFlagsApplyConfig = lockFlags;
+        lockFlagsApplyConfig.applyNixConfig = true;
         _lockedFlake = std::make_shared<flake::LockedFlake>(lockFlake(*state, flakeRef, lockFlagsApplyConfig));
     }
     return _lockedFlake;
@@ -685,6 +699,10 @@ std::vector<std::shared_ptr<Installable>> SourceExprCommand::parseInstallables(
 {
     std::vector<std::shared_ptr<Installable>> result;
 
+    if (readOnlyMode) {
+        settings.readOnlyMode = true;
+    }
+
     if (file || expr) {
         if (file && expr)
             throw UsageError("'--file' and '--expr' are exclusive");
@@ -695,7 +713,10 @@ std::vector<std::shared_ptr<Installable>> SourceExprCommand::parseInstallables(
         auto state = getEvalState();
         auto vFile = state->allocValue();
 
-        if (file)
+        if (file == "-") {
+            auto e = state->parseStdin();
+            state->eval(e, *vFile);
+        } else if (file)
             state->evalFile(lookupFileArg(*state, *file), *vFile);
         else {
             auto e = state->parseExprFromString(*expr, absPath("."));
@@ -751,55 +772,20 @@ std::shared_ptr<Installable> SourceExprCommand::parseInstallable(
     return installables.front();
 }
 
-BuiltPaths getBuiltPaths(ref<Store> evalStore, ref<Store> store, const DerivedPaths & hopefullyBuiltPaths)
+BuiltPaths Installable::build(
+    ref<Store> evalStore,
+    ref<Store> store,
+    Realise mode,
+    const std::vector<std::shared_ptr<Installable>> & installables,
+    BuildMode bMode)
 {
     BuiltPaths res;
-    for (const auto & b : hopefullyBuiltPaths)
+    for (auto & [_, builtPath] : build2(evalStore, store, mode, installables, bMode))
-        std::visit(
+        res.push_back(builtPath);
-            overloaded{
-                [&](const DerivedPath::Opaque & bo) {
-                    res.push_back(BuiltPath::Opaque{bo.path});
-                },
-                [&](const DerivedPath::Built & bfd) {
-                    OutputPathMap outputs;
-                    auto drv = evalStore->readDerivation(bfd.drvPath);
-                    auto outputHashes = staticOutputHashes(*evalStore, drv); // FIXME: expensive
-                    auto drvOutputs = drv.outputsAndOptPaths(*store);
-                    for (auto & output : bfd.outputs) {
-                        if (!outputHashes.count(output))
-                            throw Error(
-                                "the derivation '%s' doesn't have an output named '%s'",
-                                store->printStorePath(bfd.drvPath), output);
-                        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations)) {
-                            auto outputId =
-                                DrvOutput{outputHashes.at(output), output};
-                            auto realisation =
-                                store->queryRealisation(outputId);
-                            if (!realisation)
-                                throw Error(
-                                    "cannot operate on an output of unbuilt "
-                                    "content-addressed derivation '%s'",
-                                    outputId.to_string());
-                            outputs.insert_or_assign(
-                                output, realisation->outPath);
-                        } else {
-                            // If ca-derivations isn't enabled, assume that
-                            // the output path is statically known.
-                            assert(drvOutputs.count(output));
-                            assert(drvOutputs.at(output).second);
-                            outputs.insert_or_assign(
-                                output, *drvOutputs.at(output).second);
-                        }
-                    }
-                    res.push_back(BuiltPath::Built{bfd.drvPath, outputs});
-                },
-            },
-            b.raw());
-
     return res;
 }
 
-BuiltPaths Installable::build(
+std::vector<std::pair<std::shared_ptr<Installable>, BuiltPath>> Installable::build2(
     ref<Store> evalStore,
     ref<Store> store,
     Realise mode,
@@ -810,39 +796,93 @@ BuiltPaths Installable::build(
         settings.readOnlyMode = true;
 
     std::vector<DerivedPath> pathsToBuild;
+    std::map<DerivedPath, std::vector<std::shared_ptr<Installable>>> backmap;
 
     for (auto & i : installables) {
-        auto b = i->toDerivedPaths();
+        for (auto b : i->toDerivedPaths()) {
-        pathsToBuild.insert(pathsToBuild.end(), b.begin(), b.end());
+            pathsToBuild.push_back(b);
+            backmap[b].push_back(i);
+        }
     }
 
+    std::vector<std::pair<std::shared_ptr<Installable>, BuiltPath>> res;
+
     switch (mode) {
+
     case Realise::Nothing:
     case Realise::Derivation:
         printMissing(store, pathsToBuild, lvlError);
-        return getBuiltPaths(evalStore, store, pathsToBuild);
+
+        for (auto & path : pathsToBuild) {
+            for (auto & installable : backmap[path]) {
+                std::visit(overloaded {
+                    [&](const DerivedPath::Built & bfd) {
+                        OutputPathMap outputs;
+                        auto drv = evalStore->readDerivation(bfd.drvPath);
+                        auto outputHashes = staticOutputHashes(*evalStore, drv); // FIXME: expensive
+                        auto drvOutputs = drv.outputsAndOptPaths(*store);
+                        for (auto & output : bfd.outputs) {
+                            if (!outputHashes.count(output))
+                                throw Error(
+                                    "the derivation '%s' doesn't have an output named '%s'",
+                                    store->printStorePath(bfd.drvPath), output);
+                            if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations)) {
+                                DrvOutput outputId { outputHashes.at(output), output };
+                                auto realisation = store->queryRealisation(outputId);
+                                if (!realisation)
+                                    throw Error(
+                                        "cannot operate on an output of the "
+                                        "unbuilt derivation '%s'",
+                                        outputId.to_string());
+                                outputs.insert_or_assign(output, realisation->outPath);
+                            } else {
+                                // If ca-derivations isn't enabled, assume that
+                                // the output path is statically known.
+                                assert(drvOutputs.count(output));
+                                assert(drvOutputs.at(output).second);
+                                outputs.insert_or_assign(
+                                    output, *drvOutputs.at(output).second);
+                            }
+                        }
+                        res.push_back({installable, BuiltPath::Built { bfd.drvPath, outputs }});
+                    },
+                    [&](const DerivedPath::Opaque & bo) {
+                        res.push_back({installable, BuiltPath::Opaque { bo.path }});
+                    },
+                }, path.raw());
+            }
+        }
+
+        break;
+
     case Realise::Outputs: {
-        BuiltPaths res;
         for (auto & buildResult : store->buildPathsWithResults(pathsToBuild, bMode, evalStore)) {
             if (!buildResult.success())
                 buildResult.rethrow();
-            std::visit(overloaded {
+
-                [&](const DerivedPath::Built & bfd) {
+            for (auto & installable : backmap[buildResult.path]) {
-                    std::map<std::string, StorePath> outputs;
+                std::visit(overloaded {
-                    for (auto & path : buildResult.builtOutputs)
+                    [&](const DerivedPath::Built & bfd) {
-                        outputs.emplace(path.first.outputName, path.second.outPath);
+                        std::map<std::string, StorePath> outputs;
-                    res.push_back(BuiltPath::Built { bfd.drvPath, outputs });
+                        for (auto & path : buildResult.builtOutputs)
-                },
+                            outputs.emplace(path.first.outputName, path.second.outPath);
-                [&](const DerivedPath::Opaque & bo) {
+                        res.push_back({installable, BuiltPath::Built { bfd.drvPath, outputs }});
-                    res.push_back(BuiltPath::Opaque { bo.path });
+                    },
-                },
+                    [&](const DerivedPath::Opaque & bo) {
-            }, buildResult.path.raw());
+                        res.push_back({installable, BuiltPath::Opaque { bo.path }});
+                    },
+                }, buildResult.path.raw());
+            }
         }
-        return res;
+
+        break;
     }
+
     default:
         assert(false);
     }
+
+    return res;
 }
 
 BuiltPaths Installable::toBuiltPaths(
@@ -930,7 +970,7 @@ InstallablesCommand::InstallablesCommand()
 void InstallablesCommand::prepare()
 {
     if (_installables.empty() && useDefaultInstallables())
-        // FIXME: commands like "nix install" should not have a
+        // FIXME: commands like "nix profile install" should not have a
         // default, probably.
         _installables.push_back(".");
     installables = parseInstallables(getStore(), _installables);
@@ -940,13 +980,14 @@ std::optional<FlakeRef> InstallablesCommand::getFlakeRefForCompletion()
 {
     if (_installables.empty()) {
         if (useDefaultInstallables())
-            return parseFlakeRef(".", absPath("."));
+            return parseFlakeRefWithFragment(".", absPath(".")).first;
         return {};
     }
-    return parseFlakeRef(_installables.front(), absPath("."));
+    return parseFlakeRefWithFragment(_installables.front(), absPath(".")).first;
 }
 
-InstallableCommand::InstallableCommand()
+InstallableCommand::InstallableCommand(bool supportReadOnlyMode)
+    : SourceExprCommand(supportReadOnlyMode)
 {
     expectArgs({
         .label = "installable",

src/libcmd/installables.hh

@@ -68,7 +68,7 @@ struct Installable
 
     UnresolvedApp toApp(EvalState & state);
 
-    virtual std::pair<Value *, Pos> toValue(EvalState & state)
+    virtual std::pair<Value *, PosIdx> toValue(EvalState & state)
     {
         throw Error("argument '%s' cannot be evaluated", what());
     }
@@ -80,10 +80,10 @@ struct Installable
         return {};
     }
 
-    virtual std::vector<std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>>
+    virtual std::vector<ref<eval_cache::AttrCursor>>
     getCursors(EvalState & state);
 
-    std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>
+    virtual ref<eval_cache::AttrCursor>
     getCursor(EvalState & state);
 
     virtual FlakeRef nixpkgsFlakeRef() const
@@ -98,6 +98,13 @@ struct Installable
         const std::vector<std::shared_ptr<Installable>> & installables,
         BuildMode bMode = bmNormal);
 
+    static std::vector<std::pair<std::shared_ptr<Installable>, BuiltPath>> build2(
+        ref<Store> evalStore,
+        ref<Store> store,
+        Realise mode,
+        const std::vector<std::shared_ptr<Installable>> & installables,
+        BuildMode bMode = bmNormal);
+
     static std::set<StorePath> toStorePaths(
         ref<Store> evalStore,
         ref<Store> store,
@@ -171,11 +178,17 @@ struct InstallableFlake : InstallableValue
 
     std::vector<DerivationInfo> toDerivations() override;
 
-    std::pair<Value *, Pos> toValue(EvalState & state) override;
+    std::pair<Value *, PosIdx> toValue(EvalState & state) override;
 
-    std::vector<std::pair<std::shared_ptr<eval_cache::AttrCursor>, std::string>>
+    /* Get a cursor to every attrpath in getActualAttrPaths() that
+       exists. */
+    std::vector<ref<eval_cache::AttrCursor>>
     getCursors(EvalState & state) override;
 
+    /* Get a cursor to the first attrpath in getActualAttrPaths() that
+       exists, or throw an exception with suggestions if none exists. */
+    ref<eval_cache::AttrCursor> getCursor(EvalState & state) override;
+
     std::shared_ptr<flake::LockedFlake> getLockedFlake() const;
 
     FlakeRef nixpkgsFlakeRef() const override;
@@ -185,9 +198,4 @@ ref<eval_cache::EvalCache> openEvalCache(
     EvalState & state,
     std::shared_ptr<flake::LockedFlake> lockedFlake);
 
-BuiltPaths getBuiltPaths(
-    ref<Store> evalStore,
-    ref<Store> store,
-    const DerivedPaths & hopefullyBuiltPaths);
-
 }

src/libexpr/attr-path.cc

@@ -36,18 +36,18 @@ std::vector<Symbol> parseAttrPath(EvalState & state, std::string_view s)
 {
     std::vector<Symbol> res;
     for (auto & a : parseAttrPath(s))
-        res.push_back(state.symbols.create(a));
+        res.emplace_back(a);
     return res;
 }
 
 
-std::pair<Value *, Pos> findAlongAttrPath(EvalState & state, const std::string & attrPath,
+std::pair<Value *, PosIdx> findAlongAttrPath(EvalState & state, const std::string & attrPath,
     Bindings & autoArgs, Value & vIn)
 {
     Strings tokens = parseAttrPath(attrPath);
 
     Value * v = &vIn;
-    Pos pos = noPos;
+    PosIdx pos = noPos;
 
     for (auto & attr : tokens) {
 
@@ -77,13 +77,13 @@ std::pair<Value *, Pos> findAlongAttrPath(EvalState & state, const std::string &
             if (a == v->attrs->end()) {
                 std::set<std::string> attrNames;
                 for (auto & attr : *v->attrs)
-                    attrNames.insert(attr.name);
+                    attrNames.insert(state.symbols[attr.name]);
 
                 auto suggestions = Suggestions::bestMatches(attrNames, attr);
                 throw AttrPathNotFound(suggestions, "attribute '%1%' in selection path '%2%' not found", attr, attrPath);
             }
             v = &*a->value;
-            pos = *a->pos;
+            pos = a->pos;
         }
 
         else {
@@ -106,7 +106,7 @@ std::pair<Value *, Pos> findAlongAttrPath(EvalState & state, const std::string &
 }
 
 
-Pos findPackageFilename(EvalState & state, Value & v, std::string what)
+std::pair<std::string, uint32_t> findPackageFilename(EvalState & state, Value & v, std::string what)
 {
     Value * v2;
     try {
@@ -132,9 +132,7 @@ Pos findPackageFilename(EvalState & state, Value & v, std::string what)
         throw ParseError("cannot parse line number '%s'", pos);
     }
 
-    Symbol file = state.symbols.create(filename);
+    return { std::move(filename), lineno };
-
-    return { foFile, file, lineno, 0 };
 }
 
 

src/libexpr/attr-path.hh

@@ -10,14 +10,14 @@ namespace nix {
 MakeError(AttrPathNotFound, Error);
 MakeError(NoPositionInfo, Error);
 
-std::pair<Value *, Pos> findAlongAttrPath(
+std::pair<Value *, PosIdx> findAlongAttrPath(
     EvalState & state,
     const std::string & attrPath,
     Bindings & autoArgs,
     Value & vIn);
 
 /* Heuristic to find the filename and lineno or a nix value. */
-Pos findPackageFilename(EvalState & state, Value & v, std::string what);
+std::pair<std::string, uint32_t> findPackageFilename(EvalState & state, Value & v, std::string what);
 
 std::vector<Symbol> parseAttrPath(EvalState & state, std::string_view s);
 

src/libexpr/attr-set.cc

@@ -26,7 +26,7 @@ Bindings * EvalState::allocBindings(size_t capacity)
 /* Create a new attribute named 'name' on an existing attribute set stored
    in 'vAttrs' and return the newly allocated Value which is associated with
    this attribute. */
-Value * EvalState::allocAttr(Value & vAttrs, const Symbol & name)
+Value * EvalState::allocAttr(Value & vAttrs, const SymbolIdx & name)
 {
     Value * v = allocValue();
     vAttrs.attrs->push_back(Attr(name, v));
@@ -40,7 +40,7 @@ Value * EvalState::allocAttr(Value & vAttrs, std::string_view name)
 }
 
 
-Value & BindingsBuilder::alloc(const Symbol & name, ptr<Pos> pos)
+Value & BindingsBuilder::alloc(const SymbolIdx & name, PosIdx pos)
 {
     auto value = state.allocValue();
     bindings->push_back(Attr(name, value, pos));
@@ -48,7 +48,7 @@ Value & BindingsBuilder::alloc(const Symbol & name, ptr<Pos> pos)
 }
 
 
-Value & BindingsBuilder::alloc(std::string_view name, ptr<Pos> pos)
+Value & BindingsBuilder::alloc(std::string_view name, PosIdx pos)
 {
     return alloc(state.symbols.create(name), pos);
 }

src/libexpr/attr-set.hh

@@ -15,18 +15,27 @@ struct Value;
 /* Map one attribute name to its value. */
 struct Attr
 {
-    Symbol name;
+    /* the placement of `name` and `pos` in this struct is important.
+       both of them are uint32 wrappers, they are next to each other
+       to make sure that Attr has no padding on 64 bit machines. that
+       way we keep Attr size at two words with no wasted space. */
+    SymbolIdx name;
+    PosIdx pos;
     Value * value;
-    ptr<Pos> pos;
+    Attr(SymbolIdx name, Value * value, PosIdx pos = noPos)
-    Attr(Symbol name, Value * value, ptr<Pos> pos = ptr(&noPos))
+        : name(name), pos(pos), value(value) { };
-        : name(name), value(value), pos(pos) { };
+    Attr() { };
-    Attr() : pos(&noPos) { };
     bool operator < (const Attr & a) const
     {
         return name < a.name;
     }
 };
 
+static_assert(sizeof(Attr) == 2 * sizeof(uint32_t) + sizeof(Value *),
+    "performance of the evaluator is highly sensitive to the size of Attr. "
+    "avoid introducing any padding into Attr if at all possible, and do not "
+    "introduce new fields that need not be present for almost every instance.");
+
 /* Bindings contains all the attributes of an attribute set. It is defined
    by its size and its capacity, the capacity being the number of Attr
    elements allocated after this structure, while the size corresponds to
@@ -35,13 +44,13 @@ class Bindings
 {
 public:
     typedef uint32_t size_t;
-    ptr<Pos> pos;
+    PosIdx pos;
 
 private:
     size_t size_, capacity_;
     Attr attrs[0];
 
-    Bindings(size_t capacity) : pos(&noPos), size_(0), capacity_(capacity) { }
+    Bindings(size_t capacity) : size_(0), capacity_(capacity) { }
     Bindings(const Bindings & bindings) = delete;
 
 public:
@@ -57,7 +66,7 @@ public:
         attrs[size_++] = attr;
     }
 
-    iterator find(const Symbol & name)
+    iterator find(const SymbolIdx & name)
     {
         Attr key(name, 0);
         iterator i = std::lower_bound(begin(), end(), key);
@@ -65,7 +74,7 @@ public:
         return end();
     }
 
-    Attr * get(const Symbol & name)
+    Attr * get(const SymbolIdx & name)
     {
         Attr key(name, 0);
         iterator i = std::lower_bound(begin(), end(), key);
@@ -73,18 +82,6 @@ public:
         return nullptr;
     }
 
-    Attr & need(const Symbol & name, const Pos & pos = noPos)
-    {
-        auto a = get(name);
-        if (!a)
-            throw Error({
-                .msg = hintfmt("attribute '%s' missing", name),
-                .errPos = pos
-            });
-
-        return *a;
-    }
-
     iterator begin() { return &attrs[0]; }
     iterator end() { return &attrs[size_]; }
 
@@ -98,14 +95,15 @@ public:
     size_t capacity() { return capacity_; }
 
     /* Returns the attributes in lexicographically sorted order. */
-    std::vector<const Attr *> lexicographicOrder() const
+    std::vector<const Attr *> lexicographicOrder(const SymbolTable & symbols) const
     {
         std::vector<const Attr *> res;
         res.reserve(size_);
         for (size_t n = 0; n < size_; n++)
             res.emplace_back(&attrs[n]);
-        std::sort(res.begin(), res.end(), [](const Attr * a, const Attr * b) {
+        std::sort(res.begin(), res.end(), [&](const Attr * a, const Attr * b) {
-            return (const std::string &) a->name < (const std::string &) b->name;
+            std::string_view sa = symbols[a->name], sb = symbols[b->name];
+            return sa < sb;
         });
         return res;
     }
@@ -130,7 +128,7 @@ public:
         : bindings(bindings), state(state)
     { }
 
-    void insert(Symbol name, Value * value, ptr<Pos> pos = ptr(&noPos))
+    void insert(SymbolIdx name, Value * value, PosIdx pos = noPos)
     {
         insert(Attr(name, value, pos));
     }
@@ -145,9 +143,9 @@ public:
         bindings->push_back(attr);
     }
 
-    Value & alloc(const Symbol & name, ptr<Pos> pos = ptr(&noPos));
+    Value & alloc(const SymbolIdx & name, PosIdx pos = noPos);
 
-    Value & alloc(std::string_view name, ptr<Pos> pos = ptr(&noPos));
+    Value & alloc(std::string_view name, PosIdx pos = noPos);
 
     Bindings * finish()
     {

src/libexpr/eval-cache.cc

@@ -21,6 +21,8 @@ struct AttrDb
 {
     std::atomic_bool failed{false};
 
+    const Store & cfg;
+
     struct State
     {
         SQLite db;
@@ -33,8 +35,9 @@ struct AttrDb
 
     std::unique_ptr<Sync<State>> _state;
 
-    AttrDb(const Hash & fingerprint)
+    AttrDb(const Store & cfg, const Hash & fingerprint)
-        : _state(std::make_unique<Sync<State>>())
+        : cfg(cfg)
+        , _state(std::make_unique<Sync<State>>())
     {
         auto state(_state->lock());
 
@@ -250,14 +253,14 @@ struct AttrDb
                 std::vector<Symbol> attrs;
                 auto queryAttributes(state->queryAttributes.use()(rowId));
                 while (queryAttributes.next())
-                    attrs.push_back(symbols.create(queryAttributes.getStr(0)));
+                    attrs.emplace_back(queryAttributes.getStr(0));
                 return {{rowId, attrs}};
             }
             case AttrType::String: {
-                std::vector<std::pair<Path, std::string>> context;
+                NixStringContext context;
                 if (!queryAttribute.isNull(3))
                     for (auto & s : tokenizeString<std::vector<std::string>>(queryAttribute.getStr(3), ";"))
-                        context.push_back(decodeContext(s));
+                        context.push_back(decodeContext(cfg, s));
                 return {{rowId, string_t{queryAttribute.getStr(2), context}}};
             }
             case AttrType::Bool:
@@ -274,10 +277,10 @@ struct AttrDb
     }
 };
 
-static std::shared_ptr<AttrDb> makeAttrDb(const Hash & fingerprint)
+static std::shared_ptr<AttrDb> makeAttrDb(const Store & cfg, const Hash & fingerprint)
 {
     try {
-        return std::make_shared<AttrDb>(fingerprint);
+        return std::make_shared<AttrDb>(cfg, fingerprint);
     } catch (SQLiteError &) {
         ignoreException();
         return nullptr;
@@ -288,7 +291,7 @@ EvalCache::EvalCache(
     std::optional<std::reference_wrapper<const Hash>> useCache,
     EvalState & state,
     RootLoader rootLoader)
-    : db(useCache ? makeAttrDb(*useCache) : nullptr)
+    : db(useCache ? makeAttrDb(*state.store, *useCache) : nullptr)
     , state(state)
     , rootLoader(rootLoader)
 {
@@ -303,9 +306,9 @@ Value * EvalCache::getRootValue()
     return *value;
 }
 
-std::shared_ptr<AttrCursor> EvalCache::getRoot()
+ref<AttrCursor> EvalCache::getRoot()
 {
-    return std::make_shared<AttrCursor>(ref(shared_from_this()), std::nullopt);
+    return make_ref<AttrCursor>(ref(shared_from_this()), std::nullopt);
 }
 
 AttrCursor::AttrCursor(
@@ -322,7 +325,7 @@ AttrCursor::AttrCursor(
 AttrKey AttrCursor::getKey()
 {
     if (!parent)
-        return {0, root->state.sEpsilon};
+        return {0, {""}};
     if (!parent->first->cachedValue) {
         parent->first->cachedValue = root->db->getAttr(
             parent->first->getKey(), root->state.symbols);
@@ -337,7 +340,7 @@ Value & AttrCursor::getValue()
         if (parent) {
             auto & vParent = parent->first->getValue();
             root->state.forceAttrs(vParent, noPos);
-            auto attr = vParent.attrs->get(parent->second);
+            auto attr = vParent.attrs->get(root->state.symbols.create(parent->second));
             if (!attr)
                 throw Error("attribute '%s' is unexpectedly missing", getAttrPathStr());
             _value = allocRootValue(attr->value);
@@ -416,7 +419,7 @@ Suggestions AttrCursor::getSuggestionsForAttr(Symbol name)
     return Suggestions::bestMatches(strAttrNames, name);
 }
 
-std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(Symbol name, bool forceErrors)
+std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(std::string_view name, bool forceErrors)
 {
     if (root->db) {
         if (!cachedValue)
@@ -458,10 +461,10 @@ std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(Symbol name, bool forceErro
 
     for (auto & attr : *v.attrs) {
         if (root->db)
-            root->db->setPlaceholder({cachedValue->first, attr.name});
+            root->db->setPlaceholder({cachedValue->first, root->state.symbols[attr.name]});
     }
 
-    auto attr = v.attrs->get(name);
+    auto attr = v.attrs->get(root->state.symbols.create(name));
 
     if (!attr) {
         if (root->db) {
@@ -483,12 +486,7 @@ std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(Symbol name, bool forceErro
         root, std::make_pair(shared_from_this(), name), attr->value, std::move(cachedValue2));
 }
 
-std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(std::string_view name)
+ref<AttrCursor> AttrCursor::getAttr(std::string_view name, bool forceErrors)
-{
-    return maybeGetAttr(root->state.symbols.create(name));
-}
-
-ref<AttrCursor> AttrCursor::getAttr(Symbol name, bool forceErrors)
 {
     auto p = maybeGetAttr(name, forceErrors);
     if (!p)
@@ -496,11 +494,6 @@ ref<AttrCursor> AttrCursor::getAttr(Symbol name, bool forceErrors)
     return ref(p);
 }
 
-ref<AttrCursor> AttrCursor::getAttr(std::string_view name)
-{
-    return getAttr(root->state.symbols.create(name));
-}
-
 OrSuggestions<ref<AttrCursor>> AttrCursor::findAlongAttrPath(const std::vector<Symbol> & attrPath, bool force)
 {
     auto res = shared_from_this();
@@ -546,7 +539,7 @@ string_t AttrCursor::getStringWithContext()
             if (auto s = std::get_if<string_t>(&cachedValue->second)) {
                 bool valid = true;
                 for (auto & c : s->second) {
-                    if (!root->state.store->isValidPath(root->state.store->parseStorePath(c.first))) {
+                    if (!root->state.store->isValidPath(c.first)) {
                         valid = false;
                         break;
                     }
@@ -563,7 +556,7 @@ string_t AttrCursor::getStringWithContext()
     auto & v = forceValue();
 
     if (v.type() == nString)
-        return {v.string.s, v.getContext()};
+        return {v.string.s, v.getContext(*root->state.store)};
     else if (v.type() == nPath)
         return {v.path, {}};
     else
@@ -613,7 +606,7 @@ std::vector<Symbol> AttrCursor::getAttrs()
 
     std::vector<Symbol> attrs;
     for (auto & attr : *getValue().attrs)
-        attrs.push_back(attr.name);
+        attrs.push_back(root->state.symbols[attr.name]);
     std::sort(attrs.begin(), attrs.end(), [](const Symbol & a, const Symbol & b) {
         return (const std::string &) a < (const std::string &) b;
     });
@@ -632,7 +625,7 @@ bool AttrCursor::isDerivation()
 
 StorePath AttrCursor::forceDerivation()
 {
-    auto aDrvPath = getAttr(root->state.sDrvPath, true);
+    auto aDrvPath = getAttr("drvPath", true);
     auto drvPath = root->state.store->parseStorePath(aDrvPath->getString());
     if (!root->state.store->isValidPath(drvPath) && !settings.readOnlyMode) {
         /* The eval cache contains 'drvPath', but the actual path has

src/libexpr/eval-cache.hh

@@ -33,7 +33,7 @@ public:
         EvalState & state,
         RootLoader rootLoader);
 
-    std::shared_ptr<AttrCursor> getRoot();
+    ref<AttrCursor> getRoot();
 };
 
 enum AttrType {
@@ -52,7 +52,7 @@ struct misc_t {};
 struct failed_t {};
 typedef uint64_t AttrId;
 typedef std::pair<AttrId, Symbol> AttrKey;
-typedef std::pair<std::string, std::vector<std::pair<Path, std::string>>> string_t;
+typedef std::pair<std::string, NixStringContext> string_t;
 
 typedef std::variant<
     std::vector<Symbol>,
@@ -96,14 +96,12 @@ public:
 
     Suggestions getSuggestionsForAttr(Symbol name);
 
-    std::shared_ptr<AttrCursor> maybeGetAttr(Symbol name, bool forceErrors = false);
+    std::shared_ptr<AttrCursor> maybeGetAttr(std::string_view name, bool forceErrors = false);
 
-    std::shared_ptr<AttrCursor> maybeGetAttr(std::string_view name);
+    ref<AttrCursor> getAttr(std::string_view name, bool forceErrors = false);
-
-    ref<AttrCursor> getAttr(Symbol name, bool forceErrors = false);
-
-    ref<AttrCursor> getAttr(std::string_view name);
 
+    /* Get an attribute along a chain of attrsets. Note that this does
+       not auto-call functors or functions. */
     OrSuggestions<ref<AttrCursor>> findAlongAttrPath(const std::vector<Symbol> & attrPath, bool force = false);
 
     std::string getString();

src/libexpr/eval-inline.hh

@@ -2,29 +2,85 @@
 
 #include "eval.hh"
 
-#define LocalNoInline(f) static f __attribute__((noinline)); f
-#define LocalNoInlineNoReturn(f) static f __attribute__((noinline, noreturn)); f
-
 namespace nix {
 
-LocalNoInlineNoReturn(void throwEvalError(const Pos & pos, const char * s))
-{
-    throw EvalError({
-        .msg = hintfmt(s),
-        .errPos = pos
-    });
-}
 
-LocalNoInlineNoReturn(void throwTypeError(const Pos & pos, const char * s, const Value & v))
+/* Note: Various places expect the allocated memory to be zeroed. */
+[[gnu::always_inline]]
+inline void * allocBytes(size_t n)
 {
-    throw TypeError({
+    void * p;
-        .msg = hintfmt(s, showType(v)),
+#if HAVE_BOEHMGC
-        .errPos = pos
+    p = GC_MALLOC(n);
-    });
+#else
+    p = calloc(n, 1);
+#endif
+    if (!p) throw std::bad_alloc();
+    return p;
 }
 
 
-void EvalState::forceValue(Value & v, const Pos & pos)
+[[gnu::always_inline]]
+Value * EvalState::allocValue()
+{
+#if HAVE_BOEHMGC
+    /* We use the boehm batch allocator to speed up allocations of Values (of which there are many).
+       GC_malloc_many returns a linked list of objects of the given size, where the first word
+       of each object is also the pointer to the next object in the list. This also means that we
+       have to explicitly clear the first word of every object we take. */
+    if (!*valueAllocCache) {
+        *valueAllocCache = GC_malloc_many(sizeof(Value));
+        if (!*valueAllocCache) throw std::bad_alloc();
+    }
+
+    /* GC_NEXT is a convenience macro for accessing the first word of an object.
+       Take the first list item, advance the list to the next item, and clear the next pointer. */
+    void * p = *valueAllocCache;
+    *valueAllocCache = GC_NEXT(p);
+    GC_NEXT(p) = nullptr;
+#else
+    void * p = allocBytes(sizeof(Value));
+#endif
+
+    nrValues++;
+    return (Value *) p;
+}
+
+
+[[gnu::always_inline]]
+Env & EvalState::allocEnv(size_t size)
+{
+    nrEnvs++;
+    nrValuesInEnvs += size;
+
+    Env * env;
+
+#if HAVE_BOEHMGC
+    if (size == 1) {
+        /* see allocValue for explanations. */
+        if (!*env1AllocCache) {
+            *env1AllocCache = GC_malloc_many(sizeof(Env) + sizeof(Value *));
+            if (!*env1AllocCache) throw std::bad_alloc();
+        }
+
+        void * p = *env1AllocCache;
+        *env1AllocCache = GC_NEXT(p);
+        GC_NEXT(p) = nullptr;
+        env = (Env *) p;
+    } else
+#endif
+        env = (Env *) allocBytes(sizeof(Env) + size * sizeof(Value *));
+
+    env->type = Env::Plain;
+
+    /* We assume that env->values has been cleared by the allocator; maybeThunk() and lookupVar fromWith expect this. */
+
+    return *env;
+}
+
+
+[[gnu::always_inline]]
+void EvalState::forceValue(Value & v, const PosIdx pos)
 {
     forceValue(v, [&]() { return pos; });
 }
@@ -52,13 +108,15 @@ void EvalState::forceValue(Value & v, Callable getPos)
 }
 
 
-inline void EvalState::forceAttrs(Value & v, const Pos & pos)
+[[gnu::always_inline]]
+inline void EvalState::forceAttrs(Value & v, const PosIdx pos)
 {
     forceAttrs(v, [&]() { return pos; });
 }
 
 
 template <typename Callable>
+[[gnu::always_inline]]
 inline void EvalState::forceAttrs(Value & v, Callable getPos)
 {
     forceValue(v, getPos);
@@ -67,25 +125,13 @@ inline void EvalState::forceAttrs(Value & v, Callable getPos)
 }
 
 
-inline void EvalState::forceList(Value & v, const Pos & pos)
+[[gnu::always_inline]]
+inline void EvalState::forceList(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (!v.isList())
         throwTypeError(pos, "value is %1% while a list was expected", v);
 }
 
-/* Note: Various places expect the allocated memory to be zeroed. */
-inline void * allocBytes(size_t n)
-{
-    void * p;
-#if HAVE_BOEHMGC
-    p = GC_MALLOC(n);
-#else
-    p = calloc(n, 1);
-#endif
-    if (!p) throw std::bad_alloc();
-    return p;
-}
-
 
 }

src/libexpr/eval.cc

@@ -96,20 +96,21 @@ RootValue allocRootValue(Value * v)
 }
 
 
-void printValue(std::ostream & str, std::set<const void *> & seen, const Value & v)
+void Value::print(const SymbolTable & symbols, std::ostream & str,
+    std::set<const void *> * seen) const
 {
     checkInterrupt();
 
-    switch (v.internalType) {
+    switch (internalType) {
     case tInt:
-        str << v.integer;
+        str << integer;
         break;
     case tBool:
-        str << (v.boolean ? "true" : "false");
+        str << (boolean ? "true" : "false");
         break;
     case tString:
         str << "\"";
-        for (const char * i = v.string.s; *i; i++)
+        for (const char * i = string.s; *i; i++)
             if (*i == '\"' || *i == '\\') str << "\\" << *i;
             else if (*i == '\n') str << "\\n";
             else if (*i == '\r') str << "\\r";
@@ -119,19 +120,19 @@ void printValue(std::ostream & str, std::set<const void *> & seen, const Value &
         str << "\"";
         break;
     case tPath:
-        str << v.path; // !!! escaping?
+        str << path; // !!! escaping?
         break;
     case tNull:
         str << "null";
         break;
     case tAttrs: {
-        if (!v.attrs->empty() && !seen.insert(v.attrs).second)
+        if (seen && !attrs->empty() && !seen->insert(attrs).second)
-            str << "<REPEAT>";
+            str << "«repeated»";
         else {
             str << "{ ";
-            for (auto & i : v.attrs->lexicographicOrder()) {
+            for (auto & i : attrs->lexicographicOrder(symbols)) {
-                str << i->name << " = ";
+                str << symbols[i->name] << " = ";
-                printValue(str, seen, *i->value);
+                i->value->print(symbols, str, seen);
                 str << "; ";
             }
             str << "}";
@@ -141,12 +142,12 @@ void printValue(std::ostream & str, std::set<const void *> & seen, const Value &
     case tList1:
     case tList2:
     case tListN:
-        if (v.listSize() && !seen.insert(v.listElems()).second)
+        if (seen && listSize() && !seen->insert(listElems()).second)
-            str << "<REPEAT>";
+            str << "«repeated»";
         else {
             str << "[ ";
-            for (auto v2 : v.listItems()) {
+            for (auto v2 : listItems()) {
-                printValue(str, seen, *v2);
+                v2->print(symbols, str, seen);
                 str << " ";
             }
             str << "]";
@@ -166,10 +167,10 @@ void printValue(std::ostream & str, std::set<const void *> & seen, const Value &
         str << "<PRIMOP-APP>";
         break;
     case tExternal:
-        str << *v.external;
+        str << *external;
         break;
     case tFloat:
-        str << v.fpoint;
+        str << fpoint;
         break;
     default:
         abort();
@@ -177,11 +178,18 @@ void printValue(std::ostream & str, std::set<const void *> & seen, const Value &
 }
 
 
-std::ostream & operator << (std::ostream & str, const Value & v)
+void Value::print(const SymbolTable & symbols, std::ostream & str, bool showRepeated) const
 {
     std::set<const void *> seen;
-    printValue(str, seen, v);
+    print(symbols, str, showRepeated ? nullptr : &seen);
-    return str;
+}
+
+
+std::string printValue(const EvalState & state, const Value & v)
+{
+    std::ostringstream out;
+    v.print(state.symbols, out);
+    return out.str();
 }
 
 
@@ -230,10 +238,10 @@ std::string showType(const Value & v)
     }
 }
 
-Pos Value::determinePos(const Pos & pos) const
+PosIdx Value::determinePos(const PosIdx pos) const
 {
     switch (internalType) {
-        case tAttrs: return *attrs->pos;
+        case tAttrs: return attrs->pos;
         case tLambda: return lambda.fun->pos;
         case tApp: return app.left->determinePos(pos);
         default: return pos;
@@ -300,9 +308,9 @@ static BoehmGCStackAllocator boehmGCStackAllocator;
 #endif
 
 
-static Symbol getName(const AttrName & name, EvalState & state, Env & env)
+static SymbolIdx getName(const AttrName & name, EvalState & state, Env & env)
 {
-    if (name.symbol.set()) {
+    if (name.symbol) {
         return name.symbol;
     } else {
         Value nameValue;
@@ -430,6 +438,7 @@ EvalState::EvalState(
     , sBuilder(symbols.create("builder"))
     , sArgs(symbols.create("args"))
     , sContentAddressed(symbols.create("__contentAddressed"))
+    , sImpure(symbols.create("__impure"))
     , sOutputHash(symbols.create("outputHash"))
     , sOutputHashAlgo(symbols.create("outputHashAlgo"))
     , sOutputHashMode(symbols.create("outputHashMode"))
@@ -449,8 +458,10 @@ EvalState::EvalState(
     , regexCache(makeRegexCache())
 #if HAVE_BOEHMGC
     , valueAllocCache(std::allocate_shared<void *>(traceable_allocator<void *>(), nullptr))
+    , env1AllocCache(std::allocate_shared<void *>(traceable_allocator<void *>(), nullptr))
 #else
     , valueAllocCache(std::make_shared<void *>(nullptr))
+    , env1AllocCache(std::make_shared<void *>(nullptr))
 #endif
     , baseEnv(allocEnv(128))
     , staticBaseEnv(false, 0)
@@ -499,23 +510,6 @@ EvalState::~EvalState()
 }
 
 
-void EvalState::requireExperimentalFeatureOnEvaluation(
-    const ExperimentalFeature & feature,
-    const std::string_view fName,
-    const Pos & pos)
-{
-    if (!settings.isExperimentalFeatureEnabled(feature)) {
-        throw EvalError({
-            .msg = hintfmt(
-                "Cannot call '%2%' because experimental Nix feature '%1%' is disabled. You can enable it via '--extra-experimental-features %1%'.",
-                feature,
-                fName
-            ),
-            .errPos = pos
-        });
-    }
-}
-
 void EvalState::allowPath(const Path & path)
 {
     if (allowedPaths)
@@ -647,20 +641,20 @@ Value * EvalState::addPrimOp(const std::string & name,
     size_t arity, PrimOpFun primOp)
 {
     auto name2 = name.substr(0, 2) == "__" ? name.substr(2) : name;
-    Symbol sym = symbols.create(name2);
+    auto sym = symbols.create(name2);
 
     /* Hack to make constants lazy: turn them into a application of
        the primop to a dummy value. */
     if (arity == 0) {
         auto vPrimOp = allocValue();
-        vPrimOp->mkPrimOp(new PrimOp { .fun = primOp, .arity = 1, .name = sym });
+        vPrimOp->mkPrimOp(new PrimOp { .fun = primOp, .arity = 1, .name = name2 });
         Value v;
         v.mkApp(vPrimOp, vPrimOp);
         return addConstant(name, v);
     }
 
     Value * v = allocValue();
-    v->mkPrimOp(new PrimOp { .fun = primOp, .arity = arity, .name = sym });
+    v->mkPrimOp(new PrimOp { .fun = primOp, .arity = arity, .name = name2 });
     staticBaseEnv.vars.emplace_back(symbols.create(name), baseEnvDispl);
     baseEnv.values[baseEnvDispl++] = v;
     baseEnv.values[0]->attrs->push_back(Attr(sym, v));
@@ -675,21 +669,21 @@ Value * EvalState::addPrimOp(PrimOp && primOp)
     if (primOp.arity == 0) {
         primOp.arity = 1;
         auto vPrimOp = allocValue();
-        vPrimOp->mkPrimOp(new PrimOp(std::move(primOp)));
+        vPrimOp->mkPrimOp(new PrimOp(primOp));
         Value v;
         v.mkApp(vPrimOp, vPrimOp);
         return addConstant(primOp.name, v);
     }
 
-    Symbol envName = primOp.name;
+    auto envName = symbols.create(primOp.name);
     if (hasPrefix(primOp.name, "__"))
-        primOp.name = symbols.create(std::string(primOp.name, 2));
+        primOp.name = primOp.name.substr(2);
 
     Value * v = allocValue();
-    v->mkPrimOp(new PrimOp(std::move(primOp)));
+    v->mkPrimOp(new PrimOp(primOp));
     staticBaseEnv.vars.emplace_back(envName, baseEnvDispl);
     baseEnv.values[baseEnvDispl++] = v;
-    baseEnv.values[0]->attrs->push_back(Attr(primOp.name, v));
+    baseEnv.values[0]->attrs->push_back(Attr(symbols.create(primOp.name), v));
     return v;
 }
 
@@ -706,7 +700,7 @@ std::optional<EvalState::Doc> EvalState::getDoc(Value & v)
         auto v2 = &v;
         if (v2->primOp->doc)
             return Doc {
-                .pos = noPos,
+                .pos = {},
                 .name = v2->primOp->name,
                 .arity = v2->primOp->arity,
                 .args = v2->primOp->args,
@@ -722,94 +716,133 @@ std::optional<EvalState::Doc> EvalState::getDoc(Value & v)
    evaluator.  So here are some helper functions for throwing
    exceptions. */
 
-LocalNoInlineNoReturn(void throwEvalError(const char * s, const std::string & s2))
+void EvalState::throwEvalError(const PosIdx pos, const char * s) const
+{
+    throw EvalError({
+        .msg = hintfmt(s),
+        .errPos = positions[pos]
+    });
+}
+
+void EvalState::throwTypeError(const PosIdx pos, const char * s, const Value & v) const
+{
+    throw TypeError({
+        .msg = hintfmt(s, showType(v)),
+        .errPos = positions[pos]
+    });
+}
+
+void EvalState::throwEvalError(const char * s, const std::string & s2) const
 {
     throw EvalError(s, s2);
 }
 
-LocalNoInlineNoReturn(void throwEvalError(const Pos & pos, const char * s, const std::string & s2))
+void EvalState::throwEvalError(const PosIdx pos, const Suggestions & suggestions, const char * s,
+    const std::string & s2) const
 {
-    throw EvalError({
+    throw EvalError(ErrorInfo {
         .msg = hintfmt(s, s2),
-        .errPos = pos
+        .errPos = positions[pos],
+        .suggestions = suggestions,
     });
 }
 
-LocalNoInlineNoReturn(void throwEvalError(const char * s, const std::string & s2, const std::string & s3))
+void EvalState::throwEvalError(const PosIdx pos, const char * s, const std::string & s2) const
+{
+    throw EvalError(ErrorInfo {
+        .msg = hintfmt(s, s2),
+        .errPos = positions[pos]
+    });
+}
+
+void EvalState::throwEvalError(const char * s, const std::string & s2, const std::string & s3) const
 {
     throw EvalError(s, s2, s3);
 }
 
-LocalNoInlineNoReturn(void throwEvalError(const Pos & pos, const char * s, const std::string & s2, const std::string & s3))
+void EvalState::throwEvalError(const PosIdx pos, const char * s, const std::string & s2,
+    const std::string & s3) const
 {
     throw EvalError({
         .msg = hintfmt(s, s2, s3),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInlineNoReturn(void throwEvalError(const Pos & p1, const char * s, const Symbol & sym, const Pos & p2))
+void EvalState::throwEvalError(const PosIdx p1, const char * s, const SymbolIdx sym, const PosIdx p2) const
 {
     // p1 is where the error occurred; p2 is a position mentioned in the message.
     throw EvalError({
-        .msg = hintfmt(s, sym, p2),
+        .msg = hintfmt(s, symbols[sym], positions[p2]),
-        .errPos = p1
+        .errPos = positions[p1]
     });
 }
 
-LocalNoInlineNoReturn(void throwTypeError(const Pos & pos, const char * s))
+void EvalState::throwTypeError(const PosIdx pos, const char * s) const
 {
     throw TypeError({
         .msg = hintfmt(s),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInlineNoReturn(void throwTypeError(const Pos & pos, const char * s, const ExprLambda & fun, const Symbol & s2))
+void EvalState::throwTypeError(const PosIdx pos, const char * s, const ExprLambda & fun,
+    const SymbolIdx s2) const
 {
     throw TypeError({
-        .msg = hintfmt(s, fun.showNamePos(), s2),
+        .msg = hintfmt(s, fun.showNamePos(*this), symbols[s2]),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInlineNoReturn(void throwTypeError(const char * s, const Value & v))
+void EvalState::throwTypeError(const PosIdx pos, const Suggestions & suggestions, const char * s,
+    const ExprLambda & fun, const SymbolIdx s2) const
+{
+    throw TypeError(ErrorInfo {
+        .msg = hintfmt(s, fun.showNamePos(*this), symbols[s2]),
+        .errPos = positions[pos],
+        .suggestions = suggestions,
+    });
+}
+
+
+void EvalState::throwTypeError(const char * s, const Value & v) const
 {
     throw TypeError(s, showType(v));
 }
 
-LocalNoInlineNoReturn(void throwAssertionError(const Pos & pos, const char * s, const std::string & s1))
+void EvalState::throwAssertionError(const PosIdx pos, const char * s, const std::string & s1) const
 {
     throw AssertionError({
         .msg = hintfmt(s, s1),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInlineNoReturn(void throwUndefinedVarError(const Pos & pos, const char * s, const std::string & s1))
+void EvalState::throwUndefinedVarError(const PosIdx pos, const char * s, const std::string & s1) const
 {
     throw UndefinedVarError({
         .msg = hintfmt(s, s1),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInlineNoReturn(void throwMissingArgumentError(const Pos & pos, const char * s, const std::string & s1))
+void EvalState::throwMissingArgumentError(const PosIdx pos, const char * s, const std::string & s1) const
 {
     throw MissingArgumentError({
         .msg = hintfmt(s, s1),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
-LocalNoInline(void addErrorTrace(Error & e, const char * s, const std::string & s2))
+void EvalState::addErrorTrace(Error & e, const char * s, const std::string & s2) const
 {
     e.addTrace(std::nullopt, s, s2);
 }
 
-LocalNoInline(void addErrorTrace(Error & e, const Pos & pos, const char * s, const std::string & s2))
+void EvalState::addErrorTrace(Error & e, const PosIdx pos, const char * s, const std::string & s2) const
 {
-    e.addTrace(pos, s, s2);
+    e.addTrace(positions[pos], s, s2);
 }
 
 
@@ -866,52 +899,16 @@ inline Value * EvalState::lookupVar(Env * env, const ExprVar & var, bool noEval)
         }
         Bindings::iterator j = env->values[0]->attrs->find(var.name);
         if (j != env->values[0]->attrs->end()) {
-            if (countCalls) attrSelects[*j->pos]++;
+            if (countCalls) attrSelects[j->pos]++;
             return j->value;
         }
         if (!env->prevWith)
-            throwUndefinedVarError(var.pos, "undefined variable '%1%'", var.name);
+            throwUndefinedVarError(var.pos, "undefined variable '%1%'", symbols[var.name]);
         for (size_t l = env->prevWith; l; --l, env = env->up) ;
     }
 }
 
 
-Value * EvalState::allocValue()
-{
-    /* We use the boehm batch allocator to speed up allocations of Values (of which there are many).
-       GC_malloc_many returns a linked list of objects of the given size, where the first word
-       of each object is also the pointer to the next object in the list. This also means that we
-       have to explicitly clear the first word of every object we take. */
-    if (!*valueAllocCache) {
-        *valueAllocCache = GC_malloc_many(sizeof(Value));
-        if (!*valueAllocCache) throw std::bad_alloc();
-    }
-
-    /* GC_NEXT is a convenience macro for accessing the first word of an object.
-       Take the first list item, advance the list to the next item, and clear the next pointer. */
-    void * p = *valueAllocCache;
-    GC_PTR_STORE_AND_DIRTY(&*valueAllocCache, GC_NEXT(p));
-    GC_NEXT(p) = nullptr;
-
-    nrValues++;
-    auto v = (Value *) p;
-    return v;
-}
-
-
-Env & EvalState::allocEnv(size_t size)
-{
-    nrEnvs++;
-    nrValuesInEnvs += size;
-    Env * env = (Env *) allocBytes(sizeof(Env) + size * sizeof(Value *));
-    env->type = Env::Plain;
-
-    /* We assume that env->values has been cleared by the allocator; maybeThunk() and lookupVar fromWith expect this. */
-
-    return *env;
-}
-
-
 void EvalState::mkList(Value & v, size_t size)
 {
     v.mkList(size);
@@ -936,13 +933,14 @@ void EvalState::mkThunk_(Value & v, Expr * expr)
 }
 
 
-void EvalState::mkPos(Value & v, ptr<Pos> pos)
+void EvalState::mkPos(Value & v, PosIdx p)
 {
-    if (pos->file.set()) {
+    auto pos = positions[p];
+    if (!pos.file.empty()) {
         auto attrs = buildBindings(3);
-        attrs.alloc(sFile).mkString(pos->file);
+        attrs.alloc(sFile).mkString(pos.file);
-        attrs.alloc(sLine).mkInt(pos->line);
+        attrs.alloc(sLine).mkInt(pos.line);
-        attrs.alloc(sColumn).mkInt(pos->column);
+        attrs.alloc(sColumn).mkInt(pos.column);
         v.mkAttrs(attrs);
     } else
         v.mkNull();
@@ -1075,7 +1073,7 @@ inline bool EvalState::evalBool(Env & env, Expr * e)
 }
 
 
-inline bool EvalState::evalBool(Env & env, Expr * e, const Pos & pos)
+inline bool EvalState::evalBool(Env & env, Expr * e, const PosIdx pos)
 {
     Value v;
     e->eval(*this, env, v);
@@ -1149,7 +1147,7 @@ void ExprAttrs::eval(EvalState & state, Env & env, Value & v)
             } else
                 vAttr = i.second.e->maybeThunk(state, i.second.inherited ? env : env2);
             env2.values[displ++] = vAttr;
-            v.attrs->push_back(Attr(i.first, vAttr, ptr(&i.second.pos)));
+            v.attrs->push_back(Attr(i.first, vAttr, i.second.pos));
         }
 
         /* If the rec contains an attribute called `__overrides', then
@@ -1181,7 +1179,7 @@ void ExprAttrs::eval(EvalState & state, Env & env, Value & v)
 
     else
         for (auto & i : attrs)
-            v.attrs->push_back(Attr(i.first, i.second.e->maybeThunk(state, env), ptr(&i.second.pos)));
+            v.attrs->push_back(Attr(i.first, i.second.e->maybeThunk(state, env), i.second.pos));
 
     /* Dynamic attrs apply *after* rec and __overrides. */
     for (auto & i : dynamicAttrs) {
@@ -1191,18 +1189,18 @@ void ExprAttrs::eval(EvalState & state, Env & env, Value & v)
         if (nameVal.type() == nNull)
             continue;
         state.forceStringNoCtx(nameVal);
-        Symbol nameSym = state.symbols.create(nameVal.string.s);
+        auto nameSym = state.symbols.create(nameVal.string.s);
         Bindings::iterator j = v.attrs->find(nameSym);
         if (j != v.attrs->end())
-            throwEvalError(i.pos, "dynamic attribute '%1%' already defined at %2%", nameSym, *j->pos);
+            state.throwEvalError(i.pos, "dynamic attribute '%1%' already defined at %2%", nameSym, j->pos);
 
         i.valueExpr->setName(nameSym);
         /* Keep sorted order so find can catch duplicates */
-        v.attrs->push_back(Attr(nameSym, i.valueExpr->maybeThunk(state, *dynamicEnv), ptr(&i.pos)));
+        v.attrs->push_back(Attr(nameSym, i.valueExpr->maybeThunk(state, *dynamicEnv), i.pos));
         v.attrs->sort(); // FIXME: inefficient
     }
 
-    v.attrs->pos = ptr(&pos);
+    v.attrs->pos = pos;
 }
 
 
@@ -1247,10 +1245,12 @@ static std::string showAttrPath(EvalState & state, Env & env, const AttrPath & a
     for (auto & i : attrPath) {
         if (!first) out << '.'; else first = false;
         try {
-            out << getName(i, state, env);
+            out << state.symbols[getName(i, state, env)];
         } catch (Error & e) {
-            assert(!i.symbol.set());
+            assert(!i.symbol);
-            out << "\"${" << *i.expr << "}\"";
+            out << "\"${";
+            i.expr->show(state.symbols, out);
+            out << "}\"";
         }
     }
     return out.str();
@@ -1260,7 +1260,7 @@ static std::string showAttrPath(EvalState & state, Env & env, const AttrPath & a
 void ExprSelect::eval(EvalState & state, Env & env, Value & v)
 {
     Value vTmp;
-    ptr<Pos> pos2(&noPos);
+    PosIdx pos2;
     Value * vAttrs = &vTmp;
 
     e->eval(state, env, vTmp);
@@ -1270,7 +1270,7 @@ void ExprSelect::eval(EvalState & state, Env & env, Value & v)
         for (auto & i : attrPath) {
             state.nrLookups++;
             Bindings::iterator j;
-            Symbol name = getName(i, state, env);
+            auto name = getName(i, state, env);
             if (def) {
                 state.forceValue(*vAttrs, pos);
                 if (vAttrs->type() != nAttrs ||
@@ -1281,19 +1281,27 @@ void ExprSelect::eval(EvalState & state, Env & env, Value & v)
                 }
             } else {
                 state.forceAttrs(*vAttrs, pos);
-                if ((j = vAttrs->attrs->find(name)) == vAttrs->attrs->end())
+                if ((j = vAttrs->attrs->find(name)) == vAttrs->attrs->end()) {
-                    throwEvalError(pos, "attribute '%1%' missing", name);
+                    std::set<std::string> allAttrNames;
+                    for (auto & attr : *vAttrs->attrs)
+                        allAttrNames.insert(state.symbols[attr.name]);
+                    state.throwEvalError(
+                        pos,
+                        Suggestions::bestMatches(allAttrNames, state.symbols[name]),
+                        "attribute '%1%' missing", state.symbols[name]);
+                }
             }
             vAttrs = j->value;
             pos2 = j->pos;
-            if (state.countCalls) state.attrSelects[*pos2]++;
+            if (state.countCalls) state.attrSelects[pos2]++;
         }
 
-        state.forceValue(*vAttrs, (*pos2 != noPos ? *pos2 : this->pos ) );
+        state.forceValue(*vAttrs, (pos2 ? pos2 : this->pos ) );
 
     } catch (Error & e) {
-        if (*pos2 != noPos && pos2->file != state.sDerivationNix)
+        auto pos2r = state.positions[pos2];
-            addErrorTrace(e, *pos2, "while evaluating the attribute '%1%'",
+        if (pos2 && pos2r.file != state.derivationNixPath)
+            state.addErrorTrace(e, pos2, "while evaluating the attribute '%1%'",
                 showAttrPath(state, env, attrPath));
         throw;
     }
@@ -1312,7 +1320,7 @@ void ExprOpHasAttr::eval(EvalState & state, Env & env, Value & v)
     for (auto & i : attrPath) {
         state.forceValue(*vAttrs, noPos);
         Bindings::iterator j;
-        Symbol name = getName(i, state, env);
+        auto name = getName(i, state, env);
         if (vAttrs->type() != nAttrs ||
             (j = vAttrs->attrs->find(name)) == vAttrs->attrs->end())
         {
@@ -1333,9 +1341,11 @@ void ExprLambda::eval(EvalState & state, Env & env, Value & v)
 }
 
 
-void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value & vRes, const Pos & pos)
+void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value & vRes, const PosIdx pos)
 {
-    auto trace = evalSettings.traceFunctionCalls ? std::make_unique<FunctionCallTrace>(pos) : nullptr;
+    auto trace = evalSettings.traceFunctionCalls
+        ? std::make_unique<FunctionCallTrace>(positions[pos])
+        : nullptr;
 
     forceValue(fun, pos);
 
@@ -1360,7 +1370,7 @@ void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value &
             ExprLambda & lambda(*vCur.lambda.fun);
 
             auto size =
-                (lambda.arg.empty() ? 0 : 1) +
+                (!lambda.arg ? 0 : 1) +
                 (lambda.hasFormals() ? lambda.formals->formals.size() : 0);
             Env & env2(allocEnv(size));
             env2.up = vCur.lambda.env;
@@ -1373,7 +1383,7 @@ void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value &
             else {
                 forceAttrs(*args[0], pos);
 
-                if (!lambda.arg.empty())
+                if (lambda.arg)
                     env2.values[displ++] = args[0];
 
                 /* For each formal argument, get the actual argument.  If
@@ -1398,8 +1408,17 @@ void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value &
                     /* Nope, so show the first unexpected argument to the
                        user. */
                     for (auto & i : *args[0]->attrs)
-                        if (!lambda.formals->has(i.name))
+                        if (!lambda.formals->has(i.name)) {
-                            throwTypeError(pos, "%1% called with unexpected argument '%2%'", lambda, i.name);
+                            std::set<std::string> formalNames;
+                            for (auto & formal : lambda.formals->formals)
+                                formalNames.insert(symbols[formal.name]);
+                            throwTypeError(
+                                pos,
+                                Suggestions::bestMatches(formalNames, symbols[i.name]),
+                                "%1% called with unexpected argument '%2%'",
+                                lambda,
+                                i.name);
+                        }
                     abort(); // can't happen
                 }
             }
@@ -1413,8 +1432,8 @@ void EvalState::callFunction(Value & fun, size_t nrArgs, Value * * args, Value &
             } catch (Error & e) {
                 if (loggerSettings.showTrace.get()) {
                     addErrorTrace(e, lambda.pos, "while evaluating %s",
-                        (lambda.name.set()
+                        (lambda.name
-                            ? "'" + (const std::string &) lambda.name + "'"
+                            ? concatStrings("'", symbols[lambda.name], "'")
                             : "anonymous lambda"));
                     addErrorTrace(e, pos, "from call site%s", "");
                 }
@@ -1563,7 +1582,7 @@ void EvalState::autoCallFunction(Bindings & args, Value & fun, Value & res)
 Nix attempted to evaluate a function as a top level expression; in
 this case it must have its arguments supplied either by default
 values, or passed explicitly with '--arg' or '--argstr'. See
-https://nixos.org/manual/nix/stable/#ss-functions.)", i.name);
+https://nixos.org/manual/nix/stable/#ss-functions.)", symbols[i.name]);
 
             }
         }
@@ -1595,8 +1614,8 @@ void ExprAssert::eval(EvalState & state, Env & env, Value & v)
 {
     if (!state.evalBool(env, cond, pos)) {
         std::ostringstream out;
-        cond->show(out);
+        cond->show(state.symbols, out);
-        throwAssertionError(pos, "assertion '%1%' failed", out.str());
+        state.throwAssertionError(pos, "assertion '%1%' failed", out.str());
     }
     body->eval(state, env, v);
 }
@@ -1689,7 +1708,7 @@ void ExprOpConcatLists::eval(EvalState & state, Env & env, Value & v)
 }
 
 
-void EvalState::concatLists(Value & v, size_t nrLists, Value * * lists, const Pos & pos)
+void EvalState::concatLists(Value & v, size_t nrLists, Value * * lists, const PosIdx pos)
 {
     nrListConcats++;
 
@@ -1773,14 +1792,14 @@ void ExprConcatStrings::eval(EvalState & state, Env & env, Value & v)
                 nf = n;
                 nf += vTmp.fpoint;
             } else
-                throwEvalError(i_pos, "cannot add %1% to an integer", showType(vTmp));
+                state.throwEvalError(i_pos, "cannot add %1% to an integer", showType(vTmp));
         } else if (firstType == nFloat) {
             if (vTmp.type() == nInt) {
                 nf += vTmp.integer;
             } else if (vTmp.type() == nFloat) {
                 nf += vTmp.fpoint;
             } else
-                throwEvalError(i_pos, "cannot add %1% to a float", showType(vTmp));
+                state.throwEvalError(i_pos, "cannot add %1% to a float", showType(vTmp));
         } else {
             if (s.empty()) s.reserve(es->size());
             /* skip canonization of first path, which would only be not
@@ -1800,7 +1819,7 @@ void ExprConcatStrings::eval(EvalState & state, Env & env, Value & v)
         v.mkFloat(nf);
     else if (firstType == nPath) {
         if (!context.empty())
-            throwEvalError(pos, "a string that refers to a store path cannot be appended to a path");
+            state.throwEvalError(pos, "a string that refers to a store path cannot be appended to a path");
         v.mkPath(canonPath(str()));
     } else
         v.mkStringMove(c_str(), context);
@@ -1809,7 +1828,7 @@ void ExprConcatStrings::eval(EvalState & state, Env & env, Value & v)
 
 void ExprPos::eval(EvalState & state, Env & env, Value & v)
 {
-    state.mkPos(v, ptr(&pos));
+    state.mkPos(v, pos);
 }
 
 
@@ -1829,7 +1848,7 @@ void EvalState::forceValueDeep(Value & v)
                 try {
                     recurse(*i.value);
                 } catch (Error & e) {
-                    addErrorTrace(e, *i.pos, "while evaluating the attribute '%1%'", i.name);
+                    addErrorTrace(e, i.pos, "while evaluating the attribute '%1%'", symbols[i.name]);
                     throw;
                 }
         }
@@ -1844,7 +1863,7 @@ void EvalState::forceValueDeep(Value & v)
 }
 
 
-NixInt EvalState::forceInt(Value & v, const Pos & pos)
+NixInt EvalState::forceInt(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (v.type() != nInt)
@@ -1853,7 +1872,7 @@ NixInt EvalState::forceInt(Value & v, const Pos & pos)
 }
 
 
-NixFloat EvalState::forceFloat(Value & v, const Pos & pos)
+NixFloat EvalState::forceFloat(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (v.type() == nInt)
@@ -1864,7 +1883,7 @@ NixFloat EvalState::forceFloat(Value & v, const Pos & pos)
 }
 
 
-bool EvalState::forceBool(Value & v, const Pos & pos)
+bool EvalState::forceBool(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (v.type() != nBool)
@@ -1879,7 +1898,7 @@ bool EvalState::isFunctor(Value & fun)
 }
 
 
-void EvalState::forceFunction(Value & v, const Pos & pos)
+void EvalState::forceFunction(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (v.type() != nFunction && !isFunctor(v))
@@ -1887,7 +1906,7 @@ void EvalState::forceFunction(Value & v, const Pos & pos)
 }
 
 
-std::string_view EvalState::forceString(Value & v, const Pos & pos)
+std::string_view EvalState::forceString(Value & v, const PosIdx pos)
 {
     forceValue(v, pos);
     if (v.type() != nString) {
@@ -1902,13 +1921,22 @@ std::string_view EvalState::forceString(Value & v, const Pos & pos)
 
 /* Decode a context string ‘!<name>!<path>’ into a pair <path,
    name>. */
-std::pair<std::string, std::string> decodeContext(std::string_view s)
+NixStringContextElem decodeContext(const Store & store, std::string_view s)
 {
     if (s.at(0) == '!') {
         size_t index = s.find("!", 1);
-        return {std::string(s.substr(index + 1)), std::string(s.substr(1, index - 1))};
+        return {
+            store.parseStorePath(s.substr(index + 1)),
+            std::string(s.substr(1, index - 1)),
+        };
     } else
-        return {s.at(0) == '/' ? std::string(s) : std::string(s.substr(1)), ""};
+        return {
+            store.parseStorePath(
+                s.at(0) == '/'
+                ? s
+                : s.substr(1)),
+            "",
+        };
 }
 
 
@@ -1920,18 +1948,18 @@ void copyContext(const Value & v, PathSet & context)
 }
 
 
-std::vector<std::pair<Path, std::string>> Value::getContext()
+NixStringContext Value::getContext(const Store & store)
 {
-    std::vector<std::pair<Path, std::string>> res;
+    NixStringContext res;
     assert(internalType == tString);
     if (string.context)
         for (const char * * p = string.context; *p; ++p)
-            res.push_back(decodeContext(*p));
+            res.push_back(decodeContext(store, *p));
     return res;
 }
 
 
-std::string_view EvalState::forceString(Value & v, PathSet & context, const Pos & pos)
+std::string_view EvalState::forceString(Value & v, PathSet & context, const PosIdx pos)
 {
     auto s = forceString(v, pos);
     copyContext(v, context);
@@ -1939,7 +1967,7 @@ std::string_view EvalState::forceString(Value & v, PathSet & context, const Pos
 }
 
 
-std::string_view EvalState::forceStringNoCtx(Value & v, const Pos & pos)
+std::string_view EvalState::forceStringNoCtx(Value & v, const PosIdx pos)
 {
     auto s = forceString(v, pos);
     if (v.string.context) {
@@ -1959,13 +1987,13 @@ bool EvalState::isDerivation(Value & v)
     if (v.type() != nAttrs) return false;
     Bindings::iterator i = v.attrs->find(sType);
     if (i == v.attrs->end()) return false;
-    forceValue(*i->value, *i->pos);
+    forceValue(*i->value, i->pos);
     if (i->value->type() != nString) return false;
     return strcmp(i->value->string.s, "derivation") == 0;
 }
 
 
-std::optional<std::string> EvalState::tryAttrsToString(const Pos & pos, Value & v,
+std::optional<std::string> EvalState::tryAttrsToString(const PosIdx pos, Value & v,
     PathSet & context, bool coerceMore, bool copyToStore)
 {
     auto i = v.attrs->find(sToString);
@@ -1978,7 +2006,7 @@ std::optional<std::string> EvalState::tryAttrsToString(const Pos & pos, Value &
     return {};
 }
 
-BackedStringView EvalState::coerceToString(const Pos & pos, Value & v, PathSet & context,
+BackedStringView EvalState::coerceToString(const PosIdx pos, Value & v, PathSet & context,
     bool coerceMore, bool copyToStore, bool canonicalizePath)
 {
     forceValue(v, pos);
@@ -2007,7 +2035,7 @@ BackedStringView EvalState::coerceToString(const Pos & pos, Value & v, PathSet &
     }
 
     if (v.type() == nExternal)
-        return v.external->coerceToString(pos, context, coerceMore, copyToStore);
+        return v.external->coerceToString(positions[pos], context, coerceMore, copyToStore);
 
     if (coerceMore) {
 
@@ -2060,7 +2088,7 @@ std::string EvalState::copyPathToStore(PathSet & context, const Path & path)
 }
 
 
-Path EvalState::coerceToPath(const Pos & pos, Value & v, PathSet & context)
+Path EvalState::coerceToPath(const PosIdx pos, Value & v, PathSet & context)
 {
     auto path = coerceToString(pos, v, context, false, false).toOwned();
     if (path == "" || path[0] != '/')
@@ -2069,14 +2097,14 @@ Path EvalState::coerceToPath(const Pos & pos, Value & v, PathSet & context)
 }
 
 
-StorePath EvalState::coerceToStorePath(const Pos & pos, Value & v, PathSet & context)
+StorePath EvalState::coerceToStorePath(const PosIdx pos, Value & v, PathSet & context)
 {
     auto path = coerceToString(pos, v, context, false, false).toOwned();
     if (auto storePath = store->maybeParseStorePath(path))
         return *storePath;
     throw EvalError({
         .msg = hintfmt("path '%1%' is not in the Nix store", path),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 
@@ -2243,14 +2271,14 @@ void EvalState::printStats()
                 auto list = topObj.list("functions");
                 for (auto & i : functionCalls) {
                     auto obj = list.object();
-                    if (i.first->name.set())
+                    if (i.first->name)
                         obj.attr("name", (const std::string &) i.first->name);
                     else
                         obj.attr("name", nullptr);
-                    if (i.first->pos) {
+                    if (auto pos = positions[i.first->pos]) {
-                        obj.attr("file", (const std::string &) i.first->pos.file);
+                        obj.attr("file", (const std::string &) pos.file);
-                        obj.attr("line", i.first->pos.line);
+                        obj.attr("line", pos.line);
-                        obj.attr("column", i.first->pos.column);
+                        obj.attr("column", pos.column);
                     }
                     obj.attr("count", i.second);
                 }
@@ -2259,10 +2287,10 @@ void EvalState::printStats()
                 auto list = topObj.list("attributes");
                 for (auto & i : attrSelects) {
                     auto obj = list.object();
-                    if (i.first) {
+                    if (auto pos = positions[i.first]) {
-                        obj.attr("file", (const std::string &) i.first.file);
+                        obj.attr("file", (const std::string &) pos.file);
-                        obj.attr("line", i.first.line);
+                        obj.attr("line", pos.line);
-                        obj.attr("column", i.first.column);
+                        obj.attr("column", pos.column);
                     }
                     obj.attr("count", i.second);
                 }

src/libexpr/eval.hh

@@ -23,14 +23,14 @@ class StorePath;
 enum RepairFlag : bool;
 
 
-typedef void (* PrimOpFun) (EvalState & state, const Pos & pos, Value * * args, Value & v);
+typedef void (* PrimOpFun) (EvalState & state, const PosIdx pos, Value * * args, Value & v);
 
 
 struct PrimOp
 {
     PrimOpFun fun;
     size_t arity;
-    Symbol name;
+    std::string name;
     std::vector<std::string> args;
     const char * doc = nullptr;
 };
@@ -53,7 +53,8 @@ void copyContext(const Value & v, PathSet & context);
 typedef std::map<Path, StorePath> SrcToStore;
 
 
-std::ostream & operator << (std::ostream & str, const Value & v);
+std::ostream & printValue(const EvalState & state, std::ostream & str, const Value & v);
+std::string printValue(const EvalState & state, const Value & v);
 
 
 typedef std::pair<std::string, std::string> SearchPathElem;
@@ -73,17 +74,20 @@ class EvalState
 {
 public:
     SymbolTable symbols;
+    PosTable positions;
 
-    const Symbol sWith, sOutPath, sDrvPath, sType, sMeta, sName, sValue,
+    static inline std::string derivationNixPath = "//builtin/derivation.nix";
+
+    const SymbolIdx sWith, sOutPath, sDrvPath, sType, sMeta, sName, sValue,
         sSystem, sOverrides, sOutputs, sOutputName, sIgnoreNulls,
         sFile, sLine, sColumn, sFunctor, sToString,
         sRight, sWrong, sStructuredAttrs, sBuilder, sArgs,
-        sContentAddressed,
+        sContentAddressed, sImpure,
         sOutputHash, sOutputHashAlgo, sOutputHashMode,
         sRecurseForDerivations,
         sDescription, sSelf, sEpsilon, sStartSet, sOperator, sKey, sPath,
         sPrefix;
-    Symbol sDerivationNix;
+    SymbolIdx sDerivationNix;
 
     /* If set, force copying files to the Nix store even if they
        already exist there. */
@@ -133,9 +137,14 @@ private:
     /* Cache used by prim_match(). */
     std::shared_ptr<RegexCache> regexCache;
 
+#if HAVE_BOEHMGC
     /* Allocation cache for GC'd Value objects. */
     std::shared_ptr<void *> valueAllocCache;
 
+    /* Allocation cache for size-1 Env objects. */
+    std::shared_ptr<void *> env1AllocCache;
+#endif
+
 public:
 
     EvalState(
@@ -144,12 +153,6 @@ public:
         std::shared_ptr<Store> buildStore = nullptr);
     ~EvalState();
 
-    void requireExperimentalFeatureOnEvaluation(
-        const ExperimentalFeature &,
-        const std::string_view fName,
-        const Pos & pos
-    );
-
     void addToSearchPath(const std::string & s);
 
     SearchPath getSearchPath() { return searchPath; }
@@ -206,7 +209,7 @@ public:
 
     /* Look up a file in the search path. */
     Path findFile(const std::string_view path);
-    Path findFile(SearchPath & searchPath, const std::string_view path, const Pos & pos = noPos);
+    Path findFile(SearchPath & searchPath, const std::string_view path, const PosIdx pos = noPos);
 
     /* If the specified search path element is a URI, download it. */
     std::pair<bool, std::string> resolveSearchPathElem(const SearchPathElem & elem);
@@ -218,14 +221,14 @@ public:
     /* Evaluation the expression, then verify that it has the expected
        type. */
     inline bool evalBool(Env & env, Expr * e);
-    inline bool evalBool(Env & env, Expr * e, const Pos & pos);
+    inline bool evalBool(Env & env, Expr * e, const PosIdx pos);
     inline void evalAttrs(Env & env, Expr * e, Value & v);
 
     /* If `v' is a thunk, enter it and overwrite `v' with the result
        of the evaluation of the thunk.  If `v' is a delayed function
        application, call the function and overwrite `v' with the
        result.  Otherwise, this is a no-op. */
-    inline void forceValue(Value & v, const Pos & pos);
+    inline void forceValue(Value & v, const PosIdx pos);
 
     template <typename Callable>
     inline void forceValue(Value & v, Callable getPos);
@@ -235,33 +238,72 @@ public:
     void forceValueDeep(Value & v);
 
     /* Force `v', and then verify that it has the expected type. */
-    NixInt forceInt(Value & v, const Pos & pos);
+    NixInt forceInt(Value & v, const PosIdx pos);
-    NixFloat forceFloat(Value & v, const Pos & pos);
+    NixFloat forceFloat(Value & v, const PosIdx pos);
-    bool forceBool(Value & v, const Pos & pos);
+    bool forceBool(Value & v, const PosIdx pos);
 
-    void forceAttrs(Value & v, const Pos & pos);
+    void forceAttrs(Value & v, const PosIdx pos);
 
     template <typename Callable>
     inline void forceAttrs(Value & v, Callable getPos);
 
-    inline void forceList(Value & v, const Pos & pos);
+    inline void forceList(Value & v, const PosIdx pos);
-    void forceFunction(Value & v, const Pos & pos); // either lambda or primop
+    void forceFunction(Value & v, const PosIdx pos); // either lambda or primop
-    std::string_view forceString(Value & v, const Pos & pos = noPos);
+    std::string_view forceString(Value & v, const PosIdx pos = noPos);
-    std::string_view forceString(Value & v, PathSet & context, const Pos & pos = noPos);
+    std::string_view forceString(Value & v, PathSet & context, const PosIdx pos = noPos);
-    std::string_view forceStringNoCtx(Value & v, const Pos & pos = noPos);
+    std::string_view forceStringNoCtx(Value & v, const PosIdx pos = noPos);
 
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const PosIdx pos, const char * s) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwTypeError(const PosIdx pos, const char * s, const Value & v) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const char * s, const std::string & s2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const PosIdx pos, const Suggestions & suggestions, const char * s,
+        const std::string & s2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const PosIdx pos, const char * s, const std::string & s2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const char * s, const std::string & s2, const std::string & s3) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const PosIdx pos, const char * s, const std::string & s2, const std::string & s3) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwEvalError(const PosIdx p1, const char * s, const SymbolIdx sym, const PosIdx p2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwTypeError(const PosIdx pos, const char * s) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwTypeError(const PosIdx pos, const char * s, const ExprLambda & fun, const SymbolIdx s2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwTypeError(const PosIdx pos, const Suggestions & suggestions, const char * s,
+        const ExprLambda & fun, const SymbolIdx s2) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwTypeError(const char * s, const Value & v) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwAssertionError(const PosIdx pos, const char * s, const std::string & s1) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwUndefinedVarError(const PosIdx pos, const char * s, const std::string & s1) const;
+    [[gnu::noinline, gnu::noreturn]]
+    void throwMissingArgumentError(const PosIdx pos, const char * s, const std::string & s1) const;
+
+    [[gnu::noinline]]
+    void addErrorTrace(Error & e, const char * s, const std::string & s2) const;
+    [[gnu::noinline]]
+    void addErrorTrace(Error & e, const PosIdx pos, const char * s, const std::string & s2) const;
+
+public:
     /* Return true iff the value `v' denotes a derivation (i.e. a
        set with attribute `type = "derivation"'). */
     bool isDerivation(Value & v);
 
-    std::optional<std::string> tryAttrsToString(const Pos & pos, Value & v,
+    std::optional<std::string> tryAttrsToString(const PosIdx pos, Value & v,
         PathSet & context, bool coerceMore = false, bool copyToStore = true);
 
     /* String coercion.  Converts strings, paths and derivations to a
        string.  If `coerceMore' is set, also converts nulls, integers,
        booleans and lists to a string.  If `copyToStore' is set,
        referenced paths are copied to the Nix store as a side effect. */
-    BackedStringView coerceToString(const Pos & pos, Value & v, PathSet & context,
+    BackedStringView coerceToString(const PosIdx pos, Value & v, PathSet & context,
         bool coerceMore = false, bool copyToStore = true,
         bool canonicalizePath = true);
 
@@ -270,10 +312,10 @@ public:
     /* Path coercion.  Converts strings, paths and derivations to a
        path.  The result is guaranteed to be a canonicalised, absolute
        path.  Nothing is copied to the store. */
-    Path coerceToPath(const Pos & pos, Value & v, PathSet & context);
+    Path coerceToPath(const PosIdx pos, Value & v, PathSet & context);
 
     /* Like coerceToPath, but the result must be a store path. */
-    StorePath coerceToStorePath(const Pos & pos, Value & v, PathSet & context);
+    StorePath coerceToStorePath(const PosIdx pos, Value & v, PathSet & context);
 
 public:
 
@@ -306,7 +348,7 @@ public:
     struct Doc
     {
         Pos pos;
-        std::optional<Symbol> name;
+        std::optional<std::string> name;
         size_t arity;
         std::vector<std::string> args;
         const char * doc;
@@ -334,9 +376,9 @@ public:
     bool isFunctor(Value & fun);
 
     // FIXME: use std::span
-    void callFunction(Value & fun, size_t nrArgs, Value * * args, Value & vRes, const Pos & pos);
+    void callFunction(Value & fun, size_t nrArgs, Value * * args, Value & vRes, const PosIdx pos);
 
-    void callFunction(Value & fun, Value & arg, Value & vRes, const Pos & pos)
+    void callFunction(Value & fun, Value & arg, Value & vRes, const PosIdx pos)
     {
         Value * args[] = {&arg};
         callFunction(fun, 1, args, vRes, pos);
@@ -347,10 +389,10 @@ public:
     void autoCallFunction(Bindings & args, Value & fun, Value & res);
 
     /* Allocation primitives. */
-    Value * allocValue();
+    inline Value * allocValue();
-    Env & allocEnv(size_t size);
+    inline Env & allocEnv(size_t size);
 
-    Value * allocAttr(Value & vAttrs, const Symbol & name);
+    Value * allocAttr(Value & vAttrs, const SymbolIdx & name);
     Value * allocAttr(Value & vAttrs, std::string_view name);
 
     Bindings * allocBindings(size_t capacity);
@@ -362,9 +404,9 @@ public:
 
     void mkList(Value & v, size_t length);
     void mkThunk_(Value & v, Expr * expr);
-    void mkPos(Value & v, ptr<Pos> pos);
+    void mkPos(Value & v, PosIdx pos);
 
-    void concatLists(Value & v, size_t nrLists, Value * * lists, const Pos & pos);
+    void concatLists(Value & v, size_t nrLists, Value * * lists, const PosIdx pos);
 
     /* Print statistics. */
     void printStats();
@@ -392,7 +434,7 @@ private:
 
     bool countCalls;
 
-    typedef std::map<Symbol, size_t> PrimOpCalls;
+    typedef std::map<std::string, size_t> PrimOpCalls;
     PrimOpCalls primOpCalls;
 
     typedef std::map<ExprLambda *, size_t> FunctionCalls;
@@ -400,7 +442,7 @@ private:
 
     void incrFunctionCall(ExprLambda * fun);
 
-    typedef std::map<Pos, size_t> AttrSelects;
+    typedef std::map<PosIdx, size_t> AttrSelects;
     AttrSelects attrSelects;
 
     friend struct ExprOpUpdate;
@@ -411,9 +453,9 @@ private:
     friend struct ExprFloat;
     friend struct ExprPath;
     friend struct ExprSelect;
-    friend void prim_getAttr(EvalState & state, const Pos & pos, Value * * args, Value & v);
+    friend void prim_getAttr(EvalState & state, const PosIdx pos, Value * * args, Value & v);
-    friend void prim_match(EvalState & state, const Pos & pos, Value * * args, Value & v);
+    friend void prim_match(EvalState & state, const PosIdx pos, Value * * args, Value & v);
-    friend void prim_split(EvalState & state, const Pos & pos, Value * * args, Value & v);
+    friend void prim_split(EvalState & state, const PosIdx pos, Value * * args, Value & v);
 
     friend struct Value;
 };
@@ -425,7 +467,7 @@ std::string showType(const Value & v);
 
 /* Decode a context string ‘!<name>!<path>’ into a pair <path,
    name>. */
-std::pair<std::string, std::string> decodeContext(std::string_view s);
+NixStringContextElem decodeContext(const Store & store, std::string_view s);
 
 /* If `path' refers to a directory, then append "/default.nix". */
 Path resolveExprPath(Path path);
@@ -509,3 +551,5 @@ extern EvalSettings evalSettings;
 static const std::string corepkgsPrefix{"/__corepkgs__/"};
 
 }
+
+#include "eval-inline.hh"

src/libexpr/flake/flake.cc

@@ -72,7 +72,7 @@ static std::tuple<fetchers::Tree, FlakeRef, FlakeRef> fetchOrSubstituteTree(
     return {std::move(tree), resolvedRef, lockedRef};
 }
 
-static void forceTrivialValue(EvalState & state, Value & value, const Pos & pos)
+static void forceTrivialValue(EvalState & state, Value & value, const PosIdx pos)
 {
     if (value.isThunk() && value.isTrivial())
         state.forceValue(value, pos);
@@ -80,20 +80,20 @@ static void forceTrivialValue(EvalState & state, Value & value, const Pos & pos)
 
 
 static void expectType(EvalState & state, ValueType type,
-    Value & value, const Pos & pos)
+    Value & value, const PosIdx pos)
 {
     forceTrivialValue(state, value, pos);
     if (value.type() != type)
         throw Error("expected %s but got %s at %s",
-            showType(type), showType(value.type()), pos);
+            showType(type), showType(value.type()), state.positions[pos]);
 }
 
 static std::map<FlakeId, FlakeInput> parseFlakeInputs(
-    EvalState & state, Value * value, const Pos & pos,
+    EvalState & state, Value * value, const PosIdx pos,
     const std::optional<Path> & baseDir, InputPath lockRootPath);
 
 static FlakeInput parseFlakeInput(EvalState & state,
-    const std::string & inputName, Value * value, const Pos & pos,
+    const std::string & inputName, Value * value, const PosIdx pos,
     const std::optional<Path> & baseDir, InputPath lockRootPath)
 {
     expectType(state, nAttrs, *value, pos);
@@ -111,37 +111,39 @@ static FlakeInput parseFlakeInput(EvalState & state,
     for (nix::Attr attr : *(value->attrs)) {
         try {
             if (attr.name == sUrl) {
-                expectType(state, nString, *attr.value, *attr.pos);
+                expectType(state, nString, *attr.value, attr.pos);
                 url = attr.value->string.s;
                 attrs.emplace("url", *url);
             } else if (attr.name == sFlake) {
-                expectType(state, nBool, *attr.value, *attr.pos);
+                expectType(state, nBool, *attr.value, attr.pos);
                 input.isFlake = attr.value->boolean;
             } else if (attr.name == sInputs) {
-                input.overrides = parseFlakeInputs(state, attr.value, *attr.pos, baseDir, lockRootPath);
+                input.overrides = parseFlakeInputs(state, attr.value, attr.pos, baseDir, lockRootPath);
             } else if (attr.name == sFollows) {
-                expectType(state, nString, *attr.value, *attr.pos);
+                expectType(state, nString, *attr.value, attr.pos);
                 auto follows(parseInputPath(attr.value->string.s));
                 follows.insert(follows.begin(), lockRootPath.begin(), lockRootPath.end());
                 input.follows = follows;
             } else {
                 switch (attr.value->type()) {
                     case nString:
-                        attrs.emplace(attr.name, attr.value->string.s);
+                        attrs.emplace(state.symbols[attr.name], attr.value->string.s);
                         break;
                     case nBool:
-                        attrs.emplace(attr.name, Explicit<bool> { attr.value->boolean });
+                        attrs.emplace(state.symbols[attr.name], Explicit<bool> { attr.value->boolean });
                         break;
                     case nInt:
-                        attrs.emplace(attr.name, (long unsigned int)attr.value->integer);
+                        attrs.emplace(state.symbols[attr.name], (long unsigned int)attr.value->integer);
                         break;
                     default:
                         throw TypeError("flake input attribute '%s' is %s while a string, Boolean, or integer is expected",
-                            attr.name, showType(*attr.value));
+                            state.symbols[attr.name], showType(*attr.value));
                 }
             }
         } catch (Error & e) {
-            e.addTrace(*attr.pos, hintfmt("in flake attribute '%s'", attr.name));
+            e.addTrace(
+                state.positions[attr.pos],
+                hintfmt("in flake attribute '%s'", state.symbols[attr.name]));
             throw;
         }
     }
@@ -150,13 +152,13 @@ static FlakeInput parseFlakeInput(EvalState & state,
         try {
             input.ref = FlakeRef::fromAttrs(attrs);
         } catch (Error & e) {
-            e.addTrace(pos, hintfmt("in flake input"));
+            e.addTrace(state.positions[pos], hintfmt("in flake input"));
             throw;
         }
     else {
         attrs.erase("url");
         if (!attrs.empty())
-            throw Error("unexpected flake input attribute '%s', at %s", attrs.begin()->first, pos);
+            throw Error("unexpected flake input attribute '%s', at %s", attrs.begin()->first, state.positions[pos]);
         if (url)
             input.ref = parseFlakeRef(*url, baseDir, true, input.isFlake);
     }
@@ -168,7 +170,7 @@ static FlakeInput parseFlakeInput(EvalState & state,
 }
 
 static std::map<FlakeId, FlakeInput> parseFlakeInputs(
-    EvalState & state, Value * value, const Pos & pos,
+    EvalState & state, Value * value, const PosIdx pos,
     const std::optional<Path> & baseDir, InputPath lockRootPath)
 {
     std::map<FlakeId, FlakeInput> inputs;
@@ -176,11 +178,11 @@ static std::map<FlakeId, FlakeInput> parseFlakeInputs(
     expectType(state, nAttrs, *value, pos);
 
     for (nix::Attr & inputAttr : *(*value).attrs) {
-        inputs.emplace(inputAttr.name,
+        inputs.emplace(state.symbols[inputAttr.name],
             parseFlakeInput(state,
-                inputAttr.name,
+                state.symbols[inputAttr.name],
                 inputAttr.value,
-                *inputAttr.pos,
+                inputAttr.pos,
                 baseDir,
                 lockRootPath));
     }
@@ -218,28 +220,28 @@ static Flake getFlake(
     Value vInfo;
     state.evalFile(flakeFile, vInfo, true); // FIXME: symlink attack
 
-    expectType(state, nAttrs, vInfo, Pos(foFile, state.symbols.create(flakeFile), 0, 0));
+    expectType(state, nAttrs, vInfo, state.positions.add({flakeFile, foFile}, 0, 0));
 
     if (auto description = vInfo.attrs->get(state.sDescription)) {
-        expectType(state, nString, *description->value, *description->pos);
+        expectType(state, nString, *description->value, description->pos);
         flake.description = description->value->string.s;
     }
 
     auto sInputs = state.symbols.create("inputs");
 
     if (auto inputs = vInfo.attrs->get(sInputs))
-        flake.inputs = parseFlakeInputs(state, inputs->value, *inputs->pos, flakeDir, lockRootPath);
+        flake.inputs = parseFlakeInputs(state, inputs->value, inputs->pos, flakeDir, lockRootPath);
 
     auto sOutputs = state.symbols.create("outputs");
 
     if (auto outputs = vInfo.attrs->get(sOutputs)) {
-        expectType(state, nFunction, *outputs->value, *outputs->pos);
+        expectType(state, nFunction, *outputs->value, outputs->pos);
 
         if (outputs->value->isLambda() && outputs->value->lambda.fun->hasFormals()) {
             for (auto & formal : outputs->value->lambda.fun->formals->formals) {
                 if (formal.name != state.sSelf)
-                    flake.inputs.emplace(formal.name, FlakeInput {
+                    flake.inputs.emplace(state.symbols[formal.name], FlakeInput {
-                        .ref = parseFlakeRef(formal.name)
+                        .ref = parseFlakeRef(state.symbols[formal.name])
                     });
             }
         }
@@ -250,35 +252,41 @@ static Flake getFlake(
     auto sNixConfig = state.symbols.create("nixConfig");
 
     if (auto nixConfig = vInfo.attrs->get(sNixConfig)) {
-        expectType(state, nAttrs, *nixConfig->value, *nixConfig->pos);
+        expectType(state, nAttrs, *nixConfig->value, nixConfig->pos);
 
         for (auto & setting : *nixConfig->value->attrs) {
-            forceTrivialValue(state, *setting.value, *setting.pos);
+            forceTrivialValue(state, *setting.value, setting.pos);
             if (setting.value->type() == nString)
-                flake.config.settings.insert({setting.name, std::string(state.forceStringNoCtx(*setting.value, *setting.pos))});
+                flake.config.settings.emplace(
+                    state.symbols[setting.name],
+                    std::string(state.forceStringNoCtx(*setting.value, setting.pos)));
             else if (setting.value->type() == nPath) {
                 PathSet emptyContext = {};
                 flake.config.settings.emplace(
-                    setting.name,
+                    state.symbols[setting.name],
-                    state.coerceToString(*setting.pos, *setting.value, emptyContext, false, true, true) .toOwned());
+                    state.coerceToString(setting.pos, *setting.value, emptyContext, false, true, true) .toOwned());
             }
             else if (setting.value->type() == nInt)
-                flake.config.settings.insert({setting.name, state.forceInt(*setting.value, *setting.pos)});
+                flake.config.settings.emplace(
+                    state.symbols[setting.name],
+                    state.forceInt(*setting.value, setting.pos));
             else if (setting.value->type() == nBool)
-                flake.config.settings.insert({setting.name, Explicit<bool> { state.forceBool(*setting.value, *setting.pos) }});
+                flake.config.settings.emplace(
+                    state.symbols[setting.name],
+                    Explicit<bool> { state.forceBool(*setting.value, setting.pos) });
             else if (setting.value->type() == nList) {
                 std::vector<std::string> ss;
                 for (auto elem : setting.value->listItems()) {
                     if (elem->type() != nString)
                         throw TypeError("list element in flake configuration setting '%s' is %s while a string is expected",
-                            setting.name, showType(*setting.value));
+                            state.symbols[setting.name], showType(*setting.value));
-                    ss.emplace_back(state.forceStringNoCtx(*elem, *setting.pos));
+                    ss.emplace_back(state.forceStringNoCtx(*elem, setting.pos));
                 }
-                flake.config.settings.insert({setting.name, ss});
+                flake.config.settings.emplace(state.symbols[setting.name], ss);
             }
             else
                 throw TypeError("flake configuration setting '%s' is %s",
-                    setting.name, showType(*setting.value));
+                    state.symbols[setting.name], showType(*setting.value));
         }
     }
 
@@ -288,7 +296,7 @@ static Flake getFlake(
             attr.name != sOutputs &&
             attr.name != sNixConfig)
             throw Error("flake '%s' has an unsupported attribute '%s', at %s",
-                lockedRef, attr.name, *attr.pos);
+                lockedRef, state.symbols[attr.name], state.positions[attr.pos]);
     }
 
     return flake;
@@ -704,14 +712,12 @@ void callFlake(EvalState & state,
     state.callFunction(*vTmp2, *vRootSubdir, vRes, noPos);
 }
 
-static void prim_getFlake(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_getFlake(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
-    state.requireExperimentalFeatureOnEvaluation(Xp::Flakes, "builtins.getFlake", pos);
-
     std::string flakeRefS(state.forceStringNoCtx(*args[0], pos));
     auto flakeRef = parseFlakeRef(flakeRefS, {}, true);
     if (evalSettings.pureEval && !flakeRef.input.isLocked())
-        throw Error("cannot call 'getFlake' on unlocked flake reference '%s', at %s (use --impure to override)", flakeRefS, pos);
+        throw Error("cannot call 'getFlake' on unlocked flake reference '%s', at %s (use --impure to override)", flakeRefS, state.positions[pos]);
 
     callFlake(state,
         lockFlake(state, flakeRef,
@@ -723,7 +729,30 @@ static void prim_getFlake(EvalState & state, const Pos & pos, Value * * args, Va
         v);
 }
 
-static RegisterPrimOp r2("__getFlake", 1, prim_getFlake);
+static RegisterPrimOp r2({
+    .name =  "__getFlake",
+    .args = {"args"},
+    .doc = R"(
+      Fetch a flake from a flake reference, and return its output attributes and some metadata. For example:
+
+      ```nix
+      (builtins.getFlake "nix/55bc52401966fbffa525c574c14f67b00bc4fb3a").packages.x86_64-linux.nix
+      ```
+
+      Unless impure evaluation is allowed (`--impure`), the flake reference
+      must be "locked", e.g. contain a Git revision or content hash. An
+      example of an unlocked usage is:
+
+      ```nix
+      (builtins.getFlake "github:edolstra/dwarffs").rev
+      ```
+
+      This function is only available if you enable the experimental feature
+      `flakes`.
+    )",
+    .fun = prim_getFlake,
+    .experimentalFeature = Xp::Flakes,
+});
 
 }
 

src/libexpr/function-trace.hh

@@ -8,7 +8,7 @@ namespace nix {
 
 struct FunctionCallTrace
 {
-    const Pos & pos;
+    const Pos pos;
     FunctionCallTrace(const Pos & pos);
     ~FunctionCallTrace();
 };

src/libexpr/get-drvs.cc

@@ -1,6 +1,7 @@
 #include "get-drvs.hh"
 #include "util.hh"
 #include "eval-inline.hh"
+#include "derivations.hh"
 #include "store-api.hh"
 #include "path-with-outputs.hh"
 
@@ -60,7 +61,7 @@ std::string DrvInfo::querySystem() const
 {
     if (system == "" && attrs) {
         auto i = attrs->find(state->sSystem);
-        system = i == attrs->end() ? "unknown" : state->forceStringNoCtx(*i->value, *i->pos);
+        system = i == attrs->end() ? "unknown" : state->forceStringNoCtx(*i->value, i->pos);
     }
     return system;
 }
@@ -74,7 +75,7 @@ std::optional<StorePath> DrvInfo::queryDrvPath() const
         if (i == attrs->end())
             drvPath = {std::nullopt};
         else
-            drvPath = {state->coerceToStorePath(*i->pos, *i->value, context)};
+            drvPath = {state->coerceToStorePath(i->pos, *i->value, context)};
     }
     return drvPath.value_or(std::nullopt);
 }
@@ -94,7 +95,7 @@ StorePath DrvInfo::queryOutPath() const
         Bindings::iterator i = attrs->find(state->sOutPath);
         PathSet context;
         if (i != attrs->end())
-            outPath = state->coerceToStorePath(*i->pos, *i->value, context);
+            outPath = state->coerceToStorePath(i->pos, *i->value, context);
     }
     if (!outPath)
         throw UnimplementedError("CA derivations are not yet supported");
@@ -102,30 +103,34 @@ StorePath DrvInfo::queryOutPath() const
 }
 
 
-DrvInfo::Outputs DrvInfo::queryOutputs(bool onlyOutputsToInstall)
+DrvInfo::Outputs DrvInfo::queryOutputs(bool withPaths, bool onlyOutputsToInstall)
 {
     if (outputs.empty()) {
         /* Get the ‘outputs’ list. */
         Bindings::iterator i;
         if (attrs && (i = attrs->find(state->sOutputs)) != attrs->end()) {
-            state->forceList(*i->value, *i->pos);
+            state->forceList(*i->value, i->pos);
 
             /* For each output... */
             for (auto elem : i->value->listItems()) {
-                /* Evaluate the corresponding set. */
+                std::string output(state->forceStringNoCtx(*elem, i->pos));
-                std::string name(state->forceStringNoCtx(*elem, *i->pos));
-                Bindings::iterator out = attrs->find(state->symbols.create(name));
-                if (out == attrs->end()) continue; // FIXME: throw error?
-                state->forceAttrs(*out->value, *i->pos);
 
-                /* And evaluate its ‘outPath’ attribute. */
+                if (withPaths) {
-                Bindings::iterator outPath = out->value->attrs->find(state->sOutPath);
+                    /* Evaluate the corresponding set. */
-                if (outPath == out->value->attrs->end()) continue; // FIXME: throw error?
+                    Bindings::iterator out = attrs->find(state->symbols.create(output));
-                PathSet context;
+                    if (out == attrs->end()) continue; // FIXME: throw error?
-                outputs.emplace(name, state->coerceToStorePath(*outPath->pos, *outPath->value, context));
+                    state->forceAttrs(*out->value, i->pos);
+
+                    /* And evaluate its ‘outPath’ attribute. */
+                    Bindings::iterator outPath = out->value->attrs->find(state->sOutPath);
+                    if (outPath == out->value->attrs->end()) continue; // FIXME: throw error?
+                    PathSet context;
+                    outputs.emplace(output, state->coerceToStorePath(outPath->pos, *outPath->value, context));
+                } else
+                    outputs.emplace(output, std::nullopt);
             }
         } else
-            outputs.emplace("out", queryOutPath());
+            outputs.emplace("out", withPaths ? std::optional{queryOutPath()} : std::nullopt);
     }
     if (!onlyOutputsToInstall || !attrs)
         return outputs;
@@ -163,7 +168,7 @@ Bindings * DrvInfo::getMeta()
     if (!attrs) return 0;
     Bindings::iterator a = attrs->find(state->sMeta);
     if (a == attrs->end()) return 0;
-    state->forceAttrs(*a->value, *a->pos);
+    state->forceAttrs(*a->value, a->pos);
     meta = a->value->attrs;
     return meta;
 }
@@ -174,7 +179,7 @@ StringSet DrvInfo::queryMetaNames()
     StringSet res;
     if (!getMeta()) return res;
     for (auto & i : *meta)
-        res.insert(i.name);
+        res.emplace(state->symbols[i.name]);
     return res;
 }
 
@@ -264,7 +269,7 @@ void DrvInfo::setMeta(const std::string & name, Value * v)
 {
     getMeta();
     auto attrs = state->buildBindings(1 + (meta ? meta->size() : 0));
-    Symbol sym = state->symbols.create(name);
+    auto sym = state->symbols.create(name);
     if (meta)
         for (auto i : *meta)
             if (i.name != sym)
@@ -351,11 +356,11 @@ static void getDerivations(EvalState & state, Value & vIn,
            there are names clashes between derivations, the derivation
            bound to the attribute with the "lower" name should take
            precedence). */
-        for (auto & i : v.attrs->lexicographicOrder()) {
+        for (auto & i : v.attrs->lexicographicOrder(state.symbols)) {
-            debug("evaluating attribute '%1%'", i->name);
+            debug("evaluating attribute '%1%'", state.symbols[i->name]);
-            if (!std::regex_match(std::string(i->name), attrRegex))
+            if (!std::regex_match(std::string(state.symbols[i->name]), attrRegex))
                 continue;
-            std::string pathPrefix2 = addToPath(pathPrefix, i->name);
+            std::string pathPrefix2 = addToPath(pathPrefix, state.symbols[i->name]);
             if (combineChannels)
                 getDerivations(state, *i->value, pathPrefix2, autoArgs, drvs, done, ignoreAssertionFailures);
             else if (getDerivation(state, *i->value, pathPrefix2, drvs, done, ignoreAssertionFailures)) {
@@ -364,7 +369,7 @@ static void getDerivations(EvalState & state, Value & vIn,
                    `recurseForDerivations = true' attribute. */
                 if (i->value->type() == nAttrs) {
                     Bindings::iterator j = i->value->attrs->find(state.sRecurseForDerivations);
-                    if (j != i->value->attrs->end() && state.forceBool(*j->value, *j->pos))
+                    if (j != i->value->attrs->end() && state.forceBool(*j->value, j->pos))
                         getDerivations(state, *i->value, pathPrefix2, autoArgs, drvs, done, ignoreAssertionFailures);
                 }
             }

src/libexpr/get-drvs.hh

@@ -13,7 +13,7 @@ namespace nix {
 struct DrvInfo
 {
 public:
-    typedef std::map<std::string, StorePath> Outputs;
+    typedef std::map<std::string, std::optional<StorePath>> Outputs;
 
 private:
     EvalState * state;
@@ -46,8 +46,9 @@ public:
     StorePath requireDrvPath() const;
     StorePath queryOutPath() const;
     std::string queryOutputName() const;
-    /** Return the list of outputs. The "outputs to install" are determined by `meta.outputsToInstall`. */
+    /** Return the unordered map of output names to (optional) output paths.
-    Outputs queryOutputs(bool onlyOutputsToInstall = false);
+     * The "outputs to install" are determined by `meta.outputsToInstall`. */
+    Outputs queryOutputs(bool withPaths = true, bool onlyOutputsToInstall = false);
 
     StringSet queryMetaNames();
     Value * queryMeta(const std::string & name);

src/libexpr/lexer.l

@@ -28,6 +28,13 @@ using namespace nix;
 
 namespace nix {
 
+static inline PosIdx makeCurPos(const YYLTYPE & loc, ParseData * data)
+{
+    return data->state.positions.add(data->origin, loc.first_line, loc.first_column);
+}
+
+#define CUR_POS makeCurPos(*yylloc, data)
+
 // backup to recover from yyless(0)
 YYLTYPE prev_yylloc;
 
@@ -37,7 +44,6 @@ static void initLoc(YYLTYPE * loc)
     loc->first_column = loc->last_column = 1;
 }
 
-
 static void adjustLoc(YYLTYPE * loc, const char * s, size_t len)
 {
     prev_yylloc = *loc;
@@ -147,14 +153,20 @@ or          { return OR_KW; }
               try {
                   yylval->n = boost::lexical_cast<int64_t>(yytext);
               } catch (const boost::bad_lexical_cast &) {
-                  throw ParseError("invalid integer '%1%'", yytext);
+                  throw ParseError({
+                      .msg = hintfmt("invalid integer '%1%'", yytext),
+                      .errPos = data->state.positions[CUR_POS],
+                  });
               }
               return INT;
             }
 {FLOAT}     { errno = 0;
               yylval->nf = strtod(yytext, 0);
               if (errno != 0)
-                  throw ParseError("invalid float '%1%'", yytext);
+                  throw ParseError({
+                      .msg = hintfmt("invalid float '%1%'", yytext),
+                      .errPos = data->state.positions[CUR_POS],
+                  });
               return FLOAT;
             }
 
@@ -280,7 +292,10 @@ or          { return OR_KW; }
 
 <INPATH_SLASH>{ANY} |
 <INPATH_SLASH><<EOF>> {
-  throw ParseError("path has a trailing slash");
+  throw ParseError({
+      .msg = hintfmt("path has a trailing slash"),
+      .errPos = data->state.positions[CUR_POS],
+  });
 }
 
 {SPATH}     { yylval->path = {yytext, (size_t) yyleng}; return SPATH; }

src/libexpr/nixexpr.cc

@@ -1,5 +1,7 @@
 #include "nixexpr.hh"
 #include "derivations.hh"
+#include "eval.hh"
+#include "symbol-table.hh"
 #include "util.hh"
 
 #include <cstdlib>
@@ -10,12 +12,6 @@ namespace nix {
 
 /* Displaying abstract syntax trees. */
 
-std::ostream & operator << (std::ostream & str, const Expr & e)
-{
-    e.show(str);
-    return str;
-}
-
 static void showString(std::ostream & str, std::string_view s)
 {
     str << '"';
@@ -54,81 +50,101 @@ static void showId(std::ostream & str, std::string_view s)
 
 std::ostream & operator << (std::ostream & str, const Symbol & sym)
 {
-    showId(str, *sym.s);
+    showId(str, sym.s);
     return str;
 }
 
-void Expr::show(std::ostream & str) const
+void Expr::show(const SymbolTable & symbols, std::ostream & str) const
 {
     abort();
 }
 
-void ExprInt::show(std::ostream & str) const
+void ExprInt::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << n;
 }
 
-void ExprFloat::show(std::ostream & str) const
+void ExprFloat::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << nf;
 }
 
-void ExprString::show(std::ostream & str) const
+void ExprString::show(const SymbolTable & symbols, std::ostream & str) const
 {
     showString(str, s);
 }
 
-void ExprPath::show(std::ostream & str) const
+void ExprPath::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << s;
 }
 
-void ExprVar::show(std::ostream & str) const
+void ExprVar::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << name;
+    str << symbols[name];
 }
 
-void ExprSelect::show(std::ostream & str) const
+void ExprSelect::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "(" << *e << ")." << showAttrPath(attrPath);
+    str << "(";
-    if (def) str << " or (" << *def << ")";
+    e->show(symbols, str);
+    str << ")." << showAttrPath(symbols, attrPath);
+    if (def) {
+        str << " or (";
+        def->show(symbols, str);
+        str << ")";
+    }
 }
 
-void ExprOpHasAttr::show(std::ostream & str) const
+void ExprOpHasAttr::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "((" << *e << ") ? " << showAttrPath(attrPath) << ")";
+    str << "((";
+    e->show(symbols, str);
+    str << ") ? " << showAttrPath(symbols, attrPath) << ")";
 }
 
-void ExprAttrs::show(std::ostream & str) const
+void ExprAttrs::show(const SymbolTable & symbols, std::ostream & str) const
 {
     if (recursive) str << "rec ";
     str << "{ ";
     typedef const decltype(attrs)::value_type * Attr;
     std::vector<Attr> sorted;
     for (auto & i : attrs) sorted.push_back(&i);
-        std::sort(sorted.begin(), sorted.end(), [](Attr a, Attr b) {
+    std::sort(sorted.begin(), sorted.end(), [&](Attr a, Attr b) {
-            return (const std::string &) a->first < (const std::string &) b->first;
+        std::string_view sa = symbols[a->first], sb = symbols[b->first];
-        });
+        return sa < sb;
+    });
     for (auto & i : sorted) {
         if (i->second.inherited)
-            str << "inherit " << i->first << " " << "; ";
+            str << "inherit " << symbols[i->first] << " " << "; ";
-        else
+        else {
-            str << i->first << " = " << *i->second.e << "; ";
+            str << symbols[i->first] << " = ";
+            i->second.e->show(symbols, str);
+            str << "; ";
+        }
+    }
+    for (auto & i : dynamicAttrs) {
+        str << "\"${";
+        i.nameExpr->show(symbols, str);
+        str << "}\" = ";
+        i.valueExpr->show(symbols, str);
+        str << "; ";
     }
-    for (auto & i : dynamicAttrs)
-        str << "\"${" << *i.nameExpr << "}\" = " << *i.valueExpr << "; ";
     str << "}";
 }
 
-void ExprList::show(std::ostream & str) const
+void ExprList::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << "[ ";
-    for (auto & i : elems)
+    for (auto & i : elems) {
-        str << "(" << *i << ") ";
+        str << "(";
+        i->show(symbols, str);
+        str << ") ";
+    }
     str << "]";
 }
 
-void ExprLambda::show(std::ostream & str) const
+void ExprLambda::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << "(";
     if (hasFormals()) {
@@ -136,74 +152,100 @@ void ExprLambda::show(std::ostream & str) const
         bool first = true;
         for (auto & i : formals->formals) {
             if (first) first = false; else str << ", ";
-            str << i.name;
+            str << symbols[i.name];
-            if (i.def) str << " ? " << *i.def;
+            if (i.def) {
+                str << " ? ";
+                i.def->show(symbols, str);
+            }
         }
         if (formals->ellipsis) {
             if (!first) str << ", ";
             str << "...";
         }
         str << " }";
-        if (!arg.empty()) str << " @ ";
+        if (arg) str << " @ ";
     }
-    if (!arg.empty()) str << arg;
+    if (arg) str << symbols[arg];
-    str << ": " << *body << ")";
+    str << ": ";
+    body->show(symbols, str);
+    str << ")";
 }
 
-void ExprCall::show(std::ostream & str) const
+void ExprCall::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << '(' << *fun;
+    str << '(';
+    fun->show(symbols, str);
     for (auto e : args) {
         str <<  ' ';
-        str << *e;
+        e->show(symbols, str);
     }
     str << ')';
 }
 
-void ExprLet::show(std::ostream & str) const
+void ExprLet::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << "(let ";
     for (auto & i : attrs->attrs)
         if (i.second.inherited) {
-            str << "inherit " << i.first << "; ";
+            str << "inherit " << symbols[i.first] << "; ";
         }
-        else
+        else {
-            str << i.first << " = " << *i.second.e << "; ";
+            str << symbols[i.first] << " = ";
-    str << "in " << *body << ")";
+            i.second.e->show(symbols, str);
+            str << "; ";
+        }
+    str << "in ";
+    body->show(symbols, str);
+    str << ")";
 }
 
-void ExprWith::show(std::ostream & str) const
+void ExprWith::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "(with " << *attrs << "; " << *body << ")";
+    str << "(with ";
+    attrs->show(symbols, str);
+    str << "; ";
+    body->show(symbols, str);
+    str << ")";
 }
 
-void ExprIf::show(std::ostream & str) const
+void ExprIf::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "(if " << *cond << " then " << *then << " else " << *else_ << ")";
+    str << "(if ";
+    cond->show(symbols, str);
+    str << " then ";
+    then->show(symbols, str);
+    str << " else ";
+    else_->show(symbols, str);
+    str << ")";
 }
 
-void ExprAssert::show(std::ostream & str) const
+void ExprAssert::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "assert " << *cond << "; " << *body;
+    str << "assert ";
+    cond->show(symbols, str);
+    str << "; ";
+    body->show(symbols, str);
 }
 
-void ExprOpNot::show(std::ostream & str) const
+void ExprOpNot::show(const SymbolTable & symbols, std::ostream & str) const
 {
-    str << "(! " << *e << ")";
+    str << "(! ";
+    e->show(symbols, str);
+    str << ")";
 }
 
-void ExprConcatStrings::show(std::ostream & str) const
+void ExprConcatStrings::show(const SymbolTable & symbols, std::ostream & str) const
 {
     bool first = true;
     str << "(";
     for (auto & i : *es) {
         if (first) first = false; else str << " + ";
-        str << *i.second;
+        i.second->show(symbols, str);
     }
     str << ")";
 }
 
-void ExprPos::show(std::ostream & str) const
+void ExprPos::show(const SymbolTable & symbols, std::ostream & str) const
 {
     str << "__curPos";
 }
@@ -234,48 +276,49 @@ std::ostream & operator << (std::ostream & str, const Pos & pos)
 }
 
 
-std::string showAttrPath(const AttrPath & attrPath)
+std::string showAttrPath(const SymbolTable & symbols, const AttrPath & attrPath)
 {
     std::ostringstream out;
     bool first = true;
     for (auto & i : attrPath) {
         if (!first) out << '.'; else first = false;
-        if (i.symbol.set())
+        if (i.symbol)
-            out << i.symbol;
+            out << symbols[i.symbol];
-        else
+        else {
-            out << "\"${" << *i.expr << "}\"";
+            out << "\"${";
+            i.expr->show(symbols, out);
+            out << "}\"";
+        }
     }
     return out.str();
 }
 
 
-Pos noPos;
-
 
 /* Computing levels/displacements for variables. */
 
-void Expr::bindVars(const StaticEnv & env)
+void Expr::bindVars(const EvalState & es, const StaticEnv & env)
 {
     abort();
 }
 
-void ExprInt::bindVars(const StaticEnv & env)
+void ExprInt::bindVars(const EvalState & es, const StaticEnv & env)
 {
 }
 
-void ExprFloat::bindVars(const StaticEnv & env)
+void ExprFloat::bindVars(const EvalState & es, const StaticEnv & env)
 {
 }
 
-void ExprString::bindVars(const StaticEnv & env)
+void ExprString::bindVars(const EvalState & es, const StaticEnv & env)
 {
 }
 
-void ExprPath::bindVars(const StaticEnv & env)
+void ExprPath::bindVars(const EvalState & es, const StaticEnv & env)
 {
 }
 
-void ExprVar::bindVars(const StaticEnv & env)
+void ExprVar::bindVars(const EvalState & es, const StaticEnv & env)
 {
     /* Check whether the variable appears in the environment.  If so,
        set its level and displacement. */
@@ -301,31 +344,31 @@ void ExprVar::bindVars(const StaticEnv & env)
        "undefined variable" error now. */
     if (withLevel == -1)
         throw UndefinedVarError({
-            .msg = hintfmt("undefined variable '%1%'", name),
+            .msg = hintfmt("undefined variable '%1%'", es.symbols[name]),
-            .errPos = pos
+            .errPos = es.positions[pos]
         });
     fromWith = true;
     this->level = withLevel;
 }
 
-void ExprSelect::bindVars(const StaticEnv & env)
+void ExprSelect::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    e->bindVars(env);
+    e->bindVars(es, env);
-    if (def) def->bindVars(env);
+    if (def) def->bindVars(es, env);
     for (auto & i : attrPath)
-        if (!i.symbol.set())
+        if (!i.symbol)
-            i.expr->bindVars(env);
+            i.expr->bindVars(es, env);
 }
 
-void ExprOpHasAttr::bindVars(const StaticEnv & env)
+void ExprOpHasAttr::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    e->bindVars(env);
+    e->bindVars(es, env);
     for (auto & i : attrPath)
-        if (!i.symbol.set())
+        if (!i.symbol)
-            i.expr->bindVars(env);
+            i.expr->bindVars(es, env);
 }
 
-void ExprAttrs::bindVars(const StaticEnv & env)
+void ExprAttrs::bindVars(const EvalState & es, const StaticEnv & env)
 {
     const StaticEnv * dynamicEnv = &env;
     StaticEnv newEnv(false, &env, recursive ? attrs.size() : 0);
@@ -340,35 +383,35 @@ void ExprAttrs::bindVars(const StaticEnv & env)
         // No need to sort newEnv since attrs is in sorted order.
 
         for (auto & i : attrs)
-            i.second.e->bindVars(i.second.inherited ? env : newEnv);
+            i.second.e->bindVars(es, i.second.inherited ? env : newEnv);
     }
 
     else
         for (auto & i : attrs)
-            i.second.e->bindVars(env);
+            i.second.e->bindVars(es, env);
 
     for (auto & i : dynamicAttrs) {
-        i.nameExpr->bindVars(*dynamicEnv);
+        i.nameExpr->bindVars(es, *dynamicEnv);
-        i.valueExpr->bindVars(*dynamicEnv);
+        i.valueExpr->bindVars(es, *dynamicEnv);
     }
 }
 
-void ExprList::bindVars(const StaticEnv & env)
+void ExprList::bindVars(const EvalState & es, const StaticEnv & env)
 {
     for (auto & i : elems)
-        i->bindVars(env);
+        i->bindVars(es, env);
 }
 
-void ExprLambda::bindVars(const StaticEnv & env)
+void ExprLambda::bindVars(const EvalState & es, const StaticEnv & env)
 {
     StaticEnv newEnv(
         false, &env,
         (hasFormals() ? formals->formals.size() : 0) +
-        (arg.empty() ? 0 : 1));
+        (!arg ? 0 : 1));
 
     Displacement displ = 0;
 
-    if (!arg.empty()) newEnv.vars.emplace_back(arg, displ++);
+    if (arg) newEnv.vars.emplace_back(arg, displ++);
 
     if (hasFormals()) {
         for (auto & i : formals->formals)
@@ -377,20 +420,20 @@ void ExprLambda::bindVars(const StaticEnv & env)
         newEnv.sort();
 
         for (auto & i : formals->formals)
-            if (i.def) i.def->bindVars(newEnv);
+            if (i.def) i.def->bindVars(es, newEnv);
     }
 
-    body->bindVars(newEnv);
+    body->bindVars(es, newEnv);
 }
 
-void ExprCall::bindVars(const StaticEnv & env)
+void ExprCall::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    fun->bindVars(env);
+    fun->bindVars(es, env);
     for (auto e : args)
-        e->bindVars(env);
+        e->bindVars(es, env);
 }
 
-void ExprLet::bindVars(const StaticEnv & env)
+void ExprLet::bindVars(const EvalState & es, const StaticEnv & env)
 {
     StaticEnv newEnv(false, &env, attrs->attrs.size());
 
@@ -401,12 +444,12 @@ void ExprLet::bindVars(const StaticEnv & env)
     // No need to sort newEnv since attrs->attrs is in sorted order.
 
     for (auto & i : attrs->attrs)
-        i.second.e->bindVars(i.second.inherited ? env : newEnv);
+        i.second.e->bindVars(es, i.second.inherited ? env : newEnv);
 
-    body->bindVars(newEnv);
+    body->bindVars(es, newEnv);
 }
 
-void ExprWith::bindVars(const StaticEnv & env)
+void ExprWith::bindVars(const EvalState & es, const StaticEnv & env)
 {
     /* Does this `with' have an enclosing `with'?  If so, record its
        level so that `lookupVar' can look up variables in the previous
@@ -420,57 +463,60 @@ void ExprWith::bindVars(const StaticEnv & env)
             break;
         }
 
-    attrs->bindVars(env);
+    attrs->bindVars(es, env);
     StaticEnv newEnv(true, &env);
-    body->bindVars(newEnv);
+    body->bindVars(es, newEnv);
 }
 
-void ExprIf::bindVars(const StaticEnv & env)
+void ExprIf::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    cond->bindVars(env);
+    cond->bindVars(es, env);
-    then->bindVars(env);
+    then->bindVars(es, env);
-    else_->bindVars(env);
+    else_->bindVars(es, env);
 }
 
-void ExprAssert::bindVars(const StaticEnv & env)
+void ExprAssert::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    cond->bindVars(env);
+    cond->bindVars(es, env);
-    body->bindVars(env);
+    body->bindVars(es, env);
 }
 
-void ExprOpNot::bindVars(const StaticEnv & env)
+void ExprOpNot::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    e->bindVars(env);
+    e->bindVars(es, env);
 }
 
-void ExprConcatStrings::bindVars(const StaticEnv & env)
+void ExprConcatStrings::bindVars(const EvalState & es, const StaticEnv & env)
 {
-    for (auto & i : *es)
+    for (auto & i : *this->es)
-        i.second->bindVars(env);
+        i.second->bindVars(es, env);
 }
 
-void ExprPos::bindVars(const StaticEnv & env)
+void ExprPos::bindVars(const EvalState & es, const StaticEnv & env)
 {
 }
 
 
 /* Storing function names. */
 
-void Expr::setName(Symbol & name)
+void Expr::setName(SymbolIdx name)
 {
 }
 
 
-void ExprLambda::setName(Symbol & name)
+void ExprLambda::setName(SymbolIdx name)
 {
     this->name = name;
     body->setName(name);
 }
 
 
-std::string ExprLambda::showNamePos() const
+std::string ExprLambda::showNamePos(const EvalState & state) const
 {
-    return fmt("%1% at %2%", name.set() ? "'" + (std::string) name + "'" : "anonymous function", pos);
+    std::string id(name
+        ? concatStrings("'", state.symbols[name], "'")
+        : "anonymous function");
+    return fmt("%1% at %2%", id, state.positions[pos]);
 }
 
 
@@ -480,8 +526,7 @@ std::string ExprLambda::showNamePos() const
 size_t SymbolTable::totalSize() const
 {
     size_t n = 0;
-    for (auto & i : store)
+    dump([&] (const Symbol & s) { n += std::string_view(s).size(); });
-        n += i.size();
     return n;
 }
 

src/libexpr/nixexpr.hh

@@ -1,8 +1,12 @@
 #pragma once
 
+#include <map>
+#include <vector>
+
 #include "value.hh"
 #include "symbol-table.hh"
 #include "error.hh"
+#include "chunked-vector.hh"
 
 
 namespace nix {
@@ -23,33 +27,92 @@ MakeError(RestrictedPathError, Error);
 
 struct Pos
 {
+    std::string file;
     FileOrigin origin;
-    Symbol file;
+    uint32_t line;
-    unsigned int line, column;
+    uint32_t column;
 
-    Pos() : origin(foString), line(0), column(0) { }
+    explicit operator bool() const { return line > 0; }
-    Pos(FileOrigin origin, const Symbol & file, unsigned int line, unsigned int column)
+};
-        : origin(origin), file(file), line(line), column(column) { }
 
-    operator bool() const
+class PosIdx {
+    friend class PosTable;
+
+private:
+    uint32_t id;
+
+    explicit PosIdx(uint32_t id): id(id) {}
+
+public:
+    PosIdx() : id(0) {}
+
+    explicit operator bool() const { return id > 0; }
+
+    bool operator<(const PosIdx other) const { return id < other.id; }
+};
+
+class PosTable
+{
+public:
+    class Origin {
+        friend PosTable;
+    private:
+        // must always be invalid by default, add() replaces this with the actual value.
+        // subsequent add() calls use this index as a token to quickly check whether the
+        // current origins.back() can be reused or not.
+        mutable uint32_t idx = std::numeric_limits<uint32_t>::max();
+
+        explicit Origin(uint32_t idx): idx(idx), file{}, origin{} {}
+
+    public:
+        const std::string file;
+        const FileOrigin origin;
+
+        Origin(std::string file, FileOrigin origin): file(std::move(file)), origin(origin) {}
+    };
+
+    struct Offset {
+        uint32_t line, column;
+    };
+
+private:
+    std::vector<Origin> origins;
+    ChunkedVector<Offset, 8192> offsets;
+
+public:
+    PosTable(): offsets(1024)
     {
-        return line != 0;
+        origins.reserve(1024);
     }
 
-    bool operator < (const Pos & p2) const
+    PosIdx add(const Origin & origin, uint32_t line, uint32_t column)
     {
-        if (!line) return p2.line;
+        const auto idx = offsets.add({line, column}).second;
-        if (!p2.line) return false;
+        if (origins.empty() || origins.back().idx != origin.idx) {
-        int d = ((const std::string &) file).compare((const std::string &) p2.file);
+            origin.idx = idx;
-        if (d < 0) return true;
+            origins.push_back(origin);
-        if (d > 0) return false;
+        }
-        if (line < p2.line) return true;
+        return PosIdx(idx + 1);
-        if (line > p2.line) return false;
+    }
-        return column < p2.column;
+
+    Pos operator[](PosIdx p) const
+    {
+        if (p.id == 0 || p.id > offsets.size())
+            return {};
+        const auto idx = p.id - 1;
+        /* we want the last key <= idx, so we'll take prev(first key > idx).
+           this is guaranteed to never rewind origin.begin because the first
+           key is always 0. */
+        const auto pastOrigin = std::upper_bound(
+            origins.begin(), origins.end(), Origin(idx),
+            [] (const auto & a, const auto & b) { return a.idx < b.idx; });
+        const auto origin = *std::prev(pastOrigin);
+        const auto offset = offsets[idx];
+        return {origin.file, origin.origin, offset.line, offset.column};
     }
 };
 
-extern Pos noPos;
+inline PosIdx noPos = {};
 
 std::ostream & operator << (std::ostream & str, const Pos & pos);
 
@@ -63,15 +126,15 @@ struct StaticEnv;
 /* An attribute path is a sequence of attribute names. */
 struct AttrName
 {
-    Symbol symbol;
+    SymbolIdx symbol;
     Expr * expr;
-    AttrName(const Symbol & s) : symbol(s) {};
+    AttrName(const SymbolIdx & s) : symbol(s) {};
     AttrName(Expr * e) : expr(e) {};
 };
 
 typedef std::vector<AttrName> AttrPath;
 
-std::string showAttrPath(const AttrPath & attrPath);
+std::string showAttrPath(const SymbolTable & symbols, const AttrPath & attrPath);
 
 
 /* Abstract syntax of Nix expressions. */
@@ -79,19 +142,17 @@ std::string showAttrPath(const AttrPath & attrPath);
 struct Expr
 {
     virtual ~Expr() { };
-    virtual void show(std::ostream & str) const;
+    virtual void show(const SymbolTable & symbols, std::ostream & str) const;
-    virtual void bindVars(const StaticEnv & env);
+    virtual void bindVars(const EvalState & es, const StaticEnv & env);
     virtual void eval(EvalState & state, Env & env, Value & v);
     virtual Value * maybeThunk(EvalState & state, Env & env);
-    virtual void setName(Symbol & name);
+    virtual void setName(SymbolIdx name);
 };
 
-std::ostream & operator << (std::ostream & str, const Expr & e);
-
 #define COMMON_METHODS \
-    void show(std::ostream & str) const; \
+    void show(const SymbolTable & symbols, std::ostream & str) const;    \
     void eval(EvalState & state, Env & env, Value & v); \
-    void bindVars(const StaticEnv & env);
+    void bindVars(const EvalState & es, const StaticEnv & env);
 
 struct ExprInt : Expr
 {
@@ -134,8 +195,8 @@ typedef uint32_t Displacement;
 
 struct ExprVar : Expr
 {
-    Pos pos;
+    PosIdx pos;
-    Symbol name;
+    SymbolIdx name;
 
     /* Whether the variable comes from an environment (e.g. a rec, let
        or function argument) or from a "with". */
@@ -150,19 +211,19 @@ struct ExprVar : Expr
     Level level;
     Displacement displ;
 
-    ExprVar(const Symbol & name) : name(name) { };
+    ExprVar(const SymbolIdx & name) : name(name) { };
-    ExprVar(const Pos & pos, const Symbol & name) : pos(pos), name(name) { };
+    ExprVar(const PosIdx & pos, const SymbolIdx & name) : pos(pos), name(name) { };
     COMMON_METHODS
     Value * maybeThunk(EvalState & state, Env & env);
 };
 
 struct ExprSelect : Expr
 {
-    Pos pos;
+    PosIdx pos;
     Expr * e, * def;
     AttrPath attrPath;
-    ExprSelect(const Pos & pos, Expr * e, const AttrPath & attrPath, Expr * def) : pos(pos), e(e), def(def), attrPath(attrPath) { };
+    ExprSelect(const PosIdx & pos, Expr * e, const AttrPath & attrPath, Expr * def) : pos(pos), e(e), def(def), attrPath(attrPath) { };
-    ExprSelect(const Pos & pos, Expr * e, const Symbol & name) : pos(pos), e(e), def(0) { attrPath.push_back(AttrName(name)); };
+    ExprSelect(const PosIdx & pos, Expr * e, const SymbolIdx & name) : pos(pos), e(e), def(0) { attrPath.push_back(AttrName(name)); };
     COMMON_METHODS
 };
 
@@ -177,28 +238,28 @@ struct ExprOpHasAttr : Expr
 struct ExprAttrs : Expr
 {
     bool recursive;
-    Pos pos;
+    PosIdx pos;
     struct AttrDef {
         bool inherited;
         Expr * e;
-        Pos pos;
+        PosIdx pos;
         Displacement displ; // displacement
-        AttrDef(Expr * e, const Pos & pos, bool inherited=false)
+        AttrDef(Expr * e, const PosIdx & pos, bool inherited=false)
             : inherited(inherited), e(e), pos(pos) { };
         AttrDef() { };
     };
-    typedef std::map<Symbol, AttrDef> AttrDefs;
+    typedef std::map<SymbolIdx, AttrDef> AttrDefs;
     AttrDefs attrs;
     struct DynamicAttrDef {
         Expr * nameExpr, * valueExpr;
-        Pos pos;
+        PosIdx pos;
-        DynamicAttrDef(Expr * nameExpr, Expr * valueExpr, const Pos & pos)
+        DynamicAttrDef(Expr * nameExpr, Expr * valueExpr, const PosIdx & pos)
             : nameExpr(nameExpr), valueExpr(valueExpr), pos(pos) { };
     };
     typedef std::vector<DynamicAttrDef> DynamicAttrDefs;
     DynamicAttrDefs dynamicAttrs;
-    ExprAttrs(const Pos &pos) : recursive(false), pos(pos) { };
+    ExprAttrs(const PosIdx &pos) : recursive(false), pos(pos) { };
-    ExprAttrs() : recursive(false), pos(noPos) { };
+    ExprAttrs() : recursive(false) { };
     COMMON_METHODS
 };
 
@@ -211,10 +272,9 @@ struct ExprList : Expr
 
 struct Formal
 {
-    Pos pos;
+    PosIdx pos;
-    Symbol name;
+    SymbolIdx name;
     Expr * def;
-    Formal(const Pos & pos, const Symbol & name, Expr * def) : pos(pos), name(name), def(def) { };
 };
 
 struct Formals
@@ -223,18 +283,19 @@ struct Formals
     Formals_ formals;
     bool ellipsis;
 
-    bool has(Symbol arg) const {
+    bool has(SymbolIdx arg) const {
         auto it = std::lower_bound(formals.begin(), formals.end(), arg,
-            [] (const Formal & f, const Symbol & sym) { return f.name < sym; });
+            [] (const Formal & f, const SymbolIdx & sym) { return f.name < sym; });
         return it != formals.end() && it->name == arg;
     }
 
-    std::vector<Formal> lexicographicOrder() const
+    std::vector<Formal> lexicographicOrder(const SymbolTable & symbols) const
     {
         std::vector<Formal> result(formals.begin(), formals.end());
         std::sort(result.begin(), result.end(),
-            [] (const Formal & a, const Formal & b) {
+            [&] (const Formal & a, const Formal & b) {
-                return std::string_view(a.name) < std::string_view(b.name);
+                std::string_view sa = symbols[a.name], sb = symbols[b.name];
+                return sa < sb;
             });
         return result;
     }
@@ -242,17 +303,21 @@ struct Formals
 
 struct ExprLambda : Expr
 {
-    Pos pos;
+    PosIdx pos;
-    Symbol name;
+    SymbolIdx name;
-    Symbol arg;
+    SymbolIdx arg;
     Formals * formals;
     Expr * body;
-    ExprLambda(const Pos & pos, const Symbol & arg, Formals * formals, Expr * body)
+    ExprLambda(PosIdx pos, SymbolIdx arg, Formals * formals, Expr * body)
         : pos(pos), arg(arg), formals(formals), body(body)
     {
     };
-    void setName(Symbol & name);
+    ExprLambda(PosIdx pos, Formals * formals, Expr * body)
-    std::string showNamePos() const;
+        : pos(pos), formals(formals), body(body)
+    {
+    }
+    void setName(SymbolIdx name);
+    std::string showNamePos(const EvalState & state) const;
     inline bool hasFormals() const { return formals != nullptr; }
     COMMON_METHODS
 };
@@ -261,8 +326,8 @@ struct ExprCall : Expr
 {
     Expr * fun;
     std::vector<Expr *> args;
-    Pos pos;
+    PosIdx pos;
-    ExprCall(const Pos & pos, Expr * fun, std::vector<Expr *> && args)
+    ExprCall(const PosIdx & pos, Expr * fun, std::vector<Expr *> && args)
         : fun(fun), args(args), pos(pos)
     { }
     COMMON_METHODS
@@ -278,26 +343,26 @@ struct ExprLet : Expr
 
 struct ExprWith : Expr
 {
-    Pos pos;
+    PosIdx pos;
     Expr * attrs, * body;
     size_t prevWith;
-    ExprWith(const Pos & pos, Expr * attrs, Expr * body) : pos(pos), attrs(attrs), body(body) { };
+    ExprWith(const PosIdx & pos, Expr * attrs, Expr * body) : pos(pos), attrs(attrs), body(body) { };
     COMMON_METHODS
 };
 
 struct ExprIf : Expr
 {
-    Pos pos;
+    PosIdx pos;
     Expr * cond, * then, * else_;
-    ExprIf(const Pos & pos, Expr * cond, Expr * then, Expr * else_) : pos(pos), cond(cond), then(then), else_(else_) { };
+    ExprIf(const PosIdx & pos, Expr * cond, Expr * then, Expr * else_) : pos(pos), cond(cond), then(then), else_(else_) { };
     COMMON_METHODS
 };
 
 struct ExprAssert : Expr
 {
-    Pos pos;
+    PosIdx pos;
     Expr * cond, * body;
-    ExprAssert(const Pos & pos, Expr * cond, Expr * body) : pos(pos), cond(cond), body(body) { };
+    ExprAssert(const PosIdx & pos, Expr * cond, Expr * body) : pos(pos), cond(cond), body(body) { };
     COMMON_METHODS
 };
 
@@ -311,17 +376,17 @@ struct ExprOpNot : Expr
 #define MakeBinOp(name, s) \
     struct name : Expr \
     { \
-        Pos pos; \
+        PosIdx pos; \
         Expr * e1, * e2; \
         name(Expr * e1, Expr * e2) : e1(e1), e2(e2) { }; \
-        name(const Pos & pos, Expr * e1, Expr * e2) : pos(pos), e1(e1), e2(e2) { }; \
+        name(const PosIdx & pos, Expr * e1, Expr * e2) : pos(pos), e1(e1), e2(e2) { }; \
-        void show(std::ostream & str) const \
+        void show(const SymbolTable & symbols, std::ostream & str) const \
         { \
-            str << "(" << *e1 << " " s " " << *e2 << ")";   \
+            str << "("; e1->show(symbols, str); str << " " s " "; e2->show(symbols, str); str << ")"; \
         } \
-        void bindVars(const StaticEnv & env) \
+        void bindVars(const EvalState & es, const StaticEnv & env)    \
         { \
-            e1->bindVars(env); e2->bindVars(env); \
+            e1->bindVars(es, env); e2->bindVars(es, env);    \
         } \
         void eval(EvalState & state, Env & env, Value & v); \
     };
@@ -336,18 +401,18 @@ MakeBinOp(ExprOpConcatLists, "++")
 
 struct ExprConcatStrings : Expr
 {
-    Pos pos;
+    PosIdx pos;
     bool forceString;
-    std::vector<std::pair<Pos, Expr *> > * es;
+    std::vector<std::pair<PosIdx, Expr *> > * es;
-    ExprConcatStrings(const Pos & pos, bool forceString, std::vector<std::pair<Pos, Expr *> > * es)
+    ExprConcatStrings(const PosIdx & pos, bool forceString, std::vector<std::pair<PosIdx, Expr *> > * es)
         : pos(pos), forceString(forceString), es(es) { };
     COMMON_METHODS
 };
 
 struct ExprPos : Expr
 {
-    Pos pos;
+    PosIdx pos;
-    ExprPos(const Pos & pos) : pos(pos) { };
+    ExprPos(const PosIdx & pos) : pos(pos) { };
     COMMON_METHODS
 };
 
@@ -361,7 +426,7 @@ struct StaticEnv
     const StaticEnv * up;
 
     // Note: these must be in sorted order.
-    typedef std::vector<std::pair<Symbol, Displacement>> Vars;
+    typedef std::vector<std::pair<SymbolIdx, Displacement>> Vars;
     Vars vars;
 
     StaticEnv(bool isWith, const StaticEnv * up, size_t expectedSize = 0) : isWith(isWith), up(up) {
@@ -385,7 +450,7 @@ struct StaticEnv
         vars.erase(it, end);
     }
 
-    Vars::const_iterator find(const Symbol & name) const
+    Vars::const_iterator find(const SymbolIdx & name) const
     {
         Vars::value_type key(name, 0);
         auto i = std::lower_bound(vars.begin(), vars.end(), key);

src/libexpr/parser.y

@@ -32,12 +32,12 @@ namespace nix {
         SymbolTable & symbols;
         Expr * result;
         Path basePath;
-        Symbol file;
+        PosTable::Origin origin;
-        FileOrigin origin;
         std::optional<ErrorInfo> error;
-        ParseData(EvalState & state)
+        ParseData(EvalState & state, PosTable::Origin origin)
             : state(state)
             , symbols(state.symbols)
+            , origin(std::move(origin))
             { };
     };
 
@@ -77,26 +77,26 @@ using namespace nix;
 namespace nix {
 
 
-static void dupAttr(const AttrPath & attrPath, const Pos & pos, const Pos & prevPos)
+static void dupAttr(const EvalState & state, const AttrPath & attrPath, const PosIdx pos, const PosIdx prevPos)
 {
     throw ParseError({
          .msg = hintfmt("attribute '%1%' already defined at %2%",
-             showAttrPath(attrPath), prevPos),
+             showAttrPath(state.symbols, attrPath), state.positions[prevPos]),
-         .errPos = pos
+         .errPos = state.positions[pos]
     });
 }
 
-static void dupAttr(Symbol attr, const Pos & pos, const Pos & prevPos)
+static void dupAttr(const EvalState & state, SymbolIdx attr, const PosIdx pos, const PosIdx prevPos)
 {
     throw ParseError({
-        .msg = hintfmt("attribute '%1%' already defined at %2%", attr, prevPos),
+        .msg = hintfmt("attribute '%1%' already defined at %2%", state.symbols[attr], state.positions[prevPos]),
-        .errPos = pos
+        .errPos = state.positions[pos]
     });
 }
 
 
 static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
-    Expr * e, const Pos & pos)
+    Expr * e, const PosIdx pos, const nix::EvalState & state)
 {
     AttrPath::iterator i;
     // All attrpaths have at least one attr
@@ -104,15 +104,15 @@ static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
     // Checking attrPath validity.
     // ===========================
     for (i = attrPath.begin(); i + 1 < attrPath.end(); i++) {
-        if (i->symbol.set()) {
+        if (i->symbol) {
             ExprAttrs::AttrDefs::iterator j = attrs->attrs.find(i->symbol);
             if (j != attrs->attrs.end()) {
                 if (!j->second.inherited) {
                     ExprAttrs * attrs2 = dynamic_cast<ExprAttrs *>(j->second.e);
-                    if (!attrs2) dupAttr(attrPath, pos, j->second.pos);
+                    if (!attrs2) dupAttr(state, attrPath, pos, j->second.pos);
                     attrs = attrs2;
                 } else
-                    dupAttr(attrPath, pos, j->second.pos);
+                    dupAttr(state, attrPath, pos, j->second.pos);
             } else {
                 ExprAttrs * nested = new ExprAttrs;
                 attrs->attrs[i->symbol] = ExprAttrs::AttrDef(nested, pos);
@@ -126,7 +126,7 @@ static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
     }
     // Expr insertion.
     // ==========================
-    if (i->symbol.set()) {
+    if (i->symbol) {
         ExprAttrs::AttrDefs::iterator j = attrs->attrs.find(i->symbol);
         if (j != attrs->attrs.end()) {
             // This attr path is already defined. However, if both
@@ -139,11 +139,11 @@ static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
                 for (auto & ad : ae->attrs) {
                     auto j2 = jAttrs->attrs.find(ad.first);
                     if (j2 != jAttrs->attrs.end()) // Attr already defined in iAttrs, error.
-                        dupAttr(ad.first, j2->second.pos, ad.second.pos);
+                        dupAttr(state, ad.first, j2->second.pos, ad.second.pos);
                     jAttrs->attrs.emplace(ad.first, ad.second);
                 }
             } else {
-                dupAttr(attrPath, pos, j->second.pos);
+                dupAttr(state, attrPath, pos, j->second.pos);
             }
         } else {
             // This attr path is not defined. Let's create it.
@@ -157,14 +157,14 @@ static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
 
 
 static Formals * toFormals(ParseData & data, ParserFormals * formals,
-    Pos pos = noPos, Symbol arg = {})
+    PosIdx pos = noPos, SymbolIdx arg = {})
 {
     std::sort(formals->formals.begin(), formals->formals.end(),
         [] (const auto & a, const auto & b) {
             return std::tie(a.name, a.pos) < std::tie(b.name, b.pos);
         });
 
-    std::optional<std::pair<Symbol, Pos>> duplicate;
+    std::optional<std::pair<SymbolIdx, PosIdx>> duplicate;
     for (size_t i = 0; i + 1 < formals->formals.size(); i++) {
         if (formals->formals[i].name != formals->formals[i + 1].name)
             continue;
@@ -173,18 +173,18 @@ static Formals * toFormals(ParseData & data, ParserFormals * formals,
     }
     if (duplicate)
         throw ParseError({
-            .msg = hintfmt("duplicate formal function argument '%1%'", duplicate->first),
+            .msg = hintfmt("duplicate formal function argument '%1%'", data.symbols[duplicate->first]),
-            .errPos = duplicate->second
+            .errPos = data.state.positions[duplicate->second]
         });
 
     Formals result;
     result.ellipsis = formals->ellipsis;
     result.formals = std::move(formals->formals);
 
-    if (arg.set() && result.has(arg))
+    if (arg && result.has(arg))
         throw ParseError({
-            .msg = hintfmt("duplicate formal function argument '%1%'", arg),
+            .msg = hintfmt("duplicate formal function argument '%1%'", data.symbols[arg]),
-            .errPos = pos
+            .errPos = data.state.positions[pos]
         });
 
     delete formals;
@@ -192,8 +192,8 @@ static Formals * toFormals(ParseData & data, ParserFormals * formals,
 }
 
 
-static Expr * stripIndentation(const Pos & pos, SymbolTable & symbols,
+static Expr * stripIndentation(const PosIdx pos, SymbolTable & symbols,
-    std::vector<std::pair<Pos, std::variant<Expr *, StringToken> > > & es)
+    std::vector<std::pair<PosIdx, std::variant<Expr *, StringToken> > > & es)
 {
     if (es.empty()) return new ExprString("");
 
@@ -233,7 +233,7 @@ static Expr * stripIndentation(const Pos & pos, SymbolTable & symbols,
     }
 
     /* Strip spaces from each line. */
-    std::vector<std::pair<Pos, Expr *> > * es2 = new std::vector<std::pair<Pos, Expr *> >;
+    auto * es2 = new std::vector<std::pair<PosIdx, Expr *> >;
     atStartOfLine = true;
     size_t curDropped = 0;
     size_t n = es.size();
@@ -284,9 +284,9 @@ static Expr * stripIndentation(const Pos & pos, SymbolTable & symbols,
 }
 
 
-static inline Pos makeCurPos(const YYLTYPE & loc, ParseData * data)
+static inline PosIdx makeCurPos(const YYLTYPE & loc, ParseData * data)
 {
-    return Pos(data->origin, data->file, loc.first_line, loc.first_column);
+    return data->state.positions.add(data->origin, loc.first_line, loc.first_column);
 }
 
 #define CUR_POS makeCurPos(*yylocp, data)
@@ -299,7 +299,7 @@ void yyerror(YYLTYPE * loc, yyscan_t scanner, ParseData * data, const char * err
 {
     data->error = {
         .msg = hintfmt(error),
-        .errPos = makeCurPos(*loc, data)
+        .errPos = data->state.positions[makeCurPos(*loc, data)]
     };
 }
 
@@ -320,8 +320,8 @@ void yyerror(YYLTYPE * loc, yyscan_t scanner, ParseData * data, const char * err
   StringToken uri;
   StringToken str;
   std::vector<nix::AttrName> * attrNames;
-  std::vector<std::pair<nix::Pos, nix::Expr *> > * string_parts;
+  std::vector<std::pair<nix::PosIdx, nix::Expr *> > * string_parts;
-  std::vector<std::pair<nix::Pos, std::variant<nix::Expr *, StringToken> > > * ind_string_parts;
+  std::vector<std::pair<nix::PosIdx, std::variant<nix::Expr *, StringToken> > > * ind_string_parts;
 }
 
 %type <e> start expr expr_function expr_if expr_op
@@ -369,15 +369,15 @@ expr_function
   : ID ':' expr_function
     { $$ = new ExprLambda(CUR_POS, data->symbols.create($1), 0, $3); }
   | '{' formals '}' ':' expr_function
-    { $$ = new ExprLambda(CUR_POS, data->symbols.create(""), toFormals(*data, $2), $5); }
+    { $$ = new ExprLambda(CUR_POS, toFormals(*data, $2), $5); }
   | '{' formals '}' '@' ID ':' expr_function
     {
-      Symbol arg = data->symbols.create($5);
+      auto arg = data->symbols.create($5);
       $$ = new ExprLambda(CUR_POS, arg, toFormals(*data, $2, CUR_POS, arg), $7);
     }
   | ID '@' '{' formals '}' ':' expr_function
     {
-      Symbol arg = data->symbols.create($1);
+      auto arg = data->symbols.create($1);
       $$ = new ExprLambda(CUR_POS, arg, toFormals(*data, $4, CUR_POS, arg), $7);
     }
   | ASSERT expr ';' expr_function
@@ -388,7 +388,7 @@ expr_function
     { if (!$2->dynamicAttrs.empty())
         throw ParseError({
             .msg = hintfmt("dynamic attributes not allowed in let"),
-            .errPos = CUR_POS
+            .errPos = data->state.positions[CUR_POS]
         });
       $$ = new ExprLet($2, $4);
     }
@@ -415,7 +415,7 @@ expr_op
   | expr_op UPDATE expr_op { $$ = new ExprOpUpdate(CUR_POS, $1, $3); }
   | expr_op '?' attrpath { $$ = new ExprOpHasAttr($1, *$3); }
   | expr_op '+' expr_op
-    { $$ = new ExprConcatStrings(CUR_POS, false, new std::vector<std::pair<Pos, Expr *> >({{makeCurPos(@1, data), $1}, {makeCurPos(@3, data), $3}})); }
+    { $$ = new ExprConcatStrings(CUR_POS, false, new std::vector<std::pair<PosIdx, Expr *> >({{makeCurPos(@1, data), $1}, {makeCurPos(@3, data), $3}})); }
   | expr_op '-' expr_op { $$ = new ExprCall(CUR_POS, new ExprVar(data->symbols.create("__sub")), {$1, $3}); }
   | expr_op '*' expr_op { $$ = new ExprCall(CUR_POS, new ExprVar(data->symbols.create("__mul")), {$1, $3}); }
   | expr_op '/' expr_op { $$ = new ExprCall(CUR_POS, new ExprVar(data->symbols.create("__div")), {$1, $3}); }
@@ -477,7 +477,7 @@ expr_simple
       if (noURLLiterals)
           throw ParseError({
               .msg = hintfmt("URL literals are disabled"),
-              .errPos = CUR_POS
+              .errPos = data->state.positions[CUR_POS]
           });
       $$ = new ExprString(std::string($1));
   }
@@ -503,9 +503,9 @@ string_parts_interpolated
   : string_parts_interpolated STR
   { $$ = $1; $1->emplace_back(makeCurPos(@2, data), new ExprString(std::string($2))); }
   | string_parts_interpolated DOLLAR_CURLY expr '}' { $$ = $1; $1->emplace_back(makeCurPos(@2, data), $3); }
-  | DOLLAR_CURLY expr '}' { $$ = new std::vector<std::pair<Pos, Expr *> >; $$->emplace_back(makeCurPos(@1, data), $2); }
+  | DOLLAR_CURLY expr '}' { $$ = new std::vector<std::pair<PosIdx, Expr *> >; $$->emplace_back(makeCurPos(@1, data), $2); }
   | STR DOLLAR_CURLY expr '}' {
-      $$ = new std::vector<std::pair<Pos, Expr *> >;
+      $$ = new std::vector<std::pair<PosIdx, Expr *> >;
       $$->emplace_back(makeCurPos(@1, data), new ExprString(std::string($1)));
       $$->emplace_back(makeCurPos(@2, data), $3);
     }
@@ -528,17 +528,17 @@ path_start
 ind_string_parts
   : ind_string_parts IND_STR { $$ = $1; $1->emplace_back(makeCurPos(@2, data), $2); }
   | ind_string_parts DOLLAR_CURLY expr '}' { $$ = $1; $1->emplace_back(makeCurPos(@2, data), $3); }
-  | { $$ = new std::vector<std::pair<Pos, std::variant<Expr *, StringToken> > >; }
+  | { $$ = new std::vector<std::pair<PosIdx, std::variant<Expr *, StringToken> > >; }
   ;
 
 binds
-  : binds attrpath '=' expr ';' { $$ = $1; addAttr($$, *$2, $4, makeCurPos(@2, data)); }
+  : binds attrpath '=' expr ';' { $$ = $1; addAttr($$, *$2, $4, makeCurPos(@2, data), data->state); }
   | binds INHERIT attrs ';'
     { $$ = $1;
       for (auto & i : *$3) {
           if ($$->attrs.find(i.symbol) != $$->attrs.end())
-              dupAttr(i.symbol, makeCurPos(@3, data), $$->attrs[i.symbol].pos);
+              dupAttr(data->state, i.symbol, makeCurPos(@3, data), $$->attrs[i.symbol].pos);
-          Pos pos = makeCurPos(@3, data);
+          auto pos = makeCurPos(@3, data);
           $$->attrs.emplace(i.symbol, ExprAttrs::AttrDef(new ExprVar(CUR_POS, i.symbol), pos, true));
       }
     }
@@ -547,7 +547,7 @@ binds
       /* !!! Should ensure sharing of the expression in $4. */
       for (auto & i : *$6) {
           if ($$->attrs.find(i.symbol) != $$->attrs.end())
-              dupAttr(i.symbol, makeCurPos(@6, data), $$->attrs[i.symbol].pos);
+              dupAttr(data->state, i.symbol, makeCurPos(@6, data), $$->attrs[i.symbol].pos);
           $$->attrs.emplace(i.symbol, ExprAttrs::AttrDef(new ExprSelect(CUR_POS, $4, i.symbol), makeCurPos(@6, data)));
       }
     }
@@ -565,7 +565,7 @@ attrs
       } else
           throw ParseError({
               .msg = hintfmt("dynamic attributes not allowed in inherit"),
-              .errPos = makeCurPos(@2, data)
+              .errPos = data->state.positions[makeCurPos(@2, data)]
           });
     }
   | { $$ = new AttrPath; }
@@ -621,8 +621,8 @@ formals
   ;
 
 formal
-  : ID { $$ = new Formal(CUR_POS, data->symbols.create($1), 0); }
+  : ID { $$ = new Formal{CUR_POS, data->symbols.create($1), 0}; }
-  | ID '?' expr { $$ = new Formal(CUR_POS, data->symbols.create($1), $3); }
+  | ID '?' expr { $$ = new Formal{CUR_POS, data->symbols.create($1), $3}; }
   ;
 
 %%
@@ -646,19 +646,19 @@ Expr * EvalState::parse(char * text, size_t length, FileOrigin origin,
     const PathView path, const PathView basePath, StaticEnv & staticEnv)
 {
     yyscan_t scanner;
-    ParseData data(*this);
+    std::string file;
-    data.origin = origin;
     switch (origin) {
         case foFile:
-            data.file = data.symbols.create(path);
+            file = path;
             break;
         case foStdin:
         case foString:
-            data.file = data.symbols.create(text);
+            file = text;
             break;
         default:
             assert(false);
     }
+    ParseData data(*this, {file, origin});
     data.basePath = basePath;
 
     yylex_init(&scanner);
@@ -668,7 +668,7 @@ Expr * EvalState::parse(char * text, size_t length, FileOrigin origin,
 
     if (res) throw ParseError(data.error.value());
 
-    data.result->bindVars(staticEnv);
+    data.result->bindVars(*this, staticEnv);
 
     return data.result;
 }
@@ -760,7 +760,7 @@ Path EvalState::findFile(const std::string_view path)
 }
 
 
-Path EvalState::findFile(SearchPath & searchPath, const std::string_view path, const Pos & pos)
+Path EvalState::findFile(SearchPath & searchPath, const std::string_view path, const PosIdx pos)
 {
     for (auto & i : searchPath) {
         std::string suffix;
@@ -787,7 +787,7 @@ Path EvalState::findFile(SearchPath & searchPath, const std::string_view path, c
             ? "cannot look up '<%s>' in pure evaluation mode (use '--impure' to override)"
             : "file '%s' was not found in the Nix search path (add it using $NIX_PATH or -I)",
             path),
-        .errPos = pos
+        .errPos = positions[pos]
     });
 }
 

src/libexpr/primops.hh

@@ -16,6 +16,7 @@ struct RegisterPrimOp
         size_t arity = 0;
         const char * doc;
         PrimOpFun fun;
+        std::optional<ExperimentalFeature> experimentalFeature;
     };
 
     typedef std::vector<Info> PrimOps;
@@ -35,10 +36,11 @@ struct RegisterPrimOp
 /* These primops are disabled without enableNativeCode, but plugins
    may wish to use them in limited contexts without globally enabling
    them. */
+
 /* Load a ValueInitializer from a DSO and return whatever it initializes */
-void prim_importNative(EvalState & state, const Pos & pos, Value * * args, Value & v);
+void prim_importNative(EvalState & state, const PosIdx pos, Value * * args, Value & v);
 
 /* Execute a program and parse its output */
-void prim_exec(EvalState & state, const Pos & pos, Value * * args, Value & v);
+void prim_exec(EvalState & state, const PosIdx pos, Value * * args, Value & v);
 
 }

src/libexpr/primops/context.cc

@@ -1,10 +1,11 @@
 #include "primops.hh"
 #include "eval-inline.hh"
+#include "derivations.hh"
 #include "store-api.hh"
 
 namespace nix {
 
-static void prim_unsafeDiscardStringContext(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_unsafeDiscardStringContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     PathSet context;
     auto s = state.coerceToString(pos, *args[0], context);
@@ -14,7 +15,7 @@ static void prim_unsafeDiscardStringContext(EvalState & state, const Pos & pos,
 static RegisterPrimOp primop_unsafeDiscardStringContext("__unsafeDiscardStringContext", 1, prim_unsafeDiscardStringContext);
 
 
-static void prim_hasContext(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_hasContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     PathSet context;
     state.forceString(*args[0], context, pos);
@@ -30,7 +31,7 @@ static RegisterPrimOp primop_hasContext("__hasContext", 1, prim_hasContext);
    source-only deployment).  This primop marks the string context so
    that builtins.derivation adds the path to drv.inputSrcs rather than
    drv.inputDrvs. */
-static void prim_unsafeDiscardOutputDependency(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_unsafeDiscardOutputDependency(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     PathSet context;
     auto s = state.coerceToString(pos, *args[0], context);
@@ -64,7 +65,7 @@ static RegisterPrimOp primop_unsafeDiscardOutputDependency("__unsafeDiscardOutpu
    Note that for a given path any combination of the above attributes
    may be present.
 */
-static void prim_getContext(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_getContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     struct ContextInfo {
         bool path = false;
@@ -82,8 +83,8 @@ static void prim_getContext(EvalState & state, const Pos & pos, Value * * args,
             drv = std::string(p, 1);
             path = &drv;
         } else if (p.at(0) == '!') {
-            std::pair<std::string, std::string> ctx = decodeContext(p);
+            NixStringContextElem ctx = decodeContext(*state.store, p);
-            drv = ctx.first;
+            drv = state.store->printStorePath(ctx.first);
             output = ctx.second;
             path = &drv;
         }
@@ -133,7 +134,7 @@ static RegisterPrimOp primop_getContext("__getContext", 1, prim_getContext);
    See the commentary above unsafeGetContext for details of the
    context representation.
 */
-static void prim_appendContext(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_appendContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     PathSet context;
     auto orig = state.forceString(*args[0], context, pos);
@@ -143,45 +144,46 @@ static void prim_appendContext(EvalState & state, const Pos & pos, Value * * arg
     auto sPath = state.symbols.create("path");
     auto sAllOutputs = state.symbols.create("allOutputs");
     for (auto & i : *args[1]->attrs) {
-        if (!state.store->isStorePath(i.name))
+        const auto & name = state.symbols[i.name];
+        if (!state.store->isStorePath(name))
             throw EvalError({
-                .msg = hintfmt("Context key '%s' is not a store path", i.name),
+                .msg = hintfmt("Context key '%s' is not a store path", name),
-                .errPos = *i.pos
+                .errPos = state.positions[i.pos]
             });
         if (!settings.readOnlyMode)
-            state.store->ensurePath(state.store->parseStorePath(i.name));
+            state.store->ensurePath(state.store->parseStorePath(name));
-        state.forceAttrs(*i.value, *i.pos);
+        state.forceAttrs(*i.value, i.pos);
         auto iter = i.value->attrs->find(sPath);
         if (iter != i.value->attrs->end()) {
-            if (state.forceBool(*iter->value, *iter->pos))
+            if (state.forceBool(*iter->value, iter->pos))
-                context.insert(i.name);
+                context.emplace(name);
         }
 
         iter = i.value->attrs->find(sAllOutputs);
         if (iter != i.value->attrs->end()) {
-            if (state.forceBool(*iter->value, *iter->pos)) {
+            if (state.forceBool(*iter->value, iter->pos)) {
-                if (!isDerivation(i.name)) {
+                if (!isDerivation(name)) {
                     throw EvalError({
-                        .msg = hintfmt("Tried to add all-outputs context of %s, which is not a derivation, to a string", i.name),
+                        .msg = hintfmt("Tried to add all-outputs context of %s, which is not a derivation, to a string", name),
-                        .errPos = *i.pos
+                        .errPos = state.positions[i.pos]
                     });
                 }
-                context.insert("=" + std::string(i.name));
+                context.insert(concatStrings("=", name));
             }
         }
 
         iter = i.value->attrs->find(state.sOutputs);
         if (iter != i.value->attrs->end()) {
-            state.forceList(*iter->value, *iter->pos);
+            state.forceList(*iter->value, iter->pos);
-            if (iter->value->listSize() && !isDerivation(i.name)) {
+            if (iter->value->listSize() && !isDerivation(name)) {
                 throw EvalError({
-                    .msg = hintfmt("Tried to add derivation output context of %s, which is not a derivation, to a string", i.name),
+                    .msg = hintfmt("Tried to add derivation output context of %s, which is not a derivation, to a string", name),
-                    .errPos = *i.pos
+                    .errPos = state.positions[i.pos]
                 });
             }
             for (auto elem : iter->value->listItems()) {
-                auto name = state.forceStringNoCtx(*elem, *iter->pos);
+                auto outputName = state.forceStringNoCtx(*elem, iter->pos);
-                context.insert(concatStrings("!", name, "!", i.name));
+                context.insert(concatStrings("!", outputName, "!", name));
             }
         }
     }

src/libexpr/primops/fetchClosure.cc

@@ -0,0 +1,163 @@
+#include "primops.hh"
+#include "store-api.hh"
+#include "make-content-addressed.hh"
+#include "url.hh"
+
+namespace nix {
+
+static void prim_fetchClosure(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+{
+    state.forceAttrs(*args[0], pos);
+
+    std::optional<std::string> fromStoreUrl;
+    std::optional<StorePath> fromPath;
+    bool toCA = false;
+    std::optional<StorePath> toPath;
+
+    for (auto & attr : *args[0]->attrs) {
+        const auto & attrName = state.symbols[attr.name];
+
+        if (attrName == "fromPath") {
+            PathSet context;
+            fromPath = state.coerceToStorePath(attr.pos, *attr.value, context);
+        }
+
+        else if (attrName == "toPath") {
+            state.forceValue(*attr.value, attr.pos);
+            toCA = true;
+            if (attr.value->type() != nString || attr.value->string.s != std::string("")) {
+                PathSet context;
+                toPath = state.coerceToStorePath(attr.pos, *attr.value, context);
+            }
+        }
+
+        else if (attrName == "fromStore")
+            fromStoreUrl = state.forceStringNoCtx(*attr.value, attr.pos);
+
+        else
+            throw Error({
+                .msg = hintfmt("attribute '%s' isn't supported in call to 'fetchClosure'", attrName),
+                .errPos = state.positions[pos]
+            });
+    }
+
+    if (!fromPath)
+        throw Error({
+            .msg = hintfmt("attribute '%s' is missing in call to 'fetchClosure'", "fromPath"),
+            .errPos = state.positions[pos]
+        });
+
+    if (!fromStoreUrl)
+        throw Error({
+            .msg = hintfmt("attribute '%s' is missing in call to 'fetchClosure'", "fromStore"),
+            .errPos = state.positions[pos]
+        });
+
+    auto parsedURL = parseURL(*fromStoreUrl);
+
+    if (parsedURL.scheme != "http" &&
+        parsedURL.scheme != "https" &&
+        !(getEnv("_NIX_IN_TEST").has_value() && parsedURL.scheme == "file"))
+        throw Error({
+            .msg = hintfmt("'fetchClosure' only supports http:// and https:// stores"),
+            .errPos = state.positions[pos]
+        });
+
+    if (!parsedURL.query.empty())
+        throw Error({
+            .msg = hintfmt("'fetchClosure' does not support URL query parameters (in '%s')", *fromStoreUrl),
+            .errPos = state.positions[pos]
+        });
+
+    auto fromStore = openStore(parsedURL.to_string());
+
+    if (toCA) {
+        if (!toPath || !state.store->isValidPath(*toPath)) {
+            auto remappings = makeContentAddressed(*fromStore, *state.store, { *fromPath });
+            auto i = remappings.find(*fromPath);
+            assert(i != remappings.end());
+            if (toPath && *toPath != i->second)
+                throw Error({
+                    .msg = hintfmt("rewriting '%s' to content-addressed form yielded '%s', while '%s' was expected",
+                        state.store->printStorePath(*fromPath),
+                        state.store->printStorePath(i->second),
+                        state.store->printStorePath(*toPath)),
+                    .errPos = state.positions[pos]
+                });
+            if (!toPath)
+                throw Error({
+                    .msg = hintfmt(
+                        "rewriting '%s' to content-addressed form yielded '%s'; "
+                        "please set this in the 'toPath' attribute passed to 'fetchClosure'",
+                        state.store->printStorePath(*fromPath),
+                        state.store->printStorePath(i->second)),
+                    .errPos = state.positions[pos]
+                });
+        }
+    } else {
+        if (!state.store->isValidPath(*fromPath))
+            copyClosure(*fromStore, *state.store, RealisedPath::Set { *fromPath });
+        toPath = fromPath;
+    }
+
+    /* In pure mode, require a CA path. */
+    if (evalSettings.pureEval) {
+        auto info = state.store->queryPathInfo(*toPath);
+        if (!info->isContentAddressed(*state.store))
+            throw Error({
+                .msg = hintfmt("in pure mode, 'fetchClosure' requires a content-addressed path, which '%s' isn't",
+                    state.store->printStorePath(*toPath)),
+                .errPos = state.positions[pos]
+            });
+    }
+
+    auto toPathS = state.store->printStorePath(*toPath);
+    v.mkString(toPathS, {toPathS});
+}
+
+static RegisterPrimOp primop_fetchClosure({
+    .name = "__fetchClosure",
+    .args = {"args"},
+    .doc = R"(
+      Fetch a Nix store closure from a binary cache, rewriting it into
+      content-addressed form. For example,
+
+      ```nix
+      builtins.fetchClosure {
+        fromStore = "https://cache.nixos.org";
+        fromPath = /nix/store/r2jd6ygnmirm2g803mksqqjm4y39yi6i-git-2.33.1;
+        toPath = /nix/store/ldbhlwhh39wha58rm61bkiiwm6j7211j-git-2.33.1;
+      }
+      ```
+
+      fetches `/nix/store/r2jd...` from the specified binary cache,
+      and rewrites it into the content-addressed store path
+      `/nix/store/ldbh...`.
+
+      If `fromPath` is already content-addressed, or if you are
+      allowing impure evaluation (`--impure`), then `toPath` may be
+      omitted.
+
+      To find out the correct value for `toPath` given a `fromPath`,
+      you can use `nix store make-content-addressed`:
+
+      ```console
+      # nix store make-content-addressed --from https://cache.nixos.org /nix/store/r2jd6ygnmirm2g803mksqqjm4y39yi6i-git-2.33.1
+      rewrote '/nix/store/r2jd6ygnmirm2g803mksqqjm4y39yi6i-git-2.33.1' to '/nix/store/ldbhlwhh39wha58rm61bkiiwm6j7211j-git-2.33.1'
+      ```
+
+      This function is similar to `builtins.storePath` in that it
+      allows you to use a previously built store path in a Nix
+      expression. However, it is more reproducible because it requires
+      specifying a binary cache from which the path can be fetched.
+      Also, requiring a content-addressed final store path avoids the
+      need for users to configure binary cache public keys.
+
+      This function is only available if you enable the experimental
+      feature `fetch-closure`.
+    )",
+    .fun = prim_fetchClosure,
+    .experimentalFeature = Xp::FetchClosure,
+});
+
+}

src/libexpr/primops/fetchMercurial.cc

@@ -7,7 +7,7 @@
 
 namespace nix {
 
-static void prim_fetchMercurial(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_fetchMercurial(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     std::string url;
     std::optional<Hash> rev;
@@ -22,31 +22,31 @@ static void prim_fetchMercurial(EvalState & state, const Pos & pos, Value * * ar
         state.forceAttrs(*args[0], pos);
 
         for (auto & attr : *args[0]->attrs) {
-            std::string_view n(attr.name);
+            std::string_view n(state.symbols[attr.name]);
             if (n == "url")
-                url = state.coerceToString(*attr.pos, *attr.value, context, false, false).toOwned();
+                url = state.coerceToString(attr.pos, *attr.value, context, false, false).toOwned();
             else if (n == "rev") {
                 // Ugly: unlike fetchGit, here the "rev" attribute can
                 // be both a revision or a branch/tag name.
-                auto value = state.forceStringNoCtx(*attr.value, *attr.pos);
+                auto value = state.forceStringNoCtx(*attr.value, attr.pos);
                 if (std::regex_match(value.begin(), value.end(), revRegex))
                     rev = Hash::parseAny(value, htSHA1);
                 else
                     ref = value;
             }
             else if (n == "name")
-                name = state.forceStringNoCtx(*attr.value, *attr.pos);
+                name = state.forceStringNoCtx(*attr.value, attr.pos);
             else
                 throw EvalError({
-                    .msg = hintfmt("unsupported argument '%s' to 'fetchMercurial'", attr.name),
+                    .msg = hintfmt("unsupported argument '%s' to 'fetchMercurial'", state.symbols[attr.name]),
-                    .errPos = *attr.pos
+                    .errPos = state.positions[attr.pos]
                 });
         }
 
         if (url.empty())
             throw EvalError({
                 .msg = hintfmt("'url' argument required"),
-                .errPos = pos
+                .errPos = state.positions[pos]
             });
 
     } else

src/libexpr/primops/fetchTree.cc

@@ -90,7 +90,7 @@ struct FetchTreeParams {
 
 static void fetchTree(
     EvalState & state,
-    const Pos & pos,
+    const PosIdx pos,
     Value * * args,
     Value & v,
     std::optional<std::string> type,
@@ -110,43 +110,43 @@ static void fetchTree(
             if (type)
                 throw Error({
                     .msg = hintfmt("unexpected attribute 'type'"),
-                    .errPos = pos
+                    .errPos = state.positions[pos]
                 });
-            type = state.forceStringNoCtx(*aType->value, *aType->pos);
+            type = state.forceStringNoCtx(*aType->value, aType->pos);
         } else if (!type)
             throw Error({
                 .msg = hintfmt("attribute 'type' is missing in call to 'fetchTree'"),
-                .errPos = pos
+                .errPos = state.positions[pos]
             });
 
         attrs.emplace("type", type.value());
 
         for (auto & attr : *args[0]->attrs) {
             if (attr.name == state.sType) continue;
-            state.forceValue(*attr.value, *attr.pos);
+            state.forceValue(*attr.value, attr.pos);
             if (attr.value->type() == nPath || attr.value->type() == nString) {
-                auto s = state.coerceToString(*attr.pos, *attr.value, context, false, false).toOwned();
+                auto s = state.coerceToString(attr.pos, *attr.value, context, false, false).toOwned();
-                attrs.emplace(attr.name,
+                attrs.emplace(state.symbols[attr.name],
-                    attr.name == "url"
+                    state.symbols[attr.name] == "url"
                     ? type == "git"
                       ? fixURIForGit(s, state)
                       : fixURI(s, state)
                     : s);
             }
             else if (attr.value->type() == nBool)
-                attrs.emplace(attr.name, Explicit<bool>{attr.value->boolean});
+                attrs.emplace(state.symbols[attr.name], Explicit<bool>{attr.value->boolean});
             else if (attr.value->type() == nInt)
-                attrs.emplace(attr.name, uint64_t(attr.value->integer));
+                attrs.emplace(state.symbols[attr.name], uint64_t(attr.value->integer));
             else
                 throw TypeError("fetchTree argument '%s' is %s while a string, Boolean or integer is expected",
-                    attr.name, showType(*attr.value));
+                    state.symbols[attr.name], showType(*attr.value));
         }
 
         if (!params.allowNameArgument)
             if (auto nameIter = attrs.find("name"); nameIter != attrs.end())
                 throw Error({
-                    .msg = hintfmt("attribute 'name' isn’t supported in call to 'fetchTree'"),
+                    .msg = hintfmt("attribute 'name' isn't supported in call to 'fetchTree'"),
-                    .errPos = pos
+                    .errPos = state.positions[pos]
                 });
 
         input = fetchers::Input::fromAttrs(std::move(attrs));
@@ -167,7 +167,7 @@ static void fetchTree(
         input = lookupInRegistries(state.store, input).first;
 
     if (evalSettings.pureEval && !input.isLocked())
-        throw Error("in pure evaluation mode, 'fetchTree' requires a locked input, at %s", pos);
+        throw Error("in pure evaluation mode, 'fetchTree' requires a locked input, at %s", state.positions[pos]);
 
     auto [tree, input2] = input.fetch(state.store);
 
@@ -176,7 +176,7 @@ static void fetchTree(
     emitTreeAttrs(state, tree, input2, v, params.emptyRevFallback, false);
 }
 
-static void prim_fetchTree(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_fetchTree(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     settings.requireExperimentalFeature(Xp::Flakes);
     fetchTree(state, pos, args, v, std::nullopt, FetchTreeParams { .allowNameArgument = false });
@@ -185,7 +185,7 @@ static void prim_fetchTree(EvalState & state, const Pos & pos, Value * * args, V
 // FIXME: document
 static RegisterPrimOp primop_fetchTree("fetchTree", 1, prim_fetchTree);
 
-static void fetch(EvalState & state, const Pos & pos, Value * * args, Value & v,
+static void fetch(EvalState & state, const PosIdx pos, Value * * args, Value & v,
     const std::string & who, bool unpack, std::string name)
 {
     std::optional<std::string> url;
@@ -198,24 +198,24 @@ static void fetch(EvalState & state, const Pos & pos, Value * * args, Value & v,
         state.forceAttrs(*args[0], pos);
 
         for (auto & attr : *args[0]->attrs) {
-            std::string n(attr.name);
+            std::string_view n(state.symbols[attr.name]);
             if (n == "url")
-                url = state.forceStringNoCtx(*attr.value, *attr.pos);
+                url = state.forceStringNoCtx(*attr.value, attr.pos);
             else if (n == "sha256")
-                expectedHash = newHashAllowEmpty(state.forceStringNoCtx(*attr.value, *attr.pos), htSHA256);
+                expectedHash = newHashAllowEmpty(state.forceStringNoCtx(*attr.value, attr.pos), htSHA256);
             else if (n == "name")
-                name = state.forceStringNoCtx(*attr.value, *attr.pos);
+                name = state.forceStringNoCtx(*attr.value, attr.pos);
             else
                 throw EvalError({
-                    .msg = hintfmt("unsupported argument '%s' to '%s'", attr.name, who),
+                    .msg = hintfmt("unsupported argument '%s' to '%s'", n, who),
-                    .errPos = *attr.pos
+                    .errPos = state.positions[attr.pos]
                 });
             }
 
         if (!url)
             throw EvalError({
                 .msg = hintfmt("'url' argument required"),
-                .errPos = pos
+                .errPos = state.positions[pos]
             });
     } else
         url = state.forceStringNoCtx(*args[0], pos);
@@ -262,7 +262,7 @@ static void fetch(EvalState & state, const Pos & pos, Value * * args, Value & v,
     state.allowAndSetStorePathString(storePath, v);
 }
 
-static void prim_fetchurl(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_fetchurl(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     fetch(state, pos, args, v, "fetchurl", false, "");
 }
@@ -278,7 +278,7 @@ static RegisterPrimOp primop_fetchurl({
     .fun = prim_fetchurl,
 });
 
-static void prim_fetchTarball(EvalState & state, const Pos & pos, Value * * args, Value & v)
+static void prim_fetchTarball(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     fetch(state, pos, args, v, "fetchTarball", true, "source");
 }
@@ -329,7 +329,7 @@ static RegisterPrimOp primop_fetchTarball({
     .fun = prim_fetchTarball,
 });
 
-static void prim_fetchGit(EvalState &state, const Pos &pos, Value **args, Value &v)
+static void prim_fetchGit(EvalState & state, const PosIdx pos, Value * * args, Value & v)
 {
     fetchTree(state, pos, args, v, "git", FetchTreeParams { .emptyRevFallback = true, .allowNameArgument = true });
 }

src/libexpr/primops/fromTOML.cc

@@ -5,7 +5,7 @@
 
 namespace nix {
 
-static void prim_fromTOML(EvalState & state, const Pos & pos, Value * * args, Value & val)
+static void prim_fromTOML(EvalState & state, const PosIdx pos, Value * * args, Value & val)
 {
     auto toml = state.forceStringNoCtx(*args[0], pos);
 
@@ -73,7 +73,7 @@ static void prim_fromTOML(EvalState & state, const Pos & pos, Value * * args, Va
     } catch (std::exception & e) { // TODO: toml::syntax_error
         throw EvalError({
             .msg = hintfmt("while parsing a TOML string: %s", e.what()),
-            .errPos = pos
+            .errPos = state.positions[pos]
         });
     }
 }

src/libexpr/symbol-table.hh

@@ -5,6 +5,7 @@
 #include <unordered_map>
 
 #include "types.hh"
+#include "chunked-vector.hh"
 
 namespace nix {
 
@@ -16,90 +17,90 @@ namespace nix {
 
 class Symbol
 {
-private:
-    const std::string * s; // pointer into SymbolTable
-    Symbol(const std::string * s) : s(s) { };
     friend class SymbolTable;
+private:
+    std::string s;
 
 public:
-    Symbol() : s(0) { };
+    Symbol(std::string_view s) : s(s) { }
-
-    bool operator == (const Symbol & s2) const
-    {
-        return s == s2.s;
-    }
 
     // FIXME: remove
     bool operator == (std::string_view s2) const
     {
-        return s->compare(s2) == 0;
+        return s == s2;
-    }
-
-    bool operator != (const Symbol & s2) const
-    {
-        return s != s2.s;
-    }
-
-    bool operator < (const Symbol & s2) const
-    {
-        return s < s2.s;
     }
 
     operator const std::string & () const
     {
-        return *s;
+        return s;
     }
 
     operator const std::string_view () const
-    {
-        return *s;
-    }
-
-    bool set() const
     {
         return s;
     }
 
-    bool empty() const
-    {
-        return s->empty();
-    }
-
     friend std::ostream & operator << (std::ostream & str, const Symbol & sym);
 };
 
+class SymbolIdx
+{
+    friend class SymbolTable;
+
+private:
+    uint32_t id;
+
+    explicit SymbolIdx(uint32_t id): id(id) {}
+
+public:
+    SymbolIdx() : id(0) {}
+
+    explicit operator bool() const { return id > 0; }
+
+    bool operator<(const SymbolIdx other) const { return id < other.id; }
+    bool operator==(const SymbolIdx other) const { return id == other.id; }
+    bool operator!=(const SymbolIdx other) const { return id != other.id; }
+};
+
 class SymbolTable
 {
 private:
-    std::unordered_map<std::string_view, Symbol> symbols;
+    std::unordered_map<std::string_view, std::pair<const Symbol *, uint32_t>> symbols;
-    std::list<std::string> store;
+    ChunkedVector<Symbol, 8192> store{16};
 
 public:
-    Symbol create(std::string_view s)
+    SymbolIdx create(std::string_view s)
     {
         // Most symbols are looked up more than once, so we trade off insertion performance
         // for lookup performance.
         // TODO: could probably be done more efficiently with transparent Hash and Equals
         // on the original implementation using unordered_set
         auto it = symbols.find(s);
-        if (it != symbols.end()) return it->second;
+        if (it != symbols.end()) return SymbolIdx(it->second.second + 1);
 
-        auto & rawSym = store.emplace_back(s);
+        const auto & [rawSym, idx] = store.add(s);
-        return symbols.emplace(rawSym, Symbol(&rawSym)).first->second;
+        symbols.emplace(rawSym, std::make_pair(&rawSym, idx));
+        return SymbolIdx(idx + 1);
+    }
+
+    const Symbol & operator[](SymbolIdx s) const
+    {
+        if (s.id == 0 || s.id > store.size())
+            abort();
+        return store[s.id - 1];
     }
 
     size_t size() const
     {
-        return symbols.size();
+        return store.size();
     }
 
     size_t totalSize() const;
 
     template<typename T>
-    void dump(T callback)
+    void dump(T callback) const
     {
-        for (auto & s : store)
+        store.forEach(callback);
-            callback(s);
     }
 };
 

src/libexpr/value-to-json.cc

@@ -10,7 +10,7 @@
 namespace nix {
 
 void printValueAsJSON(EvalState & state, bool strict,
-    Value & v, const Pos & pos, JSONPlaceholder & out, PathSet & context)
+    Value & v, const PosIdx pos, JSONPlaceholder & out, PathSet & context)
 {
     checkInterrupt();
 
@@ -50,14 +50,14 @@ void printValueAsJSON(EvalState & state, bool strict,
                 auto obj(out.object());
                 StringSet names;
                 for (auto & j : *v.attrs)
-                    names.insert(j.name);
+                    names.emplace(state.symbols[j.name]);
                 for (auto & j : names) {
                     Attr & a(*v.attrs->find(state.symbols.create(j)));
                     auto placeholder(obj.placeholder(j));
-                    printValueAsJSON(state, strict, *a.value, *a.pos, placeholder, context);
+                    printValueAsJSON(state, strict, *a.value, a.pos, placeholder, context);
                 }
             } else
-                printValueAsJSON(state, strict, *i->value, *i->pos, out, context);
+                printValueAsJSON(state, strict, *i->value, i->pos, out, context);
             break;
         }
 
@@ -82,14 +82,15 @@ void printValueAsJSON(EvalState & state, bool strict,
         case nFunction:
             auto e = TypeError({
                 .msg = hintfmt("cannot convert %1% to JSON", showType(v)),
-                .errPos = v.determinePos(pos)
+                .errPos = state.positions[v.determinePos(pos)]
             });
-            throw e.addTrace(pos, hintfmt("message for the trace"));
+            e.addTrace(state.positions[pos], hintfmt("message for the trace"));
+            throw e;
     }
 }
 
 void printValueAsJSON(EvalState & state, bool strict,
-    Value & v, const Pos & pos, std::ostream & str, PathSet & context)
+    Value & v, const PosIdx pos, std::ostream & str, PathSet & context)
 {
     JSONPlaceholder out(str);
     printValueAsJSON(state, strict, v, pos, out, context);

src/libexpr/value-to-json.hh

@@ -11,9 +11,9 @@ namespace nix {
 class JSONPlaceholder;
 
 void printValueAsJSON(EvalState & state, bool strict,
-    Value & v, const Pos & pos, JSONPlaceholder & out, PathSet & context);
+    Value & v, const PosIdx pos, JSONPlaceholder & out, PathSet & context);
 
 void printValueAsJSON(EvalState & state, bool strict,
-    Value & v, const Pos & pos, std::ostream & str, PathSet & context);
+    Value & v, const PosIdx pos, std::ostream & str, PathSet & context);
 
 }

src/libexpr/value-to-xml.cc

@@ -19,10 +19,10 @@ static XMLAttrs singletonAttrs(const std::string & name, const std::string & val
 
 static void printValueAsXML(EvalState & state, bool strict, bool location,
     Value & v, XMLWriter & doc, PathSet & context, PathSet & drvsSeen,
-    const Pos & pos);
+    const PosIdx pos);
 
 
-static void posToXML(XMLAttrs & xmlAttrs, const Pos & pos)
+static void posToXML(EvalState & state, XMLAttrs & xmlAttrs, const Pos & pos)
 {
     xmlAttrs["path"] = pos.file;
     xmlAttrs["line"] = (format("%1%") % pos.line).str();
@@ -36,25 +36,25 @@ static void showAttrs(EvalState & state, bool strict, bool location,
     StringSet names;
 
     for (auto & i : attrs)
-        names.insert(i.name);
+        names.emplace(state.symbols[i.name]);
 
     for (auto & i : names) {
         Attr & a(*attrs.find(state.symbols.create(i)));
 
         XMLAttrs xmlAttrs;
         xmlAttrs["name"] = i;
-        if (location && a.pos != ptr(&noPos)) posToXML(xmlAttrs, *a.pos);
+        if (location && a.pos) posToXML(state, xmlAttrs, state.positions[a.pos]);
 
         XMLOpenElement _(doc, "attr", xmlAttrs);
         printValueAsXML(state, strict, location,
-            *a.value, doc, context, drvsSeen, *a.pos);
+            *a.value, doc, context, drvsSeen, a.pos);
     }
 }
 
 
 static void printValueAsXML(EvalState & state, bool strict, bool location,
     Value & v, XMLWriter & doc, PathSet & context, PathSet & drvsSeen,
-    const Pos & pos)
+    const PosIdx pos)
 {
     checkInterrupt();
 
@@ -93,14 +93,14 @@ static void printValueAsXML(EvalState & state, bool strict, bool location,
                 Path drvPath;
                 a = v.attrs->find(state.sDrvPath);
                 if (a != v.attrs->end()) {
-                    if (strict) state.forceValue(*a->value, *a->pos);
+                    if (strict) state.forceValue(*a->value, a->pos);
                     if (a->value->type() == nString)
                         xmlAttrs["drvPath"] = drvPath = a->value->string.s;
                 }
 
                 a = v.attrs->find(state.sOutPath);
                 if (a != v.attrs->end()) {
-                    if (strict) state.forceValue(*a->value, *a->pos);
+                    if (strict) state.forceValue(*a->value, a->pos);
                     if (a->value->type() == nString)
                         xmlAttrs["outPath"] = a->value->string.s;
                 }
@@ -134,18 +134,18 @@ static void printValueAsXML(EvalState & state, bool strict, bool location,
                 break;
             }
             XMLAttrs xmlAttrs;
-            if (location) posToXML(xmlAttrs, v.lambda.fun->pos);
+            if (location) posToXML(state, xmlAttrs, state.positions[v.lambda.fun->pos]);
             XMLOpenElement _(doc, "function", xmlAttrs);
 
             if (v.lambda.fun->hasFormals()) {
                 XMLAttrs attrs;
-                if (!v.lambda.fun->arg.empty()) attrs["name"] = v.lambda.fun->arg;
+                if (v.lambda.fun->arg) attrs["name"] = state.symbols[v.lambda.fun->arg];
                 if (v.lambda.fun->formals->ellipsis) attrs["ellipsis"] = "1";
                 XMLOpenElement _(doc, "attrspat", attrs);
-                for (auto & i : v.lambda.fun->formals->lexicographicOrder())
+                for (auto & i : v.lambda.fun->formals->lexicographicOrder(state.symbols))
-                    doc.writeEmptyElement("attr", singletonAttrs("name", i.name));
+                    doc.writeEmptyElement("attr", singletonAttrs("name", state.symbols[i.name]));
             } else
-                doc.writeEmptyElement("varpat", singletonAttrs("name", v.lambda.fun->arg));
+                doc.writeEmptyElement("varpat", singletonAttrs("name", state.symbols[v.lambda.fun->arg]));
 
             break;
         }
@@ -166,14 +166,14 @@ static void printValueAsXML(EvalState & state, bool strict, bool location,
 
 void ExternalValueBase::printValueAsXML(EvalState & state, bool strict,
     bool location, XMLWriter & doc, PathSet & context, PathSet & drvsSeen,
-    const Pos & pos) const
+    const PosIdx pos) const
 {
     doc.writeEmptyElement("unevaluated");
 }
 
 
 void printValueAsXML(EvalState & state, bool strict, bool location,
-    Value & v, std::ostream & out, PathSet & context, const Pos & pos)
+    Value & v, std::ostream & out, PathSet & context, const PosIdx pos)
 {
     XMLWriter doc(true, out);
     XMLOpenElement root(doc, "expr");

src/libexpr/value-to-xml.hh

@@ -9,6 +9,6 @@
 namespace nix {
 
 void printValueAsXML(EvalState & state, bool strict, bool location,
-    Value & v, std::ostream & out, PathSet & context, const Pos & pos);
+    Value & v, std::ostream & out, PathSet & context, const PosIdx pos);
 
 }

src/libexpr/value.hh

@@ -55,8 +55,11 @@ struct Env;
 struct Expr;
 struct ExprLambda;
 struct PrimOp;
-class Symbol;
+class SymbolIdx;
+class PosIdx;
 struct Pos;
+class StorePath;
+class Store;
 class EvalState;
 class XMLWriter;
 class JSONPlaceholder;
@@ -64,6 +67,8 @@ class JSONPlaceholder;
 
 typedef int64_t NixInt;
 typedef double NixFloat;
+typedef std::pair<StorePath, std::string> NixStringContextElem;
+typedef std::vector<NixStringContextElem> NixStringContext;
 
 /* External values must descend from ExternalValueBase, so that
  * type-agnostic nix functions (e.g. showType) can be implemented
@@ -99,7 +104,7 @@ class ExternalValueBase
     /* Print the value as XML. Defaults to unevaluated */
     virtual void printValueAsXML(EvalState & state, bool strict, bool location,
         XMLWriter & doc, PathSet & context, PathSet & drvsSeen,
-        const Pos & pos) const;
+        const PosIdx pos) const;
 
     virtual ~ExternalValueBase()
     {
@@ -115,10 +120,13 @@ private:
     InternalType internalType;
 
     friend std::string showType(const Value & v);
-    friend void printValue(std::ostream & str, std::set<const void *> & seen, const Value & v);
+
+    void print(const SymbolTable & symbols, std::ostream & str, std::set<const void *> * seen) const;
 
 public:
 
+    void print(const SymbolTable & symbols, std::ostream & str, bool showRepeated = false) const;
+
     // Functions needed to distinguish the type
     // These should be removed eventually, by putting the functionality that's
     // needed by callers into methods of this type
@@ -245,7 +253,7 @@ public:
 
     inline void mkString(const Symbol & s)
     {
-        mkString(((const std::string &) s).c_str());
+        mkString(std::string_view(s).data());
     }
 
     inline void mkPath(const char * s)
@@ -361,14 +369,14 @@ public:
         return internalType == tList1 ? 1 : internalType == tList2 ? 2 : bigList.size;
     }
 
-    Pos determinePos(const Pos & pos) const;
+    PosIdx determinePos(const PosIdx pos) const;
 
     /* Check whether forcing this value requires a trivial amount of
        computation. In particular, function applications are
        non-trivial. */
     bool isTrivial() const;
 
-    std::vector<std::pair<Path, std::string>> getContext();
+    NixStringContext getContext(const Store &);
 
     auto listItems()
     {
@@ -402,12 +410,12 @@ public:
 
 #if HAVE_BOEHMGC
 typedef std::vector<Value *, traceable_allocator<Value *> > ValueVector;
-typedef std::map<Symbol, Value *, std::less<Symbol>, traceable_allocator<std::pair<const Symbol, Value *> > > ValueMap;
+typedef std::map<SymbolIdx, Value *, std::less<SymbolIdx>, traceable_allocator<std::pair<const SymbolIdx, Value *> > > ValueMap;
-typedef std::map<Symbol, ValueVector, std::less<Symbol>, traceable_allocator<std::pair<const Symbol, ValueVector> > > ValueVectorMap;
+typedef std::map<SymbolIdx, ValueVector, std::less<SymbolIdx>, traceable_allocator<std::pair<const SymbolIdx, ValueVector> > > ValueVectorMap;
 #else
 typedef std::vector<Value *> ValueVector;
-typedef std::map<Symbol, Value *> ValueMap;
+typedef std::map<SymbolIdx, Value *> ValueMap;
-typedef std::map<Symbol, ValueVector> ValueVectorMap;
+typedef std::map<SymbolIdx, ValueVector> ValueVectorMap;
 #endif
 
 

src/libfetchers/fetchers.cc

@@ -238,9 +238,18 @@ std::optional<std::string> Input::getRef() const
 
 std::optional<Hash> Input::getRev() const
 {
-    if (auto s = maybeGetStrAttr(attrs, "rev"))
+    std::optional<Hash> hash = {};
-        return Hash::parseAny(*s, htSHA1);
+
-    return {};
+    if (auto s = maybeGetStrAttr(attrs, "rev")) {
+        try {
+            hash = Hash::parseAnyPrefixed(*s);
+        } catch (BadHash &e) {
+            // Default to sha1 for backwards compatibility with existing flakes
+            hash = Hash::parseAny(*s, htSHA1);
+        }
+    }
+
+    return hash;
 }
 
 std::optional<uint64_t> Input::getRevCount() const

src/libfetchers/git.cc

@@ -28,9 +28,7 @@ static std::string readHead(const Path & path)
 
 static bool isNotDotGitDirectory(const Path & path)
 {
-    static const std::regex gitDirRegex("^(?:.*/)?\\.git$");
+    return baseNameOf(path) != ".git";
-
-    return not std::regex_match(path, gitDirRegex);
 }
 
 struct GitInputScheme : InputScheme
@@ -189,8 +187,16 @@ struct GitInputScheme : InputScheme
         if (submodules) cacheType += "-submodules";
         if (allRefs) cacheType += "-all-refs";
 
+        auto checkHashType = [&](const std::optional<Hash> & hash)
+        {
+            if (hash.has_value() && !(hash->type == htSHA1 || hash->type == htSHA256))
+                throw Error("Hash '%s' is not supported by Git. Supported types are sha1 and sha256.", hash->to_string(Base16, true));
+        };
+
         auto getLockedAttrs = [&]()
         {
+            checkHashType(input.getRev());
+
             return Attrs({
                 {"type", cacheType},
                 {"name", name},
@@ -222,22 +228,46 @@ struct GitInputScheme : InputScheme
         if (!input.getRef() && !input.getRev() && isLocal) {
             bool clean = false;
 
-            /* Check whether this repo has any commits. There are
+            auto env = getEnv();
-               probably better ways to do this. */
+            // Set LC_ALL to C: because we rely on the error messages from git rev-parse to determine what went wrong
-            auto gitDir = actualUrl + "/.git";
+            // that way unknown errors can lead to a failure instead of continuing through the wrong code path
-            auto commonGitDir = chomp(runProgram(
+            env["LC_ALL"] = "C";
-                "git",
-                true,
-                { "-C", actualUrl, "rev-parse", "--git-common-dir" }
-            ));
-            if (commonGitDir != ".git")
-                    gitDir = commonGitDir;
 
-            bool haveCommits = !readDirectory(gitDir + "/refs/heads").empty();
+            /* Check whether HEAD points to something that looks like a commit,
+               since that is the refrence we want to use later on. */
+            auto result = runProgram(RunOptions {
+                .program = "git",
+                .args = { "-C", actualUrl, "--git-dir=.git", "rev-parse", "--verify", "--no-revs", "HEAD^{commit}" },
+                .environment = env,
+                .mergeStderrToStdout = true
+            });
+            auto exitCode = WEXITSTATUS(result.first);
+            auto errorMessage = result.second;
 
+            if (errorMessage.find("fatal: not a git repository") != std::string::npos) {
+                throw Error("'%s' is not a Git repository", actualUrl);
+            } else if (errorMessage.find("fatal: Needed a single revision") != std::string::npos) {
+                // indicates that the repo does not have any commits
+                // we want to proceed and will consider it dirty later
+            } else if (exitCode != 0) {
+                // any other errors should lead to a failure
+                throw Error("getting the HEAD of the Git tree '%s' failed with exit code %d:\n%s", actualUrl, exitCode, errorMessage);
+            }
+
+            bool hasHead = exitCode == 0;
             try {
-                if (haveCommits) {
+                if (hasHead) {
-                    runProgram("git", true, { "-C", actualUrl, "diff-index", "--quiet", "HEAD", "--" });
+                    // Using git diff is preferrable over lower-level operations here,
+                    // because its conceptually simpler and we only need the exit code anyways.
+                    auto gitDiffOpts = Strings({ "-C", actualUrl, "diff", "HEAD", "--quiet"});
+                    if (!submodules) {
+                        // Changes in submodules should only make the tree dirty
+                        // when those submodules will be copied as well.
+                        gitDiffOpts.emplace_back("--ignore-submodules");
+                    }
+                    gitDiffOpts.emplace_back("--");
+                    runProgram("git", true, gitDiffOpts);
+
                     clean = true;
                 }
             } catch (ExecError & e) {
@@ -261,9 +291,11 @@ struct GitInputScheme : InputScheme
                 auto files = tokenizeString<std::set<std::string>>(
                     runProgram("git", true, gitOpts), "\0"s);
 
+                Path actualPath(absPath(actualUrl));
+
                 PathFilter filter = [&](const Path & p) -> bool {
-                    assert(hasPrefix(p, actualUrl));
+                    assert(hasPrefix(p, actualPath));
-                    std::string file(p, actualUrl.size() + 1);
+                    std::string file(p, actualPath.size() + 1);
 
                     auto st = lstat(p);
 
@@ -276,13 +308,13 @@ struct GitInputScheme : InputScheme
                     return files.count(file);
                 };
 
-                auto storePath = store->addToStore(input.getName(), actualUrl, FileIngestionMethod::Recursive, htSHA256, filter);
+                auto storePath = store->addToStore(input.getName(), actualPath, FileIngestionMethod::Recursive, htSHA256, filter);
 
                 // FIXME: maybe we should use the timestamp of the last
                 // modified dirty file?
                 input.attrs.insert_or_assign(
                     "lastModified",
-                    haveCommits ? std::stoull(runProgram("git", true, { "-C", actualUrl, "log", "-1", "--format=%ct", "--no-show-signature", "HEAD" })) : 0);
+                    hasHead ? std::stoull(runProgram("git", true, { "-C", actualPath, "log", "-1", "--format=%ct", "--no-show-signature", "HEAD" })) : 0);
 
                 return {std::move(storePath), input};
             }

src/libfetchers/github.cc

@@ -390,7 +390,7 @@ struct SourceHutInputScheme : GitArchiveInputScheme
 
             ref_uri = line.substr(ref_index+5, line.length()-1);
         } else
-            ref_uri = fmt("refs/heads/%s", ref);
+            ref_uri = fmt("refs/(heads|tags)/%s", ref);
 
         auto file = store->toRealPath(
             downloadFile(store, fmt("%s/info/refs", base_url), "source", false, headers).storePath);
@@ -399,9 +399,11 @@ struct SourceHutInputScheme : GitArchiveInputScheme
         std::string line;
         std::string id;
         while(getline(is, line)) {
-            auto index = line.find(ref_uri);
+            // Append $ to avoid partial name matches
-            if (index != std::string::npos) {
+            std::regex pattern(fmt("%s$", ref_uri));
-                id = line.substr(0, index-1);
+
+            if (std::regex_search(line, pattern)) {
+                id = line.substr(0, line.find('\t'));
                 break;
             }
         }

src/libfetchers/mercurial.cc

@@ -36,7 +36,7 @@ static std::string runHg(const Strings & args, const std::optional<std::string>
     auto res = runProgram(std::move(opts));
 
     if (!statusOk(res.first))
-        throw ExecError(res.first, fmt("hg %1%", statusToString(res.first)));
+        throw ExecError(res.first, "hg %1%", statusToString(res.first));
 
     return res.second;
 }
@@ -178,9 +178,11 @@ struct MercurialInputScheme : InputScheme
                 auto files = tokenizeString<std::set<std::string>>(
                     runHg({ "status", "-R", actualUrl, "--clean", "--modified", "--added", "--no-status", "--print0" }), "\0"s);
 
+                Path actualPath(absPath(actualUrl));
+
                 PathFilter filter = [&](const Path & p) -> bool {
-                    assert(hasPrefix(p, actualUrl));
+                    assert(hasPrefix(p, actualPath));
-                    std::string file(p, actualUrl.size() + 1);
+                    std::string file(p, actualPath.size() + 1);
 
                     auto st = lstat(p);
 
@@ -193,7 +195,7 @@ struct MercurialInputScheme : InputScheme
                     return files.count(file);
                 };
 
-                auto storePath = store->addToStore(input.getName(), actualUrl, FileIngestionMethod::Recursive, htSHA256, filter);
+                auto storePath = store->addToStore(input.getName(), actualPath, FileIngestionMethod::Recursive, htSHA256, filter);
 
                 return {std::move(storePath), input};
             }
@@ -201,8 +203,17 @@ struct MercurialInputScheme : InputScheme
 
         if (!input.getRef()) input.attrs.insert_or_assign("ref", "default");
 
+        auto checkHashType = [&](const std::optional<Hash> & hash)
+        {
+            if (hash.has_value() && hash->type != htSHA1)
+                throw Error("Hash '%s' is not supported by Mercurial. Only sha1 is supported.", hash->to_string(Base16, true));
+        };
+
+
         auto getLockedAttrs = [&]()
         {
+            checkHashType(input.getRev());
+
             return Attrs({
                 {"type", "hg"},
                 {"name", name},
@@ -262,7 +273,7 @@ struct MercurialInputScheme : InputScheme
                         runHg({ "recover", "-R", cacheDir });
                         runHg({ "pull", "-R", cacheDir, "--", actualUrl });
                     } else {
-                        throw ExecError(e.status, fmt("'hg pull' %s", statusToString(e.status)));
+                        throw ExecError(e.status, "'hg pull' %s", statusToString(e.status));
                     }
                 }
             } else {

src/libfetchers/path.cc

@@ -1,5 +1,6 @@
 #include "fetchers.hh"
 #include "store-api.hh"
+#include "archive.hh"
 
 namespace nix::fetchers {
 
@@ -80,8 +81,9 @@ struct PathInputScheme : InputScheme
         // nothing to do
     }
 
-    std::pair<StorePath, Input> fetch(ref<Store> store, const Input & input) override
+    std::pair<StorePath, Input> fetch(ref<Store> store, const Input & _input) override
     {
+        Input input(_input);
         std::string absPath;
         auto path = getStrAttr(input.attrs, "path");
 
@@ -111,9 +113,15 @@ struct PathInputScheme : InputScheme
         if (storePath)
             store->addTempRoot(*storePath);
 
-        if (!storePath || storePath->name() != "source" || !store->isValidPath(*storePath))
+        time_t mtime = 0;
+        if (!storePath || storePath->name() != "source" || !store->isValidPath(*storePath)) {
             // FIXME: try to substitute storePath.
-            storePath = store->addToStore("source", absPath);
+            auto src = sinkToSource([&](Sink & sink) {
+                mtime = dumpPathAndGetMtime(absPath, sink, defaultPathFilter);
+            });
+            storePath = store->addToStoreFromDump(*src, "source");
+        }
+        input.attrs.insert_or_assign("lastModified", uint64_t(mtime));
 
         return {std::move(*storePath), input};
     }

src/libmain/shared.cc

@@ -60,37 +60,37 @@ void printMissing(ref<Store> store, const StorePathSet & willBuild,
 {
     if (!willBuild.empty()) {
         if (willBuild.size() == 1)
-            printMsg(lvl, fmt("this derivation will be built:"));
+            printMsg(lvl, "this derivation will be built:");
         else
-            printMsg(lvl, fmt("these %d derivations will be built:", willBuild.size()));
+            printMsg(lvl, "these %d derivations will be built:", willBuild.size());
         auto sorted = store->topoSortPaths(willBuild);
         reverse(sorted.begin(), sorted.end());
         for (auto & i : sorted)
-            printMsg(lvl, fmt("  %s", store->printStorePath(i)));
+            printMsg(lvl, "  %s", store->printStorePath(i));
     }
 
     if (!willSubstitute.empty()) {
         const float downloadSizeMiB = downloadSize / (1024.f * 1024.f);
         const float narSizeMiB = narSize / (1024.f * 1024.f);
         if (willSubstitute.size() == 1) {
-            printMsg(lvl, fmt("this path will be fetched (%.2f MiB download, %.2f MiB unpacked):",
+            printMsg(lvl, "this path will be fetched (%.2f MiB download, %.2f MiB unpacked):",
                 downloadSizeMiB,
-                narSizeMiB));
+                narSizeMiB);
         } else {
-            printMsg(lvl, fmt("these %d paths will be fetched (%.2f MiB download, %.2f MiB unpacked):",
+            printMsg(lvl, "these %d paths will be fetched (%.2f MiB download, %.2f MiB unpacked):",
                 willSubstitute.size(),
                 downloadSizeMiB,
-                narSizeMiB));
+                narSizeMiB);
         }
         for (auto & i : willSubstitute)
-            printMsg(lvl, fmt("  %s", store->printStorePath(i)));
+            printMsg(lvl, "  %s", store->printStorePath(i));
     }
 
     if (!unknown.empty()) {
-        printMsg(lvl, fmt("don't know how to build these paths%s:",
+        printMsg(lvl, "don't know how to build these paths%s:",
-                (settings.readOnlyMode ? " (may be caused by read-only store access)" : "")));
+                (settings.readOnlyMode ? " (may be caused by read-only store access)" : ""));
         for (auto & i : unknown)
-            printMsg(lvl, fmt("  %s", store->printStorePath(i)));
+            printMsg(lvl, "  %s", store->printStorePath(i));
     }
 }
 

src/libstore/binary-cache-store.hh

@@ -2,6 +2,7 @@
 
 #include "crypto.hh"
 #include "store-api.hh"
+#include "log-store.hh"
 
 #include "pool.hh"
 
@@ -28,7 +29,9 @@ struct BinaryCacheStoreConfig : virtual StoreConfig
         "other than -1 which we reserve to indicate Nix defaults should be used"};
 };
 
-class BinaryCacheStore : public virtual BinaryCacheStoreConfig, public virtual Store
+class BinaryCacheStore : public virtual BinaryCacheStoreConfig,
+    public virtual Store,
+    public virtual LogStore
 {
 
 private:

src/libstore/build-result.hh

@@ -1,6 +1,7 @@
 #pragma once
 
 #include "realisation.hh"
+#include "derived-path.hh"
 
 #include <string>
 #include <chrono>
@@ -30,6 +31,8 @@ struct BuildResult
         ResolvesToAlreadyValid,
         NoSubstituters,
     } status = MiscFailure;
+
+    // FIXME: include entire ErrorInfo object.
     std::string errorMsg;
 
     std::string toString() const {

src/libstore/build/derivation-goal.cc

@@ -204,10 +204,33 @@ void DerivationGoal::haveDerivation()
 {
     trace("have derivation");
 
-    if (drv->type() == DerivationType::CAFloating)
+    parsedDrv = std::make_unique<ParsedDerivation>(drvPath, *drv);
+
+    if (!drv->type().hasKnownOutputPaths())
         settings.requireExperimentalFeature(Xp::CaDerivations);
 
-    retrySubstitution = false;
+    if (!drv->type().isPure()) {
+        settings.requireExperimentalFeature(Xp::ImpureDerivations);
+
+        for (auto & [outputName, output] : drv->outputs) {
+            auto randomPath = StorePath::random(outputPathName(drv->name, outputName));
+            assert(!worker.store.isValidPath(randomPath));
+            initialOutputs.insert({
+                outputName,
+                InitialOutput {
+                    .wanted = true,
+                    .outputHash = impureOutputHash,
+                    .known = InitialOutputStatus {
+                        .path = randomPath,
+                        .status = PathStatus::Absent
+                    }
+                }
+            });
+        }
+
+        gaveUpOnSubstitution();
+        return;
+    }
 
     for (auto & i : drv->outputsAndOptPaths(worker.store))
         if (i.second.second)
@@ -232,9 +255,6 @@ void DerivationGoal::haveDerivation()
         return;
     }
 
-    parsedDrv = std::make_unique<ParsedDerivation>(drvPath, *drv);
-
-
     /* We are first going to try to create the invalid output paths
        through substitutes.  If that doesn't work, we'll build
        them. */
@@ -268,6 +288,8 @@ void DerivationGoal::outputsSubstitutionTried()
 {
     trace("all outputs substituted (maybe)");
 
+    assert(drv->type().isPure());
+
     if (nrFailed > 0 && nrFailed > nrNoSubstituters + nrIncompleteClosure && !settings.tryFallback) {
         done(BuildResult::TransientFailure, {},
             Error("some substitutes for the outputs of derivation '%s' failed (usually happens due to networking issues); try '--fallback' to build derivation from source ",
@@ -311,18 +333,27 @@ void DerivationGoal::outputsSubstitutionTried()
     gaveUpOnSubstitution();
 }
 
+
 /* At least one of the output paths could not be
    produced using a substitute.  So we have to build instead. */
 void DerivationGoal::gaveUpOnSubstitution()
 {
-    /* Make sure checkPathValidity() from now on checks all
-       outputs. */
-    wantedOutputs.clear();
-
     /* The inputs must be built before we can build this goal. */
+    inputDrvOutputs.clear();
     if (useDerivation)
-        for (auto & i : dynamic_cast<Derivation *>(drv.get())->inputDrvs)
+        for (auto & i : dynamic_cast<Derivation *>(drv.get())->inputDrvs) {
+            /* Ensure that pure, non-fixed-output derivations don't
+               depend on impure derivations. */
+            if (drv->type().isPure() && !drv->type().isFixed()) {
+                auto inputDrv = worker.evalStore.readDerivation(i.first);
+                if (!inputDrv.type().isPure())
+                    throw Error("pure derivation '%s' depends on impure derivation '%s'",
+                        worker.store.printStorePath(drvPath),
+                        worker.store.printStorePath(i.first));
+            }
+
             addWaitee(worker.makeDerivationGoal(i.first, i.second, buildMode == bmRepair ? bmRepair : bmNormal));
+        }
 
     /* Copy the input sources from the eval store to the build
        store. */
@@ -350,6 +381,8 @@ void DerivationGoal::gaveUpOnSubstitution()
 
 void DerivationGoal::repairClosure()
 {
+    assert(drv->type().isPure());
+
     /* If we're repairing, we now know that our own outputs are valid.
        Now check whether the other paths in the outputs closure are
        good.  If not, then start derivation goals for the derivations
@@ -426,7 +459,8 @@ void DerivationGoal::inputsRealised()
         return;
     }
 
-    if (retrySubstitution) {
+    if (retrySubstitution && !retriedSubstitution) {
+        retriedSubstitution = true;
         haveDerivation();
         return;
     }
@@ -440,19 +474,40 @@ void DerivationGoal::inputsRealised()
     if (useDerivation) {
         auto & fullDrv = *dynamic_cast<Derivation *>(drv.get());
 
-        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations) &&
+        auto drvType = fullDrv.type();
-            ((!fullDrv.inputDrvs.empty() && derivationIsCA(fullDrv.type()))
+        bool resolveDrv = std::visit(overloaded {
-            || fullDrv.type() == DerivationType::DeferredInputAddressed)) {
+            [&](const DerivationType::InputAddressed & ia) {
+                /* must resolve if deferred. */
+                return ia.deferred;
+            },
+            [&](const DerivationType::ContentAddressed & ca) {
+                return !fullDrv.inputDrvs.empty() && (
+                    ca.fixed
+                    /* Can optionally resolve if fixed, which is good
+                       for avoiding unnecessary rebuilds. */
+                    ? settings.isExperimentalFeatureEnabled(Xp::CaDerivations)
+                    /* Must resolve if floating and there are any inputs
+                       drvs. */
+                    : true);
+            },
+            [&](const DerivationType::Impure &) {
+                return true;
+            }
+        }, drvType.raw());
+
+        if (resolveDrv && !fullDrv.inputDrvs.empty()) {
+            settings.requireExperimentalFeature(Xp::CaDerivations);
+
             /* We are be able to resolve this derivation based on the
-               now-known results of dependencies. If so, we become a stub goal
+               now-known results of dependencies. If so, we become a
-               aliasing that resolved derivation goal */
+               stub goal aliasing that resolved derivation goal. */
-            std::optional attempt = fullDrv.tryResolve(worker.store);
+            std::optional attempt = fullDrv.tryResolve(worker.store, inputDrvOutputs);
             assert(attempt);
             Derivation drvResolved { *std::move(attempt) };
 
             auto pathResolved = writeDerivation(worker.store, drvResolved);
 
-            auto msg = fmt("Resolved derivation: '%s' -> '%s'",
+            auto msg = fmt("resolved derivation: '%s' -> '%s'",
                 worker.store.printStorePath(drvPath),
                 worker.store.printStorePath(pathResolved));
             act = std::make_unique<Activity>(*logger, lvlInfo, actBuildWaiting, msg,
@@ -473,21 +528,13 @@ void DerivationGoal::inputsRealised()
             /* Add the relevant output closures of the input derivation
                `i' as input paths.  Only add the closures of output paths
                that are specified as inputs. */
-            assert(worker.evalStore.isValidPath(drvPath));
+            for (auto & j : wantedDepOutputs)
-            auto outputs = worker.evalStore.queryPartialDerivationOutputMap(depDrvPath);
+                if (auto outPath = get(inputDrvOutputs, { depDrvPath, j }))
-            for (auto & j : wantedDepOutputs) {
+                    worker.store.computeFSClosure(*outPath, inputPaths);
-                if (outputs.count(j) > 0) {
+                else
-                    auto optRealizedInput = outputs.at(j);
-                    if (!optRealizedInput)
-                        throw Error(
-                            "derivation '%s' requires output '%s' from input derivation '%s', which is supposedly realized already, yet we still don't know what path corresponds to that output",
-                            worker.store.printStorePath(drvPath), j, worker.store.printStorePath(depDrvPath));
-                    worker.store.computeFSClosure(*optRealizedInput, inputPaths);
-                } else
                     throw Error(
                         "derivation '%s' requires non-existent output '%s' from input derivation '%s'",
                         worker.store.printStorePath(drvPath), j, worker.store.printStorePath(depDrvPath));
-            }
         }
     }
 
@@ -501,7 +548,7 @@ void DerivationGoal::inputsRealised()
 
     /* Don't repeat fixed-output derivations since they're already
        verified by their output hash.*/
-    nrRounds = derivationIsFixed(derivationType) ? 1 : settings.buildRepeat + 1;
+    nrRounds = derivationType.isFixed() ? 1 : settings.buildRepeat + 1;
 
     /* Okay, try to build.  Note that here we don't wait for a build
        slot to become available, since we don't need one if there is a
@@ -908,7 +955,7 @@ void DerivationGoal::buildDone()
             st =
                 dynamic_cast<NotDeterministic*>(&e) ? BuildResult::NotDeterministic :
                 statusOk(status) ? BuildResult::OutputRejected :
-                derivationIsImpure(derivationType) || diskFull ? BuildResult::TransientFailure :
+                !derivationType.isSandboxed() || diskFull ? BuildResult::TransientFailure :
                 BuildResult::PermanentFailure;
         }
 
@@ -919,60 +966,53 @@ void DerivationGoal::buildDone()
 
 void DerivationGoal::resolvedFinished()
 {
+    trace("resolved derivation finished");
+
     assert(resolvedDrvGoal);
     auto resolvedDrv = *resolvedDrvGoal->drv;
-
+    auto & resolvedResult = resolvedDrvGoal->buildResult;
-    auto resolvedHashes = staticOutputHashes(worker.store, resolvedDrv);
-
-    StorePathSet outputPaths;
-
-    // `wantedOutputs` might be empty, which means “all the outputs”
-    auto realWantedOutputs = wantedOutputs;
-    if (realWantedOutputs.empty())
-        realWantedOutputs = resolvedDrv.outputNames();
 
     DrvOutputs builtOutputs;
 
-    for (auto & wantedOutput : realWantedOutputs) {
+    if (resolvedResult.success()) {
-        assert(initialOutputs.count(wantedOutput) != 0);
+        auto resolvedHashes = staticOutputHashes(worker.store, resolvedDrv);
-        assert(resolvedHashes.count(wantedOutput) != 0);
+
-        auto realisation = worker.store.queryRealisation(
+        StorePathSet outputPaths;
-            DrvOutput{resolvedHashes.at(wantedOutput), wantedOutput}
+
-        );
+        // `wantedOutputs` might be empty, which means “all the outputs”
-        // We've just built it, but maybe the build failed, in which case the
+        auto realWantedOutputs = wantedOutputs;
-        // realisation won't be there
+        if (realWantedOutputs.empty())
-        if (realisation) {
+            realWantedOutputs = resolvedDrv.outputNames();
-            auto newRealisation = *realisation;
+
-            newRealisation.id = DrvOutput{initialOutputs.at(wantedOutput).outputHash, wantedOutput};
+        for (auto & wantedOutput : realWantedOutputs) {
-            newRealisation.signatures.clear();
+            assert(initialOutputs.count(wantedOutput) != 0);
-            newRealisation.dependentRealisations = drvOutputReferences(worker.store, *drv, realisation->outPath);
+            assert(resolvedHashes.count(wantedOutput) != 0);
-            signRealisation(newRealisation);
+            auto realisation = resolvedResult.builtOutputs.at(
-            worker.store.registerDrvOutput(newRealisation);
+                DrvOutput { resolvedHashes.at(wantedOutput), wantedOutput });
-            outputPaths.insert(realisation->outPath);
+            if (drv->type().isPure()) {
-            builtOutputs.emplace(realisation->id, *realisation);
+                auto newRealisation = realisation;
-        } else {
+                newRealisation.id = DrvOutput { initialOutputs.at(wantedOutput).outputHash, wantedOutput };
-            // If we don't have a realisation, then it must mean that something
+                newRealisation.signatures.clear();
-            // failed when building the resolved drv
+                if (!drv->type().isFixed())
-            assert(!buildResult.success());
+                    newRealisation.dependentRealisations = drvOutputReferences(worker.store, *drv, realisation.outPath);
+                signRealisation(newRealisation);
+                worker.store.registerDrvOutput(newRealisation);
+            }
+            outputPaths.insert(realisation.outPath);
+            builtOutputs.emplace(realisation.id, realisation);
         }
+
+        runPostBuildHook(
+            worker.store,
+            *logger,
+            drvPath,
+            outputPaths
+        );
     }
 
-    runPostBuildHook(
+    auto status = resolvedResult.status;
-        worker.store,
+    if (status == BuildResult::AlreadyValid)
-        *logger,
+        status = BuildResult::ResolvesToAlreadyValid;
-        drvPath,
-        outputPaths
-    );
-
-    auto status = [&]() {
-        auto & resolvedResult = resolvedDrvGoal->buildResult;
-        switch (resolvedResult.status) {
-            case BuildResult::AlreadyValid:
-                return BuildResult::ResolvesToAlreadyValid;
-            default:
-                return resolvedResult.status;
-        }
-    }();
 
     done(status, std::move(builtOutputs));
 }
@@ -1221,7 +1261,8 @@ void DerivationGoal::flushLine()
 
 std::map<std::string, std::optional<StorePath>> DerivationGoal::queryPartialDerivationOutputMap()
 {
-    if (!useDerivation || drv->type() != DerivationType::CAFloating) {
+    assert(drv->type().isPure());
+    if (!useDerivation || drv->type().hasKnownOutputPaths()) {
         std::map<std::string, std::optional<StorePath>> res;
         for (auto & [name, output] : drv->outputs)
             res.insert_or_assign(name, output.path(worker.store, drv->name, name));
@@ -1233,7 +1274,8 @@ std::map<std::string, std::optional<StorePath>> DerivationGoal::queryPartialDeri
 
 OutputPathMap DerivationGoal::queryDerivationOutputMap()
 {
-    if (!useDerivation || drv->type() != DerivationType::CAFloating) {
+    assert(drv->type().isPure());
+    if (!useDerivation || drv->type().hasKnownOutputPaths()) {
         OutputPathMap res;
         for (auto & [name, output] : drv->outputsAndOptPaths(worker.store))
             res.insert_or_assign(name, *output.second);
@@ -1246,6 +1288,8 @@ OutputPathMap DerivationGoal::queryDerivationOutputMap()
 
 std::pair<bool, DrvOutputs> DerivationGoal::checkPathValidity()
 {
+    if (!drv->type().isPure()) return { false, {} };
+
     bool checkHash = buildMode == bmRepair;
     auto wantedOutputsLeft = wantedOutputs;
     DrvOutputs validOutputs;
@@ -1289,6 +1333,7 @@ std::pair<bool, DrvOutputs> DerivationGoal::checkPathValidity()
         if (info.wanted && info.known && info.known->isValid())
             validOutputs.emplace(drvOutput, Realisation { drvOutput, info.known->path });
     }
+
     // If we requested all the outputs via the empty set, we are always fine.
     // If we requested specific elements, the loop above removes all the valid
     // ones, so any that are left must be invalid.
@@ -1326,9 +1371,7 @@ void DerivationGoal::done(
 {
     buildResult.status = status;
     if (ex)
-        // FIXME: strip: "error: "
+        buildResult.errorMsg = fmt("%s", normaltxt(ex->info().msg));
-        buildResult.errorMsg = ex->what();
-    amDone(buildResult.success() ? ecSuccess : ecFailed, ex);
     if (buildResult.status == BuildResult::TimedOut)
         worker.timedOut = true;
     if (buildResult.status == BuildResult::PermanentFailure)
@@ -1355,7 +1398,21 @@ void DerivationGoal::done(
         fs.open(traceBuiltOutputsFile, std::fstream::out);
         fs << worker.store.printStorePath(drvPath) << "\t" << buildResult.toString() << std::endl;
     }
+
+    amDone(buildResult.success() ? ecSuccess : ecFailed, ex);
 }
 
 
+void DerivationGoal::waiteeDone(GoalPtr waitee, ExitCode result)
+{
+    Goal::waiteeDone(waitee, result);
+
+    if (waitee->buildResult.success())
+        if (auto bfd = std::get_if<DerivedPath::Built>(&waitee->buildResult.path))
+            for (auto & [output, realisation] : waitee->buildResult.builtOutputs)
+                inputDrvOutputs.insert_or_assign(
+                    { bfd->drvPath, output.outputName },
+                    realisation.outPath);
+}
+
 }

src/libstore/build/derivation-goal.hh

@@ -57,12 +57,21 @@ struct DerivationGoal : public Goal
        them. */
     StringSet wantedOutputs;
 
+    /* Mapping from input derivations + output names to actual store
+       paths. This is filled in by waiteeDone() as each dependency
+       finishes, before inputsRealised() is reached, */
+    std::map<std::pair<StorePath, std::string>, StorePath> inputDrvOutputs;
+
     /* Whether additional wanted outputs have been added. */
     bool needRestart = false;
 
     /* Whether to retry substituting the outputs after building the
-       inputs. */
+       inputs. This is done in case of an incomplete closure. */
-    bool retrySubstitution;
+    bool retrySubstitution = false;
+
+    /* Whether we've retried substitution, in which case we won't try
+       again. */
+    bool retriedSubstitution = false;
 
     /* The derivation stored at drvPath. */
     std::unique_ptr<Derivation> drv;
@@ -220,6 +229,8 @@ struct DerivationGoal : public Goal
         DrvOutputs builtOutputs = {},
         std::optional<Error> ex = {});
 
+    void waiteeDone(GoalPtr waitee, ExitCode result) override;
+
     StorePathSet exportReferences(const StorePathSet & storePaths);
 };
 

src/libstore/build/drv-output-substitution-goal.cc

@@ -41,7 +41,7 @@ void DrvOutputSubstitutionGoal::tryNext()
     if (subs.size() == 0) {
         /* None left.  Terminate this goal and let someone else deal
            with it. */
-        debug("drv output '%s' is required, but there is no substituter that can provide it", id.to_string());
+        debug("derivation output '%s' is required, but there is no substituter that can provide it", id.to_string());
 
         /* Hack: don't indicate failure if there were no substituters.
            In that case the calling derivation should just do a

src/libstore/build/goal.cc

@@ -28,7 +28,7 @@ void Goal::addWaitee(GoalPtr waitee)
 
 void Goal::waiteeDone(GoalPtr waitee, ExitCode result)
 {
-    assert(waitees.find(waitee) != waitees.end());
+    assert(waitees.count(waitee));
     waitees.erase(waitee);
 
     trace(fmt("waitee '%s' done; %d left", waitee->name, waitees.size()));

src/libstore/build/goal.hh

@@ -40,21 +40,21 @@ struct Goal : public std::enable_shared_from_this<Goal>
     WeakGoals waiters;
 
     /* Number of goals we are/were waiting for that have failed. */
-    unsigned int nrFailed;
+    size_t nrFailed = 0;
 
     /* Number of substitution goals we are/were waiting for that
        failed because there are no substituters. */
-    unsigned int nrNoSubstituters;
+    size_t nrNoSubstituters = 0;
 
     /* Number of substitution goals we are/were waiting for that
        failed because they had unsubstitutable references. */
-    unsigned int nrIncompleteClosure;
+    size_t nrIncompleteClosure = 0;
 
     /* Name of this goal for debugging purposes. */
     std::string name;
 
     /* Whether the goal is finished. */
-    ExitCode exitCode;
+    ExitCode exitCode = ecBusy;
 
     /* Build result. */
     BuildResult buildResult;
@@ -65,10 +65,7 @@ struct Goal : public std::enable_shared_from_this<Goal>
     Goal(Worker & worker, DerivedPath path)
         : worker(worker)
         , buildResult { .path = std::move(path) }
-    {
+    { }
-        nrFailed = nrNoSubstituters = nrIncompleteClosure = 0;
-        exitCode = ecBusy;
-    }
 
     virtual ~Goal()
     {

src/libstore/build/local-derivation-goal.cc

@@ -395,7 +395,7 @@ void LocalDerivationGoal::startBuilder()
         else if (settings.sandboxMode == smDisabled)
             useChroot = false;
         else if (settings.sandboxMode == smRelaxed)
-            useChroot = !(derivationIsImpure(derivationType)) && !noChroot;
+            useChroot = derivationType.isSandboxed() && !noChroot;
     }
 
     auto & localStore = getLocalStore();
@@ -608,7 +608,7 @@ void LocalDerivationGoal::startBuilder()
                 "nogroup:x:65534:\n", sandboxGid()));
 
         /* Create /etc/hosts with localhost entry. */
-        if (!(derivationIsImpure(derivationType)))
+        if (derivationType.isSandboxed())
             writeFile(chrootRootDir + "/etc/hosts", "127.0.0.1 localhost\n::1 localhost\n");
 
         /* Make the closure of the inputs available in the chroot,
@@ -704,6 +704,9 @@ void LocalDerivationGoal::startBuilder()
 
     /* Run the builder. */
     printMsg(lvlChatty, "executing builder '%1%'", drv->builder);
+    printMsg(lvlChatty, "using builder args '%1%'", concatStringsSep(" ", drv->args));
+    for (auto & i : drv->env)
+        printMsg(lvlVomit, "setting builder env variable '%1%'='%2%'", i.first, i.second);
 
     /* Create the log file. */
     Path logFile = openLogFile();
@@ -796,7 +799,7 @@ void LocalDerivationGoal::startBuilder()
            us.
         */
 
-        if (!(derivationIsImpure(derivationType)))
+        if (derivationType.isSandboxed())
             privateNetwork = true;
 
         userNamespaceSync.create();
@@ -1049,7 +1052,7 @@ void LocalDerivationGoal::initEnv()
        derivation, tell the builder, so that for instance `fetchurl'
        can skip checking the output.  On older Nixes, this environment
        variable won't be set, so `fetchurl' will do the check. */
-    if (derivationIsFixed(derivationType)) env["NIX_OUTPUT_CHECKED"] = "1";
+    if (derivationType.isFixed()) env["NIX_OUTPUT_CHECKED"] = "1";
 
     /* *Only* if this is a fixed-output derivation, propagate the
        values of the environment variables specified in the
@@ -1060,7 +1063,7 @@ void LocalDerivationGoal::initEnv()
        to the builder is generally impure, but the output of
        fixed-output derivations is by definition pure (since we
        already know the cryptographic hash of the output). */
-    if (derivationIsImpure(derivationType)) {
+    if (!derivationType.isSandboxed()) {
         for (auto & i : parsedDrv->getStringsAttr("impureEnvVars").value_or(Strings()))
             env[i] = getEnv(i).value_or("");
     }
@@ -1340,6 +1343,12 @@ struct RestrictedStore : public virtual RestrictedStoreConfig, public virtual Lo
         next->queryMissing(allowed, willBuild, willSubstitute,
             unknown, downloadSize, narSize);
     }
+
+    virtual std::optional<std::string> getBuildLog(const StorePath & path) override
+    { return std::nullopt; }
+
+    virtual void addBuildLog(const StorePath & path, std::string_view log) override
+    { unsupported("addBuildLog"); }
 };
 
 
@@ -1668,7 +1677,7 @@ void LocalDerivationGoal::runChild()
             /* Fixed-output derivations typically need to access the
                network, so give them access to /etc/resolv.conf and so
                on. */
-            if (derivationIsImpure(derivationType)) {
+            if (!derivationType.isSandboxed()) {
                 // Only use nss functions to resolve hosts and
                 // services. Don’t use it for anything else that may
                 // be configured for this system. This limits the
@@ -1912,7 +1921,7 @@ void LocalDerivationGoal::runChild()
 
                 sandboxProfile += "(import \"sandbox-defaults.sb\")\n";
 
-                if (derivationIsImpure(derivationType))
+                if (!derivationType.isSandboxed())
                     sandboxProfile += "(import \"sandbox-network.sb\")\n";
 
                 /* Add the output paths we'll use at build-time to the chroot */
@@ -2273,7 +2282,7 @@ DrvOutputs LocalDerivationGoal::registerOutputs()
             return res;
         };
 
-        auto newInfoFromCA = [&](const DerivationOutputCAFloating outputHash) -> ValidPathInfo {
+        auto newInfoFromCA = [&](const DerivationOutput::CAFloating outputHash) -> ValidPathInfo {
             auto & st = outputStats.at(outputName);
             if (outputHash.method == FileIngestionMethod::Flat) {
                 /* The output path should be a regular file without execute permission. */
@@ -2340,7 +2349,7 @@ DrvOutputs LocalDerivationGoal::registerOutputs()
 
         ValidPathInfo newInfo = std::visit(overloaded {
 
-            [&](const DerivationOutputInputAddressed & output) {
+            [&](const DerivationOutput::InputAddressed & output) {
                 /* input-addressed case */
                 auto requiredFinalPath = output.path;
                 /* Preemptively add rewrite rule for final hash, as that is
@@ -2360,8 +2369,8 @@ DrvOutputs LocalDerivationGoal::registerOutputs()
                 return newInfo0;
             },
 
-            [&](const DerivationOutputCAFixed & dof) {
+            [&](const DerivationOutput::CAFixed & dof) {
-                auto newInfo0 = newInfoFromCA(DerivationOutputCAFloating {
+                auto newInfo0 = newInfoFromCA(DerivationOutput::CAFloating {
                     .method = dof.hash.method,
                     .hashType = dof.hash.hash.type,
                 });
@@ -2383,17 +2392,24 @@ DrvOutputs LocalDerivationGoal::registerOutputs()
                 return newInfo0;
             },
 
-            [&](DerivationOutputCAFloating & dof) {
+            [&](const DerivationOutput::CAFloating & dof) {
                 return newInfoFromCA(dof);
             },
 
-            [&](DerivationOutputDeferred) -> ValidPathInfo {
+            [&](const DerivationOutput::Deferred &) -> ValidPathInfo {
                 // No derivation should reach that point without having been
                 // rewritten first
                 assert(false);
             },
 
-        }, output.output);
+            [&](const DerivationOutput::Impure & doi) {
+                return newInfoFromCA(DerivationOutput::CAFloating {
+                    .method = doi.method,
+                    .hashType = doi.hashType,
+                });
+            },
+
+        }, output.raw());
 
         /* FIXME: set proper permissions in restorePath() so
             we don't have to do another traversal. */
@@ -2603,11 +2619,14 @@ DrvOutputs LocalDerivationGoal::registerOutputs()
             },
             .outPath = newInfo.path
         };
-        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations)) {
+        if (settings.isExperimentalFeatureEnabled(Xp::CaDerivations)
+            && drv->type().isPure())
+        {
             signRealisation(thisRealisation);
             worker.store.registerDrvOutput(thisRealisation);
         }
-        builtOutputs.emplace(thisRealisation.id, thisRealisation);
+        if (wantOutput(outputName, wantedOutputs))
+            builtOutputs.emplace(thisRealisation.id, thisRealisation);
     }
 
     return builtOutputs;

src/libstore/build/substitution-goal.cc

@@ -24,9 +24,16 @@ PathSubstitutionGoal::~PathSubstitutionGoal()
 }
 
 
-void PathSubstitutionGoal::done(ExitCode result, BuildResult::Status status)
+void PathSubstitutionGoal::done(
+    ExitCode result,
+    BuildResult::Status status,
+    std::optional<std::string> errorMsg)
 {
     buildResult.status = status;
+    if (errorMsg) {
+        debug(*errorMsg);
+        buildResult.errorMsg = *errorMsg;
+    }
     amDone(result);
 }
 
@@ -67,12 +74,14 @@ void PathSubstitutionGoal::tryNext()
     if (subs.size() == 0) {
         /* None left.  Terminate this goal and let someone else deal
            with it. */
-        debug("path '%s' is required, but there is no substituter that can build it", worker.store.printStorePath(storePath));
 
         /* Hack: don't indicate failure if there were no substituters.
            In that case the calling derivation should just do a
            build. */
-        done(substituterFailed ? ecFailed : ecNoSubstituters, BuildResult::NoSubstituters);
+        done(
+            substituterFailed ? ecFailed : ecNoSubstituters,
+            BuildResult::NoSubstituters,
+            fmt("path '%s' is required, but there is no substituter that can build it", worker.store.printStorePath(storePath)));
 
         if (substituterFailed) {
             worker.failedSubstitutions++;
@@ -169,10 +178,10 @@ void PathSubstitutionGoal::referencesValid()
     trace("all references realised");
 
     if (nrFailed > 0) {
-        debug("some references of path '%s' could not be realised", worker.store.printStorePath(storePath));
         done(
             nrNoSubstituters > 0 || nrIncompleteClosure > 0 ? ecIncompleteClosure : ecFailed,
-            BuildResult::DependencyFailed);
+            BuildResult::DependencyFailed,
+            fmt("some references of path '%s' could not be realised", worker.store.printStorePath(storePath)));
         return;
     }
 

src/libstore/build/substitution-goal.hh

@@ -53,7 +53,10 @@ struct PathSubstitutionGoal : public Goal
     /* Content address for recomputing store path */
     std::optional<ContentAddress> ca;
 
-    void done(ExitCode result, BuildResult::Status status);
+    void done(
+        ExitCode result,
+        BuildResult::Status status,
+        std::optional<std::string> errorMsg = {});
 
 public:
     PathSubstitutionGoal(const StorePath & storePath, Worker & worker, RepairFlag repair = NoRepair, std::optional<ContentAddress> ca = std::nullopt);

src/libstore/builtins/buildenv.cc

@@ -47,9 +47,9 @@ static void createLinks(State & state, const Path & srcDir, const Path & dstDir,
             throw;
         }
 
-        /* The files below are special-cased to that they don't show up
+        /* The files below are special-cased to that they don't show
-         * in user profiles, either because they are useless, or
+         * up in user profiles, either because they are useless, or
-         * because they would cauase pointless collisions (e.g., each
+         * because they would cause pointless collisions (e.g., each
          * Python package brings its own
          * `$out/lib/pythonX.Y/site-packages/easy-install.pth'.)
          */
@@ -57,7 +57,9 @@ static void createLinks(State & state, const Path & srcDir, const Path & dstDir,
             hasSuffix(srcFile, "/nix-support") ||
             hasSuffix(srcFile, "/perllocal.pod") ||
             hasSuffix(srcFile, "/info/dir") ||
-            hasSuffix(srcFile, "/log"))
+            hasSuffix(srcFile, "/log") ||
+            hasSuffix(srcFile, "/manifest.nix") ||
+            hasSuffix(srcFile, "/manifest.json"))
             continue;
 
         else if (S_ISDIR(srcSt.st_mode)) {

src/libstore/ca-specific-schema.sql

@@ -13,12 +13,27 @@ create table if not exists Realisations (
 
 create index if not exists IndexRealisations on Realisations(drvPath, outputName);
 
+-- We can end-up in a weird edge-case where a path depends on itself because
+-- it’s an output of a CA derivation, that happens to be the same as one of its
+-- dependencies.
+-- In that case we have a dependency loop (path -> realisation1 -> realisation2
+-- -> path) that we need to break by removing the dependencies between the
+-- realisations
+create trigger if not exists DeleteSelfRefsViaRealisations before delete on ValidPaths
+  begin
+    delete from RealisationsRefs where realisationReference in (
+      select id from Realisations where outputPath = old.id
+    );
+  end;
+
 create table if not exists RealisationsRefs (
     referrer integer not null,
     realisationReference integer,
     foreign key (referrer) references Realisations(id) on delete cascade,
     foreign key (realisationReference) references Realisations(id) on delete restrict
 );
+-- used by deletion trigger
+create index if not exists IndexRealisationsRefsRealisationReference on RealisationsRefs(realisationReference);
 
 -- used by QueryRealisationReferences
 create index if not exists IndexRealisationsRefs on RealisationsRefs(referrer);

src/libstore/daemon.cc

@@ -3,7 +3,9 @@
 #include "worker-protocol.hh"
 #include "build-result.hh"
 #include "store-api.hh"
+#include "store-cast.hh"
 #include "gc-store.hh"
+#include "log-store.hh"
 #include "path-with-outputs.hh"
 #include "finally.hh"
 #include "archive.hh"
@@ -558,6 +560,8 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
         BuildMode buildMode = (BuildMode) readInt(from);
         logger->startWork();
 
+        auto drvType = drv.type();
+
         /* Content-addressed derivations are trustless because their output paths
            are verified by their content alone, so any derivation is free to
            try to produce such a path.
@@ -590,12 +594,12 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
            derivations, we throw out the precomputed output paths and just
            store the hashes, so there aren't two competing sources of truth an
            attacker could exploit. */
-        if (drv.type() == DerivationType::InputAddressed && !trusted)
+        if (!(drvType.isCA() || trusted))
             throw Error("you are not privileged to build input-addressed derivations");
 
         /* Make sure that the non-input-addressed derivations that got this far
            are in fact content-addressed if we don't trust them. */
-        assert(derivationIsCA(drv.type()) || trusted);
+        assert(drvType.isCA() || trusted);
 
         /* Recompute the derivation path when we cannot trust the original. */
         if (!trusted) {
@@ -604,7 +608,7 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
                original not-necessarily-resolved derivation to verify the drv
                derivation as adequate claim to the input-addressed output
                paths. */
-            assert(derivationIsCA(drv.type()));
+            assert(drvType.isCA());
 
             Derivation drv2;
             static_cast<BasicDerivation &>(drv2) = drv;
@@ -645,7 +649,7 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
         Path path = absPath(readString(from));
 
         logger->startWork();
-        auto & gcStore = requireGcStore(*store);
+        auto & gcStore = require<GcStore>(*store);
         gcStore.addIndirectRoot(path);
         logger->stopWork();
 
@@ -663,7 +667,7 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
 
     case wopFindRoots: {
         logger->startWork();
-        auto & gcStore = requireGcStore(*store);
+        auto & gcStore = require<GcStore>(*store);
         Roots roots = gcStore.findRoots(!trusted);
         logger->stopWork();
 
@@ -695,7 +699,7 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
         logger->startWork();
         if (options.ignoreLiveness)
             throw Error("you are not allowed to ignore liveness");
-        auto & gcStore = requireGcStore(*store);
+        auto & gcStore = require<GcStore>(*store);
         gcStore.collectGarbage(options, results);
         logger->stopWork();
 
@@ -953,11 +957,12 @@ static void performOp(TunnelLogger * logger, ref<Store> store,
         logger->startWork();
         if (!trusted)
             throw Error("you are not privileged to add logs");
+        auto & logStore = require<LogStore>(*store);
         {
             FramedSource source(from);
             StringSink sink;
             source.drainInto(sink);
-            store->addBuildLog(path, sink.s);
+            logStore.addBuildLog(path, sink.s);
         }
         logger->stopWork();
         to << 1;

src/libstore/derivations.cc

@@ -11,72 +11,114 @@ namespace nix {
 std::optional<StorePath> DerivationOutput::path(const Store & store, std::string_view drvName, std::string_view outputName) const
 {
     return std::visit(overloaded {
-        [](const DerivationOutputInputAddressed & doi) -> std::optional<StorePath> {
+        [](const DerivationOutput::InputAddressed & doi) -> std::optional<StorePath> {
             return { doi.path };
         },
-        [&](const DerivationOutputCAFixed & dof) -> std::optional<StorePath> {
+        [&](const DerivationOutput::CAFixed & dof) -> std::optional<StorePath> {
             return {
                 dof.path(store, drvName, outputName)
             };
         },
-        [](const DerivationOutputCAFloating & dof) -> std::optional<StorePath> {
+        [](const DerivationOutput::CAFloating & dof) -> std::optional<StorePath> {
             return std::nullopt;
         },
-        [](const DerivationOutputDeferred &) -> std::optional<StorePath> {
+        [](const DerivationOutput::Deferred &) -> std::optional<StorePath> {
             return std::nullopt;
         },
-    }, output);
+        [](const DerivationOutput::Impure &) -> std::optional<StorePath> {
+            return std::nullopt;
+        },
+    }, raw());
 }
 
 
-StorePath DerivationOutputCAFixed::path(const Store & store, std::string_view drvName, std::string_view outputName) const {
+StorePath DerivationOutput::CAFixed::path(const Store & store, std::string_view drvName, std::string_view outputName) const
+{
     return store.makeFixedOutputPath(
         hash.method, hash.hash,
         outputPathName(drvName, outputName));
 }
 
 
-bool derivationIsCA(DerivationType dt) {
+bool DerivationType::isCA() const
-    switch (dt) {
+{
-    case DerivationType::InputAddressed: return false;
+    /* Normally we do the full `std::visit` to make sure we have
-    case DerivationType::CAFixed: return true;
+       exhaustively handled all variants, but so long as there is a
-    case DerivationType::CAFloating: return true;
+       variant called `ContentAddressed`, it must be the only one for
-    case DerivationType::DeferredInputAddressed: return false;
+       which `isCA` is true for this to make sense!. */
-    };
+    return std::visit(overloaded {
-    // Since enums can have non-variant values, but making a `default:` would
+        [](const InputAddressed & ia) {
-    // disable exhaustiveness warnings.
+            return false;
-    assert(false);
+        },
+        [](const ContentAddressed & ca) {
+            return true;
+        },
+        [](const Impure &) {
+            return true;
+        },
+    }, raw());
 }
 
-bool derivationIsFixed(DerivationType dt) {
+bool DerivationType::isFixed() const
-    switch (dt) {
+{
-    case DerivationType::InputAddressed: return false;
+    return std::visit(overloaded {
-    case DerivationType::CAFixed: return true;
+        [](const InputAddressed & ia) {
-    case DerivationType::CAFloating: return false;
+            return false;
-    case DerivationType::DeferredInputAddressed: return false;
+        },
-    };
+        [](const ContentAddressed & ca) {
-    assert(false);
+            return ca.fixed;
+        },
+        [](const Impure &) {
+            return false;
+        },
+    }, raw());
 }
 
-bool derivationHasKnownOutputPaths(DerivationType dt) {
+bool DerivationType::hasKnownOutputPaths() const
-    switch (dt) {
+{
-    case DerivationType::InputAddressed: return true;
+    return std::visit(overloaded {
-    case DerivationType::CAFixed: return true;
+        [](const InputAddressed & ia) {
-    case DerivationType::CAFloating: return false;
+            return !ia.deferred;
-    case DerivationType::DeferredInputAddressed: return false;
+        },
-    };
+        [](const ContentAddressed & ca) {
-    assert(false);
+            return ca.fixed;
+        },
+        [](const Impure &) {
+            return false;
+        },
+    }, raw());
 }
 
 
-bool derivationIsImpure(DerivationType dt) {
+bool DerivationType::isSandboxed() const
-    switch (dt) {
+{
-    case DerivationType::InputAddressed: return false;
+    return std::visit(overloaded {
-    case DerivationType::CAFixed: return true;
+        [](const InputAddressed & ia) {
-    case DerivationType::CAFloating: return false;
+            return true;
-    case DerivationType::DeferredInputAddressed: return false;
+        },
-    };
+        [](const ContentAddressed & ca) {
-    assert(false);
+            return ca.sandboxed;
+        },
+        [](const Impure &) {
+            return false;
+        },
+    }, raw());
+}
+
+
+bool DerivationType::isPure() const
+{
+    return std::visit(overloaded {
+        [](const InputAddressed & ia) {
+            return true;
+        },
+        [](const ContentAddressed & ca) {
+            return true;
+        },
+        [](const Impure &) {
+            return false;
+        },
+    }, raw());
 }
 
 
@@ -177,37 +219,36 @@ static DerivationOutput parseDerivationOutput(const Store & store,
             hashAlgo = hashAlgo.substr(2);
         }
         const auto hashType = parseHashType(hashAlgo);
-        if (hash != "") {
+        if (hash == "impure") {
+            settings.requireExperimentalFeature(Xp::ImpureDerivations);
+            assert(pathS == "");
+            return DerivationOutput::Impure {
+                .method = std::move(method),
+                .hashType = std::move(hashType),
+            };
+        } else if (hash != "") {
             validatePath(pathS);
-            return DerivationOutput {
+            return DerivationOutput::CAFixed {
-                .output = DerivationOutputCAFixed {
+                .hash = FixedOutputHash {
-                    .hash = FixedOutputHash {
+                    .method = std::move(method),
-                        .method = std::move(method),
+                    .hash = Hash::parseNonSRIUnprefixed(hash, hashType),
-                        .hash = Hash::parseNonSRIUnprefixed(hash, hashType),
-                    },
                 },
             };
         } else {
             settings.requireExperimentalFeature(Xp::CaDerivations);
             assert(pathS == "");
-            return DerivationOutput {
+            return DerivationOutput::CAFloating {
-                .output = DerivationOutputCAFloating {
+                .method = std::move(method),
-                    .method = std::move(method),
+                .hashType = std::move(hashType),
-                    .hashType = std::move(hashType),
-                },
             };
         }
     } else {
         if (pathS == "") {
-            return DerivationOutput {
+            return DerivationOutput::Deferred { };
-                .output = DerivationOutputDeferred { }
-            };
         }
         validatePath(pathS);
-        return DerivationOutput {
+        return DerivationOutput::InputAddressed {
-            .output = DerivationOutputInputAddressed {
+            .path = store.parseStorePath(pathS),
-                .path = store.parseStorePath(pathS),
-            }
         };
     }
 }
@@ -335,27 +376,33 @@ std::string Derivation::unparse(const Store & store, bool maskOutputs,
         if (first) first = false; else s += ',';
         s += '('; printUnquotedString(s, i.first);
         std::visit(overloaded {
-            [&](const DerivationOutputInputAddressed & doi) {
+            [&](const DerivationOutput::InputAddressed & doi) {
                 s += ','; printUnquotedString(s, maskOutputs ? "" : store.printStorePath(doi.path));
                 s += ','; printUnquotedString(s, "");
                 s += ','; printUnquotedString(s, "");
             },
-            [&](const DerivationOutputCAFixed & dof) {
+            [&](const DerivationOutput::CAFixed & dof) {
                 s += ','; printUnquotedString(s, maskOutputs ? "" : store.printStorePath(dof.path(store, name, i.first)));
                 s += ','; printUnquotedString(s, dof.hash.printMethodAlgo());
                 s += ','; printUnquotedString(s, dof.hash.hash.to_string(Base16, false));
             },
-            [&](const DerivationOutputCAFloating & dof) {
+            [&](const DerivationOutput::CAFloating & dof) {
                 s += ','; printUnquotedString(s, "");
                 s += ','; printUnquotedString(s, makeFileIngestionPrefix(dof.method) + printHashType(dof.hashType));
                 s += ','; printUnquotedString(s, "");
             },
-            [&](const DerivationOutputDeferred &) {
+            [&](const DerivationOutput::Deferred &) {
                 s += ','; printUnquotedString(s, "");
                 s += ','; printUnquotedString(s, "");
                 s += ','; printUnquotedString(s, "");
+            },
+            [&](const DerivationOutputImpure & doi) {
+                // FIXME
+                s += ','; printUnquotedString(s, "");
+                s += ','; printUnquotedString(s, makeFileIngestionPrefix(doi.method) + printHashType(doi.hashType));
+                s += ','; printUnquotedString(s, "impure");
             }
-        }, i.second.output);
+        }, i.second.raw());
         s += ')';
     }
 
@@ -419,49 +466,100 @@ std::string outputPathName(std::string_view drvName, std::string_view outputName
 
 DerivationType BasicDerivation::type() const
 {
-    std::set<std::string_view> inputAddressedOutputs, fixedCAOutputs, floatingCAOutputs, deferredIAOutputs;
+    std::set<std::string_view>
+        inputAddressedOutputs,
+        fixedCAOutputs,
+        floatingCAOutputs,
+        deferredIAOutputs,
+        impureOutputs;
     std::optional<HashType> floatingHashType;
+
     for (auto & i : outputs) {
         std::visit(overloaded {
-            [&](const DerivationOutputInputAddressed &) {
+            [&](const DerivationOutput::InputAddressed &) {
                inputAddressedOutputs.insert(i.first);
             },
-            [&](const DerivationOutputCAFixed &) {
+            [&](const DerivationOutput::CAFixed &) {
                 fixedCAOutputs.insert(i.first);
             },
-            [&](const DerivationOutputCAFloating & dof) {
+            [&](const DerivationOutput::CAFloating & dof) {
                 floatingCAOutputs.insert(i.first);
                 if (!floatingHashType) {
                     floatingHashType = dof.hashType;
                 } else {
                     if (*floatingHashType != dof.hashType)
-                        throw Error("All floating outputs must use the same hash type");
+                        throw Error("all floating outputs must use the same hash type");
                 }
             },
-            [&](const DerivationOutputDeferred &) {
+            [&](const DerivationOutput::Deferred &) {
-               deferredIAOutputs.insert(i.first);
+                deferredIAOutputs.insert(i.first);
             },
-        }, i.second.output);
+            [&](const DerivationOutput::Impure &) {
+                impureOutputs.insert(i.first);
+            },
+        }, i.second.raw());
     }
 
-    if (inputAddressedOutputs.empty() && fixedCAOutputs.empty() && floatingCAOutputs.empty() && deferredIAOutputs.empty()) {
+    if (inputAddressedOutputs.empty()
-        throw Error("Must have at least one output");
+        && fixedCAOutputs.empty()
-    } else if (! inputAddressedOutputs.empty() && fixedCAOutputs.empty() && floatingCAOutputs.empty() && deferredIAOutputs.empty()) {
+        && floatingCAOutputs.empty()
-        return DerivationType::InputAddressed;
+        && deferredIAOutputs.empty()
-    } else if (inputAddressedOutputs.empty() && ! fixedCAOutputs.empty() && floatingCAOutputs.empty() && deferredIAOutputs.empty()) {
+        && impureOutputs.empty())
+        throw Error("must have at least one output");
+
+    if (!inputAddressedOutputs.empty()
+        && fixedCAOutputs.empty()
+        && floatingCAOutputs.empty()
+        && deferredIAOutputs.empty()
+        && impureOutputs.empty())
+        return DerivationType::InputAddressed {
+            .deferred = false,
+        };
+
+    if (inputAddressedOutputs.empty()
+        && !fixedCAOutputs.empty()
+        && floatingCAOutputs.empty()
+        && deferredIAOutputs.empty()
+        && impureOutputs.empty())
+    {
         if (fixedCAOutputs.size() > 1)
             // FIXME: Experimental feature?
-            throw Error("Only one fixed output is allowed for now");
+            throw Error("only one fixed output is allowed for now");
         if (*fixedCAOutputs.begin() != "out")
-            throw Error("Single fixed output must be named \"out\"");
+            throw Error("single fixed output must be named \"out\"");
-        return DerivationType::CAFixed;
+        return DerivationType::ContentAddressed {
-    } else if (inputAddressedOutputs.empty() && fixedCAOutputs.empty() && ! floatingCAOutputs.empty() && deferredIAOutputs.empty()) {
+            .sandboxed = false,
-        return DerivationType::CAFloating;
+            .fixed = true,
-    } else if (inputAddressedOutputs.empty() && fixedCAOutputs.empty() && floatingCAOutputs.empty() && !deferredIAOutputs.empty()) {
+        };
-        return DerivationType::DeferredInputAddressed;
-    } else {
-        throw Error("Can't mix derivation output types");
     }
+
+    if (inputAddressedOutputs.empty()
+        && fixedCAOutputs.empty()
+        && !floatingCAOutputs.empty()
+        && deferredIAOutputs.empty()
+        && impureOutputs.empty())
+        return DerivationType::ContentAddressed {
+            .sandboxed = true,
+            .fixed = false,
+        };
+
+    if (inputAddressedOutputs.empty()
+        && fixedCAOutputs.empty()
+        && floatingCAOutputs.empty()
+        && !deferredIAOutputs.empty()
+        && impureOutputs.empty())
+        return DerivationType::InputAddressed {
+            .deferred = true,
+        };
+
+    if (inputAddressedOutputs.empty()
+        && fixedCAOutputs.empty()
+        && floatingCAOutputs.empty()
+        && deferredIAOutputs.empty()
+        && !impureOutputs.empty())
+        return DerivationType::Impure { };
+
+    throw Error("can't mix derivation output types");
 }
 
 
@@ -473,7 +571,7 @@ Sync<DrvHashes> drvHashes;
 /* Look up the derivation by value and memoize the
    `hashDerivationModulo` call.
  */
-static const DrvHashModulo pathDerivationModulo(Store & store, const StorePath & drvPath)
+static const DrvHash pathDerivationModulo(Store & store, const StorePath & drvPath)
 {
     {
         auto hashes = drvHashes.lock();
@@ -508,88 +606,83 @@ static const DrvHashModulo pathDerivationModulo(Store & store, const StorePath &
    don't leak the provenance of fixed outputs, reducing pointless cache
    misses as the build itself won't know this.
  */
-DrvHashModulo hashDerivationModulo(Store & store, const Derivation & drv, bool maskOutputs)
+DrvHash hashDerivationModulo(Store & store, const Derivation & drv, bool maskOutputs)
 {
-    bool isDeferred = false;
+    auto type = drv.type();
+
     /* Return a fixed hash for fixed-output derivations. */
-    switch (drv.type()) {
+    if (type.isFixed()) {
-    case DerivationType::CAFixed: {
         std::map<std::string, Hash> outputHashes;
         for (const auto & i : drv.outputs) {
-            auto & dof = std::get<DerivationOutputCAFixed>(i.second.output);
+            auto & dof = std::get<DerivationOutput::CAFixed>(i.second.raw());
             auto hash = hashString(htSHA256, "fixed:out:"
                 + dof.hash.printMethodAlgo() + ":"
                 + dof.hash.hash.to_string(Base16, false) + ":"
                 + store.printStorePath(dof.path(store, drv.name, i.first)));
             outputHashes.insert_or_assign(i.first, std::move(hash));
         }
-        return outputHashes;
+        return DrvHash {
-    }
+            .hashes = outputHashes,
-    case DerivationType::CAFloating:
+            .kind = DrvHash::Kind::Regular,
-        isDeferred = true;
+        };
-        break;
-    case DerivationType::InputAddressed:
-        break;
-    case DerivationType::DeferredInputAddressed:
-        break;
     }
 
-    /* For other derivations, replace the inputs paths with recursive
+    if (!type.isPure()) {
-       calls to this function. */
+        std::map<std::string, Hash> outputHashes;
+        for (const auto & [outputName, _] : drv.outputs)
+            outputHashes.insert_or_assign(outputName, impureOutputHash);
+        return DrvHash {
+            .hashes = outputHashes,
+            .kind = DrvHash::Kind::Deferred,
+        };
+    }
+
+    auto kind = std::visit(overloaded {
+        [](const DerivationType::InputAddressed & ia) {
+            /* This might be a "pesimistically" deferred output, so we don't
+               "taint" the kind yet. */
+            return DrvHash::Kind::Regular;
+        },
+        [](const DerivationType::ContentAddressed & ca) {
+            return ca.fixed
+                ? DrvHash::Kind::Regular
+                : DrvHash::Kind::Deferred;
+        },
+        [](const DerivationType::Impure &) -> DrvHash::Kind {
+            assert(false);
+        }
+    }, drv.type().raw());
+
     std::map<std::string, StringSet> inputs2;
-    for (auto & i : drv.inputDrvs) {
+    for (auto & [drvPath, inputOutputs0] : drv.inputDrvs) {
-        const auto & res = pathDerivationModulo(store, i.first);
+        // Avoid lambda capture restriction with standard / Clang
-        std::visit(overloaded {
+        auto & inputOutputs = inputOutputs0;
-            // Regular non-CA derivation, replace derivation
+        const auto & res = pathDerivationModulo(store, drvPath);
-            [&](const Hash & drvHash) {
+        if (res.kind == DrvHash::Kind::Deferred)
-                inputs2.insert_or_assign(drvHash.to_string(Base16, false), i.second);
+            kind = DrvHash::Kind::Deferred;
-            },
+        for (auto & outputName : inputOutputs) {
-            [&](const DeferredHash & deferredHash) {
+            const auto h = res.hashes.at(outputName);
-                isDeferred = true;
+            inputs2[h.to_string(Base16, false)].insert(outputName);
-                inputs2.insert_or_assign(deferredHash.hash.to_string(Base16, false), i.second);
+        }
-            },
-            // CA derivation's output hashes
-            [&](const CaOutputHashes & outputHashes) {
-                std::set<std::string> justOut = { "out" };
-                for (auto & output : i.second) {
-                    /* Put each one in with a single "out" output.. */
-                    const auto h = outputHashes.at(output);
-                    inputs2.insert_or_assign(
-                        h.to_string(Base16, false),
-                        justOut);
-                }
-            },
-        }, res);
     }
 
     auto hash = hashString(htSHA256, drv.unparse(store, maskOutputs, &inputs2));
 
-    if (isDeferred)
+    std::map<std::string, Hash> outputHashes;
-        return DeferredHash { hash };
+    for (const auto & [outputName, _] : drv.outputs) {
-    else
+        outputHashes.insert_or_assign(outputName, hash);
-        return hash;
+    }
+
+    return DrvHash {
+        .hashes = outputHashes,
+        .kind = kind,
+    };
 }
 
 
 std::map<std::string, Hash> staticOutputHashes(Store & store, const Derivation & drv)
 {
-    std::map<std::string, Hash> res;
+    return hashDerivationModulo(store, drv, true).hashes;
-    std::visit(overloaded {
-        [&](const Hash & drvHash) {
-            for (auto & outputName : drv.outputNames()) {
-                res.insert({outputName, drvHash});
-            }
-        },
-        [&](const DeferredHash & deferredHash) {
-            for (auto & outputName : drv.outputNames()) {
-                res.insert({outputName, deferredHash.hash});
-            }
-        },
-        [&](const CaOutputHashes & outputHashes) {
-            res = outputHashes;
-        },
-    }, hashDerivationModulo(store, drv, true));
-    return res;
 }
 
 
@@ -616,7 +709,8 @@ StringSet BasicDerivation::outputNames() const
     return names;
 }
 
-DerivationOutputsAndOptPaths BasicDerivation::outputsAndOptPaths(const Store & store) const {
+DerivationOutputsAndOptPaths BasicDerivation::outputsAndOptPaths(const Store & store) const
+{
     DerivationOutputsAndOptPaths outsAndOptPaths;
     for (auto output : outputs)
         outsAndOptPaths.insert(std::make_pair(
@@ -627,7 +721,8 @@ DerivationOutputsAndOptPaths BasicDerivation::outputsAndOptPaths(const Store & s
     return outsAndOptPaths;
 }
 
-std::string_view BasicDerivation::nameFromPath(const StorePath & drvPath) {
+std::string_view BasicDerivation::nameFromPath(const StorePath & drvPath)
+{
     auto nameWithSuffix = drvPath.name();
     constexpr std::string_view extension = ".drv";
     assert(hasSuffix(nameWithSuffix, extension));
@@ -669,27 +764,32 @@ void writeDerivation(Sink & out, const Store & store, const BasicDerivation & dr
     for (auto & i : drv.outputs) {
         out << i.first;
         std::visit(overloaded {
-            [&](const DerivationOutputInputAddressed & doi) {
+            [&](const DerivationOutput::InputAddressed & doi) {
                 out << store.printStorePath(doi.path)
                     << ""
                     << "";
             },
-            [&](const DerivationOutputCAFixed & dof) {
+            [&](const DerivationOutput::CAFixed & dof) {
                 out << store.printStorePath(dof.path(store, drv.name, i.first))
                     << dof.hash.printMethodAlgo()
                     << dof.hash.hash.to_string(Base16, false);
             },
-            [&](const DerivationOutputCAFloating & dof) {
+            [&](const DerivationOutput::CAFloating & dof) {
                 out << ""
                     << (makeFileIngestionPrefix(dof.method) + printHashType(dof.hashType))
                     << "";
             },
-            [&](const DerivationOutputDeferred &) {
+            [&](const DerivationOutput::Deferred &) {
                 out << ""
                     << ""
                     << "";
             },
-        }, i.second.output);
+            [&](const DerivationOutput::Impure & doi) {
+                out << ""
+                    << (makeFileIngestionPrefix(doi.method) + printHashType(doi.hashType))
+                    << "impure";
+            },
+        }, i.second.raw());
     }
     worker_proto::write(store, out, drv.inputSrcs);
     out << drv.platform << drv.builder << drv.args;
@@ -714,21 +814,19 @@ std::string downstreamPlaceholder(const Store & store, const StorePath & drvPath
 }
 
 
-static void rewriteDerivation(Store & store, BasicDerivation & drv, const StringMap & rewrites) {
+static void rewriteDerivation(Store & store, BasicDerivation & drv, const StringMap & rewrites)
-
+{
-    debug("Rewriting the derivation");
+    for (auto & rewrite : rewrites) {
-
-    for (auto &rewrite: rewrites) {
         debug("rewriting %s as %s", rewrite.first, rewrite.second);
     }
 
     drv.builder = rewriteStrings(drv.builder, rewrites);
-    for (auto & arg: drv.args) {
+    for (auto & arg : drv.args) {
         arg = rewriteStrings(arg, rewrites);
     }
 
     StringPairs newEnv;
-    for (auto & envVar: drv.env) {
+    for (auto & envVar : drv.env) {
         auto envName = rewriteStrings(envVar.first, rewrites);
         auto envValue = rewriteStrings(envVar.second, rewrites);
         newEnv.emplace(envName, envValue);
@@ -737,43 +835,52 @@ static void rewriteDerivation(Store & store, BasicDerivation & drv, const String
 
     auto hashModulo = hashDerivationModulo(store, Derivation(drv), true);
     for (auto & [outputName, output] : drv.outputs) {
-        if (std::holds_alternative<DerivationOutputDeferred>(output.output)) {
+        if (std::holds_alternative<DerivationOutput::Deferred>(output.raw())) {
-            Hash h = std::get<Hash>(hashModulo);
+            auto & h = hashModulo.hashes.at(outputName);
             auto outPath = store.makeOutputPath(outputName, h, drv.name);
             drv.env[outputName] = store.printStorePath(outPath);
-            output = DerivationOutput {
+            output = DerivationOutput::InputAddressed {
-                .output = DerivationOutputInputAddressed {
+                .path = std::move(outPath),
-                    .path = std::move(outPath),
-                },
             };
         }
     }
 
 }
 
-std::optional<BasicDerivation> Derivation::tryResolve(Store & store) {
+std::optional<BasicDerivation> Derivation::tryResolve(Store & store) const
+{
+    std::map<std::pair<StorePath, std::string>, StorePath> inputDrvOutputs;
+
+    for (auto & input : inputDrvs)
+        for (auto & [outputName, outputPath] : store.queryPartialDerivationOutputMap(input.first))
+            if (outputPath)
+                inputDrvOutputs.insert_or_assign({input.first, outputName}, *outputPath);
+
+    return tryResolve(store, inputDrvOutputs);
+}
+
+std::optional<BasicDerivation> Derivation::tryResolve(
+    Store & store,
+    const std::map<std::pair<StorePath, std::string>, StorePath> & inputDrvOutputs) const
+{
     BasicDerivation resolved { *this };
 
     // Input paths that we'll want to rewrite in the derivation
     StringMap inputRewrites;
 
-    for (auto & input : inputDrvs) {
+    for (auto & [inputDrv, inputOutputs] : inputDrvs) {
-        auto inputDrvOutputs = store.queryPartialDerivationOutputMap(input.first);
+        for (auto & outputName : inputOutputs) {
-        StringSet newOutputNames;
+            if (auto actualPath = get(inputDrvOutputs, { inputDrv, outputName })) {
-        for (auto & outputName : input.second) {
+                inputRewrites.emplace(
-            auto actualPathOpt = inputDrvOutputs.at(outputName);
+                    downstreamPlaceholder(store, inputDrv, outputName),
-            if (!actualPathOpt) {
+                    store.printStorePath(*actualPath));
-                warn("output %s of input %s missing, aborting the resolving",
+                resolved.inputSrcs.insert(*actualPath);
+            } else {
+                warn("output '%s' of input '%s' missing, aborting the resolving",
                     outputName,
-                    store.printStorePath(input.first)
+                    store.printStorePath(inputDrv));
-                );
+                return {};
-                return std::nullopt;
             }
-            auto actualPath = *actualPathOpt;
-            inputRewrites.emplace(
-                downstreamPlaceholder(store, input.first, outputName),
-                store.printStorePath(actualPath));
-            resolved.inputSrcs.insert(std::move(actualPath));
         }
     }
 
@@ -782,4 +889,6 @@ std::optional<BasicDerivation> Derivation::tryResolve(Store & store) {
     return resolved;
 }
 
+const Hash impureOutputHash = hashString(htSHA256, "impure");
+
 }

src/libstore/derivations.hh

@@ -4,6 +4,7 @@
 #include "types.hh"
 #include "hash.hh"
 #include "content-address.hh"
+#include "repair-flag.hh"
 #include "sync.hh"
 
 #include <map>
@@ -40,23 +41,47 @@ struct DerivationOutputCAFloating
 };
 
 /* Input-addressed output which depends on a (CA) derivation whose hash isn't
- * known atm
+ * known yet.
  */
 struct DerivationOutputDeferred {};
 
-struct DerivationOutput
+/* Impure output which is moved to a content-addressed location (like
+   CAFloating) but isn't registered as a realization.
+ */
+struct DerivationOutputImpure
 {
-    std::variant<
+    /* information used for expected hash computation */
-        DerivationOutputInputAddressed,
+    FileIngestionMethod method;
-        DerivationOutputCAFixed,
+    HashType hashType;
-        DerivationOutputCAFloating,
+};
-        DerivationOutputDeferred
+
-    > output;
+typedef std::variant<
+    DerivationOutputInputAddressed,
+    DerivationOutputCAFixed,
+    DerivationOutputCAFloating,
+    DerivationOutputDeferred,
+    DerivationOutputImpure
+> _DerivationOutputRaw;
+
+struct DerivationOutput : _DerivationOutputRaw
+{
+    using Raw = _DerivationOutputRaw;
+    using Raw::Raw;
+
+    using InputAddressed = DerivationOutputInputAddressed;
+    using CAFixed = DerivationOutputCAFixed;
+    using CAFloating = DerivationOutputCAFloating;
+    using Deferred = DerivationOutputDeferred;
+    using Impure = DerivationOutputImpure;
 
     /* Note, when you use this function you should make sure that you're passing
        the right derivation name. When in doubt, you should use the safer
        interface provided by BasicDerivation::outputsAndOptPaths */
     std::optional<StorePath> path(const Store & store, std::string_view drvName, std::string_view outputName) const;
+
+    inline const Raw & raw() const {
+        return static_cast<const Raw &>(*this);
+    }
 };
 
 typedef std::map<std::string, DerivationOutput> DerivationOutputs;
@@ -72,30 +97,62 @@ typedef std::map<std::string, std::pair<DerivationOutput, std::optional<StorePat
    output IDs we are interested in. */
 typedef std::map<StorePath, StringSet> DerivationInputs;
 
-enum struct DerivationType : uint8_t {
+struct DerivationType_InputAddressed {
-    InputAddressed,
+    bool deferred;
-    DeferredInputAddressed,
-    CAFixed,
-    CAFloating,
 };
 
-/* Do the outputs of the derivation have paths calculated from their content,
+struct DerivationType_ContentAddressed {
-   or from the derivation itself? */
+    bool sandboxed;
-bool derivationIsCA(DerivationType);
+    bool fixed;
+};
 
-/* Is the content of the outputs fixed a-priori via a hash? Never true for
+struct DerivationType_Impure {
-   non-CA derivations. */
+};
-bool derivationIsFixed(DerivationType);
 
-/* Is the derivation impure and needs to access non-deterministic resources, or
+typedef std::variant<
-   pure and can be sandboxed? Note that whether or not we actually sandbox the
+    DerivationType_InputAddressed,
-   derivation is controlled separately. Never true for non-CA derivations. */
+    DerivationType_ContentAddressed,
-bool derivationIsImpure(DerivationType);
+    DerivationType_Impure
+> _DerivationTypeRaw;
 
-/* Does the derivation knows its own output paths?
+struct DerivationType : _DerivationTypeRaw {
- * Only true when there's no floating-ca derivation involved in the closure.
+    using Raw = _DerivationTypeRaw;
- */
+    using Raw::Raw;
-bool derivationHasKnownOutputPaths(DerivationType);
+    using InputAddressed = DerivationType_InputAddressed;
+    using ContentAddressed = DerivationType_ContentAddressed;
+    using Impure = DerivationType_Impure;
+
+    /* Do the outputs of the derivation have paths calculated from their content,
+       or from the derivation itself? */
+    bool isCA() const;
+
+    /* Is the content of the outputs fixed a-priori via a hash? Never true for
+       non-CA derivations. */
+    bool isFixed() const;
+
+    /* Whether the derivation is fully sandboxed. If false, the
+       sandbox is opened up, e.g. the derivation has access to the
+       network. Note that whether or not we actually sandbox the
+       derivation is controlled separately. Always true for non-CA
+       derivations. */
+    bool isSandboxed() const;
+
+    /* Whether the derivation is expected to produce the same result
+       every time, and therefore it only needs to be built once. This
+       is only false for derivations that have the attribute '__impure
+       = true'. */
+    bool isPure() const;
+
+    /* Does the derivation knows its own output paths?
+       Only true when there's no floating-ca derivation involved in the
+       closure, or if fixed output.
+     */
+    bool hasKnownOutputPaths() const;
+
+    inline const Raw & raw() const {
+        return static_cast<const Raw &>(*this);
+    }
+};
 
 struct BasicDerivation
 {
@@ -140,7 +197,14 @@ struct Derivation : BasicDerivation
           added directly to input sources.
 
        2. Input placeholders are replaced with realized input store paths. */
-    std::optional<BasicDerivation> tryResolve(Store & store);
+    std::optional<BasicDerivation> tryResolve(Store & store) const;
+
+    /* Like the above, but instead of querying the Nix database for
+       realisations, uses a given mapping from input derivation paths
+       + output names to actual output store paths. */
+    std::optional<BasicDerivation> tryResolve(
+        Store & store,
+        const std::map<std::pair<StorePath, std::string>, StorePath> & inputDrvOutputs) const;
 
     Derivation() = default;
     Derivation(const BasicDerivation & bd) : BasicDerivation(bd) { }
@@ -150,8 +214,6 @@ struct Derivation : BasicDerivation
 
 class Store;
 
-enum RepairFlag : bool { NoRepair = false, Repair = true };
-
 /* Write a derivation to the Nix store, and return its path. */
 StorePath writeDerivation(Store & store,
     const Derivation & drv,
@@ -171,17 +233,27 @@ bool isDerivation(const std::string & fileName);
    the output name is "out". */
 std::string outputPathName(std::string_view drvName, std::string_view outputName);
 
-// known CA drv's output hashes, current just for fixed-output derivations
-// whose output hashes are always known since they are fixed up-front.
-typedef std::map<std::string, Hash> CaOutputHashes;
 
-struct DeferredHash { Hash hash; };
+// The hashes modulo of a derivation.
+//
+// Each output is given a hash, although in practice only the content-addressed
+// derivations (fixed-output or not) will have a different hash for each
+// output.
+struct DrvHash {
+    std::map<std::string, Hash> hashes;
 
-typedef std::variant<
+    enum struct Kind : bool {
-    Hash, // regular DRV normalized hash
+        // Statically determined derivations.
-    CaOutputHashes, // Fixed-output derivation hashes
+        // This hash will be directly used to compute the output paths
-    DeferredHash // Deferred hashes for floating outputs drvs and their dependencies
+        Regular,
-> DrvHashModulo;
+        // Floating-output derivations (and their reverse dependencies).
+        Deferred,
+    };
+
+    Kind kind;
+};
+
+void operator |= (DrvHash::Kind & self, const DrvHash::Kind & other) noexcept;
 
 /* Returns hashes with the details of fixed-output subderivations
    expunged.
@@ -206,16 +278,18 @@ typedef std::variant<
    ATerm, after subderivations have been likewise expunged from that
    derivation.
  */
-DrvHashModulo hashDerivationModulo(Store & store, const Derivation & drv, bool maskOutputs);
+DrvHash hashDerivationModulo(Store & store, const Derivation & drv, bool maskOutputs);
 
 /*
    Return a map associating each output to a hash that uniquely identifies its
    derivation (modulo the self-references).
+
+   FIXME: what is the Hash in this map?
  */
-std::map<std::string, Hash> staticOutputHashes(Store& store, const Derivation& drv);
+std::map<std::string, Hash> staticOutputHashes(Store & store, const Derivation & drv);
 
 /* Memoisation of hashDerivationModulo(). */
-typedef std::map<StorePath, DrvHashModulo> DrvHashes;
+typedef std::map<StorePath, DrvHash> DrvHashes;
 
 // FIXME: global, though at least thread-safe.
 extern Sync<DrvHashes> drvHashes;
@@ -245,4 +319,6 @@ std::string hashPlaceholder(const std::string_view outputName);
    dependency which is a CA derivation. */
 std::string downstreamPlaceholder(const Store & store, const StorePath & drvPath, std::string_view outputName);
 
+extern const Hash impureOutputHash;
+
 }

src/libstore/derived-path.cc

@@ -1,4 +1,5 @@
 #include "derived-path.hh"
+#include "derivations.hh"
 #include "store-api.hh"
 
 #include <nlohmann/json.hpp>
@@ -11,6 +12,21 @@ nlohmann::json DerivedPath::Opaque::toJSON(ref<Store> store) const {
     return res;
 }
 
+nlohmann::json DerivedPath::Built::toJSON(ref<Store> store) const {
+    nlohmann::json res;
+    res["drvPath"] = store->printStorePath(drvPath);
+    // Fallback for the input-addressed derivation case: We expect to always be
+    // able to print the output paths, so let’s do it
+    auto knownOutputs = store->queryPartialDerivationOutputMap(drvPath);
+    for (const auto& output : outputs) {
+        if (knownOutputs.at(output))
+            res["outputs"][output] = store->printStorePath(knownOutputs.at(output).value());
+        else
+            res["outputs"][output] = nullptr;
+    }
+    return res;
+}
+
 nlohmann::json BuiltPath::Built::toJSON(ref<Store> store) const {
     nlohmann::json res;
     res["drvPath"] = store->printStorePath(drvPath);
@@ -35,16 +51,22 @@ StorePathSet BuiltPath::outPaths() const
     );
 }
 
-nlohmann::json derivedPathsWithHintsToJSON(const BuiltPaths & buildables, ref<Store> store) {
+template<typename T>
+nlohmann::json stuffToJSON(const std::vector<T> & ts, ref<Store> store) {
     auto res = nlohmann::json::array();
-    for (const BuiltPath & buildable : buildables) {
+    for (const T & t : ts) {
-        std::visit([&res, store](const auto & buildable) {
+        std::visit([&res, store](const auto & t) {
-            res.push_back(buildable.toJSON(store));
+            res.push_back(t.toJSON(store));
-        }, buildable.raw());
+        }, t.raw());
     }
     return res;
 }
 
+nlohmann::json derivedPathsWithHintsToJSON(const BuiltPaths & buildables, ref<Store> store)
+{ return stuffToJSON<BuiltPath>(buildables, store); }
+nlohmann::json derivedPathsToJSON(const DerivedPaths & paths, ref<Store> store)
+{ return stuffToJSON<DerivedPath>(paths, store); }
+
 
 std::string DerivedPath::Opaque::to_string(const Store & store) const {
     return store.printStorePath(path);

src/libstore/derived-path.hh

@@ -25,6 +25,9 @@ struct DerivedPathOpaque {
     nlohmann::json toJSON(ref<Store> store) const;
     std::string to_string(const Store & store) const;
     static DerivedPathOpaque parse(const Store & store, std::string_view);
+
+    bool operator < (const DerivedPathOpaque & b) const
+    { return path < b.path; }
 };
 
 /**
@@ -45,6 +48,10 @@ struct DerivedPathBuilt {
 
     std::string to_string(const Store & store) const;
     static DerivedPathBuilt parse(const Store & store, std::string_view);
+    nlohmann::json toJSON(ref<Store> store) const;
+
+    bool operator < (const DerivedPathBuilt & b) const
+    { return std::make_pair(drvPath, outputs) < std::make_pair(b.drvPath, b.outputs); }
 };
 
 using _DerivedPathRaw = std::variant<
@@ -119,5 +126,6 @@ typedef std::vector<DerivedPath> DerivedPaths;
 typedef std::vector<BuiltPath> BuiltPaths;
 
 nlohmann::json derivedPathsWithHintsToJSON(const BuiltPaths & buildables, ref<Store> store);
+nlohmann::json derivedPathsToJSON(const DerivedPaths & , ref<Store> store);
 
 }

src/libstore/filetransfer.cc

@@ -443,14 +443,13 @@ struct curlFileTransfer : public FileTransfer
                     : httpStatus != 0
                     ? FileTransferError(err,
                         std::move(response),
-                        fmt("unable to %s '%s': HTTP error %d ('%s')",
+                        "unable to %s '%s': HTTP error %d%s",
-                            request.verb(), request.uri, httpStatus, statusMsg)
+                        request.verb(), request.uri, httpStatus,
-                        + (code == CURLE_OK ? "" : fmt(" (curl error: %s)", curl_easy_strerror(code)))
+                        code == CURLE_OK ? "" : fmt(" (curl error: %s)", curl_easy_strerror(code)))
-                        )
                     : FileTransferError(err,
                         std::move(response),
-                        fmt("unable to %s '%s': %s (%d)",
+                        "unable to %s '%s': %s (%d)",
-                            request.verb(), request.uri, curl_easy_strerror(code), code));
+                        request.verb(), request.uri, curl_easy_strerror(code), code);
 
                 /* If this is a transient error, then maybe retry the
                    download after a while. If we're writing to a
@@ -704,7 +703,7 @@ struct curlFileTransfer : public FileTransfer
                 auto s3Res = s3Helper.getObject(bucketName, key);
                 FileTransferResult res;
                 if (!s3Res.data)
-                    throw FileTransferError(NotFound, {}, "S3 object '%s' does not exist", request.uri);
+                    throw FileTransferError(NotFound, "S3 object '%s' does not exist", request.uri);
                 res.data = std::move(*s3Res.data);
                 callback(std::move(res));
 #else

src/libstore/filetransfer.hh

@@ -31,7 +31,7 @@ struct FileTransferSettings : Config
         R"(
           The timeout (in seconds) for establishing connections in the
           binary cache substituter. It corresponds to `curl`’s
-          `--connect-timeout` option.
+          `--connect-timeout` option. A value of 0 means no limit.
         )"};
 
     Setting<unsigned long> stalledDownloadTimeout{
@@ -123,8 +123,6 @@ public:
 
     template<typename... Args>
     FileTransferError(FileTransfer::Error error, std::optional<std::string> response, const Args & ... args);
-
-    virtual const char* sname() const override { return "FileTransferError"; }
 };
 
 bool isUri(std::string_view s);

src/libstore/gc-store.cc

@@ -1,13 +0,0 @@
-#include "gc-store.hh"
-
-namespace nix {
-
-GcStore & requireGcStore(Store & store)
-{
-    auto * gcStore = dynamic_cast<GcStore *>(&store);
-    if (!gcStore)
-        throw UsageError("Garbage collection not supported by this store");
-    return *gcStore;
-}
-
-}

src/libstore/gc-store.hh

@@ -61,6 +61,8 @@ struct GCResults
 
 struct GcStore : public virtual Store
 {
+    inline static std::string operationName = "Garbage collection";
+
     /* Add an indirect root, which is merely a symlink to `path' from
        /nix/var/nix/gcroots/auto/<hash of `path'>.  `path' is supposed
        to be a symlink to a store path.  The garbage collector will
@@ -79,6 +81,4 @@ struct GcStore : public virtual Store
     virtual void collectGarbage(const GCOptions & options, GCResults & results) = 0;
 };
 
-GcStore & requireGcStore(Store & store);
-
 }

src/libstore/gc.cc

@@ -678,7 +678,8 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results)
                 alive.insert(start);
                 try {
                     StorePathSet closure;
-                    computeFSClosure(*path, closure);
+                    computeFSClosure(*path, closure,
+                        /* flipDirection */ false, gcKeepOutputs, gcKeepDerivations);
                     for (auto & p : closure)
                         alive.insert(p);
                 } catch (InvalidPath &) { }
@@ -841,7 +842,8 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results)
             if (unlink(path.c_str()) == -1)
                 throw SysError("deleting '%1%'", path);
 
-            results.bytesFreed += st.st_size;
+            /* Do not accound for deleted file here. Rely on deletePath()
+               accounting.  */
         }
 
         struct stat st;

src/libstore/local-fs-store.hh

@@ -2,6 +2,7 @@
 
 #include "store-api.hh"
 #include "gc-store.hh"
+#include "log-store.hh"
 
 namespace nix {
 
@@ -24,7 +25,10 @@ struct LocalFSStoreConfig : virtual StoreConfig
         "physical path to the Nix store"};
 };
 
-class LocalFSStore : public virtual LocalFSStoreConfig, public virtual Store, virtual GcStore
+class LocalFSStore : public virtual LocalFSStoreConfig,
+    public virtual Store,
+    public virtual GcStore,
+    public virtual LogStore
 {
 public:
 

src/libstore/local-store.cc

@@ -81,7 +81,7 @@ int getSchema(Path schemaPath)
 
 void migrateCASchema(SQLite& db, Path schemaPath, AutoCloseFD& lockFd)
 {
-    const int nixCASchemaVersion = 3;
+    const int nixCASchemaVersion = 4;
     int curCASchema = getSchema(schemaPath);
     if (curCASchema != nixCASchemaVersion) {
         if (curCASchema > nixCASchemaVersion) {
@@ -143,6 +143,21 @@ void migrateCASchema(SQLite& db, Path schemaPath, AutoCloseFD& lockFd)
             )");
             txn.commit();
         }
+        if (curCASchema < 4) {
+            SQLiteTxn txn(db);
+            db.exec(R"(
+                create trigger if not exists DeleteSelfRefsViaRealisations before delete on ValidPaths
+                begin
+                    delete from RealisationsRefs where realisationReference in (
+                    select id from Realisations where outputPath = old.id
+                    );
+                end;
+                -- used by deletion trigger
+                create index if not exists IndexRealisationsRefsRealisationReference on RealisationsRefs(realisationReference);
+            )");
+            txn.commit();
+        }
+
         writeFile(schemaPath, fmt("%d", nixCASchemaVersion));
         lockFile(lockFd.get(), ltRead, true);
     }
@@ -482,18 +497,18 @@ void LocalStore::openDB(State & state, bool create)
         SQLiteStmt stmt;
         stmt.create(db, "pragma main.journal_mode;");
         if (sqlite3_step(stmt) != SQLITE_ROW)
-            throwSQLiteError(db, "querying journal mode");
+            SQLiteError::throw_(db, "querying journal mode");
         prevMode = std::string((const char *) sqlite3_column_text(stmt, 0));
     }
     if (prevMode != mode &&
         sqlite3_exec(db, ("pragma main.journal_mode = " + mode + ";").c_str(), 0, 0, 0) != SQLITE_OK)
-        throwSQLiteError(db, "setting journal mode");
+        SQLiteError::throw_(db, "setting journal mode");
 
     /* Increase the auto-checkpoint interval to 40000 pages.  This
        seems enough to ensure that instantiating the NixOS system
        derivation is done in a single fsync(). */
     if (mode == "wal" && sqlite3_exec(db, "pragma wal_autocheckpoint = 40000;", 0, 0, 0) != SQLITE_OK)
-        throwSQLiteError(db, "setting autocheckpoint interval");
+        SQLiteError::throw_(db, "setting autocheckpoint interval");
 
     /* Initialise the database schema, if necessary. */
     if (create) {
@@ -695,31 +710,34 @@ void LocalStore::checkDerivationOutputs(const StorePath & drvPath, const Derivat
     // combinations that are currently prohibited.
     drv.type();
 
-    std::optional<Hash> h;
+    std::optional<DrvHash> hashesModulo;
     for (auto & i : drv.outputs) {
         std::visit(overloaded {
-            [&](const DerivationOutputInputAddressed & doia) {
+            [&](const DerivationOutput::InputAddressed & doia) {
-                if (!h) {
+                if (!hashesModulo) {
                     // somewhat expensive so we do lazily
-                    auto temp = hashDerivationModulo(*this, drv, true);
+                    hashesModulo = hashDerivationModulo(*this, drv, true);
-                    h = std::get<Hash>(temp);
                 }
-                StorePath recomputed = makeOutputPath(i.first, *h, drvName);
+                StorePath recomputed = makeOutputPath(i.first, hashesModulo->hashes.at(i.first), drvName);
                 if (doia.path != recomputed)
                     throw Error("derivation '%s' has incorrect output '%s', should be '%s'",
                         printStorePath(drvPath), printStorePath(doia.path), printStorePath(recomputed));
                 envHasRightPath(doia.path, i.first);
             },
-            [&](const DerivationOutputCAFixed & dof) {
+            [&](const DerivationOutput::CAFixed & dof) {
                 StorePath path = makeFixedOutputPath(dof.hash.method, dof.hash.hash, drvName);
                 envHasRightPath(path, i.first);
             },
-            [&](const DerivationOutputCAFloating &) {
+            [&](const DerivationOutput::CAFloating &) {
                 /* Nothing to check */
             },
-            [&](const DerivationOutputDeferred &) {
+            [&](const DerivationOutput::Deferred &) {
+                /* Nothing to check */
             },
-        }, i.second.output);
+            [&](const DerivationOutput::Impure &) {
+                /* Nothing to check */
+            },
+        }, i.second.raw());
     }
 }
 

src/libstore/log-store.hh

@@ -0,0 +1,21 @@
+#pragma once
+
+#include "store-api.hh"
+
+
+namespace nix {
+
+struct LogStore : public virtual Store
+{
+    inline static std::string operationName = "Build log storage and retrieval";
+
+    /* Return the build log of the specified store path, if available,
+       or null otherwise. */
+    virtual std::optional<std::string> getBuildLog(const StorePath & path) = 0;
+
+    virtual void addBuildLog(const StorePath & path, std::string_view log) = 0;
+
+    static LogStore & require(Store & store);
+};
+
+}

src/libstore/make-content-addressed.cc

@@ -0,0 +1,80 @@
+#include "make-content-addressed.hh"
+#include "references.hh"
+
+namespace nix {
+
+std::map<StorePath, StorePath> makeContentAddressed(
+    Store & srcStore,
+    Store & dstStore,
+    const StorePathSet & storePaths)
+{
+    StorePathSet closure;
+    srcStore.computeFSClosure(storePaths, closure);
+
+    auto paths = srcStore.topoSortPaths(closure);
+
+    std::reverse(paths.begin(), paths.end());
+
+    std::map<StorePath, StorePath> remappings;
+
+    for (auto & path : paths) {
+        auto pathS = srcStore.printStorePath(path);
+        auto oldInfo = srcStore.queryPathInfo(path);
+        std::string oldHashPart(path.hashPart());
+
+        StringSink sink;
+        srcStore.narFromPath(path, sink);
+
+        StringMap rewrites;
+
+        StorePathSet references;
+        bool hasSelfReference = false;
+        for (auto & ref : oldInfo->references) {
+            if (ref == path)
+                hasSelfReference = true;
+            else {
+                auto i = remappings.find(ref);
+                auto replacement = i != remappings.end() ? i->second : ref;
+                // FIXME: warn about unremapped paths?
+                if (replacement != ref)
+                    rewrites.insert_or_assign(srcStore.printStorePath(ref), srcStore.printStorePath(replacement));
+                references.insert(std::move(replacement));
+            }
+        }
+
+        sink.s = rewriteStrings(sink.s, rewrites);
+
+        HashModuloSink hashModuloSink(htSHA256, oldHashPart);
+        hashModuloSink(sink.s);
+
+        auto narModuloHash = hashModuloSink.finish().first;
+
+        auto dstPath = dstStore.makeFixedOutputPath(
+            FileIngestionMethod::Recursive, narModuloHash, path.name(), references, hasSelfReference);
+
+        printInfo("rewriting '%s' to '%s'", pathS, srcStore.printStorePath(dstPath));
+
+        StringSink sink2;
+        RewritingSink rsink2(oldHashPart, std::string(dstPath.hashPart()), sink2);
+        rsink2(sink.s);
+        rsink2.flush();
+
+        ValidPathInfo info { dstPath, hashString(htSHA256, sink2.s) };
+        info.references = std::move(references);
+        if (hasSelfReference) info.references.insert(info.path);
+        info.narSize = sink.s.size();
+        info.ca = FixedOutputHash {
+            .method = FileIngestionMethod::Recursive,
+            .hash = narModuloHash,
+        };
+
+        StringSource source(sink2.s);
+        dstStore.addToStore(info, source);
+
+        remappings.insert_or_assign(std::move(path), std::move(info.path));
+    }
+
+    return remappings;
+}
+
+}

src/libstore/make-content-addressed.hh

@@ -0,0 +1,12 @@
+#pragma once
+
+#include "store-api.hh"
+
+namespace nix {
+
+std::map<StorePath, StorePath> makeContentAddressed(
+    Store & srcStore,
+    Store & dstStore,
+    const StorePathSet & storePaths);
+
+}

src/libstore/misc.cc

@@ -87,7 +87,7 @@ std::optional<ContentAddress> getDerivationCA(const BasicDerivation & drv)
 {
     auto out = drv.outputs.find("out");
     if (out != drv.outputs.end()) {
-        if (auto v = std::get_if<DerivationOutputCAFixed>(&out->second.output))
+        if (const auto * v = std::get_if<DerivationOutput::CAFixed>(&out->second.raw()))
             return v->hash;
     }
     return std::nullopt;
@@ -277,15 +277,15 @@ std::map<DrvOutput, StorePath> drvOutputReferences(
 {
     std::set<Realisation> inputRealisations;
 
-    for (const auto& [inputDrv, outputNames] : drv.inputDrvs) {
+    for (const auto & [inputDrv, outputNames] : drv.inputDrvs) {
         auto outputHashes =
             staticOutputHashes(store, store.readDerivation(inputDrv));
-        for (const auto& outputName : outputNames) {
+        for (const auto & outputName : outputNames) {
             auto thisRealisation = store.queryRealisation(
                 DrvOutput{outputHashes.at(outputName), outputName});
             if (!thisRealisation)
                 throw Error(
-                    "output '%s' of derivation '%s' isn’t built", outputName,
+                    "output '%s' of derivation '%s' isn't built", outputName,
                     store.printStorePath(inputDrv));
             inputRealisations.insert(*thisRealisation);
         }
@@ -295,4 +295,5 @@ std::map<DrvOutput, StorePath> drvOutputReferences(
 
     return drvOutputReferences(Realisation::closure(store, inputRealisations), info->references);
 }
+
 }

src/libstore/parsed-derivations.cc

@@ -93,7 +93,7 @@ StringSet ParsedDerivation::getRequiredSystemFeatures() const
     StringSet res;
     for (auto & i : getStringsAttr("requiredSystemFeatures").value_or(Strings()))
         res.insert(i);
-    if (!derivationHasKnownOutputPaths(drv.type()))
+    if (!drv.type().hasKnownOutputPaths())
         res.insert("ca-derivations");
     return res;
 }

src/libstore/parsed-derivations.hh

@@ -1,5 +1,6 @@
 #pragma once
 
+#include "derivations.hh"
 #include "store-api.hh"
 
 #include <nlohmann/json_fwd.hpp>

src/libstore/path.cc

@@ -1,5 +1,7 @@
 #include "store-api.hh"
 
+#include <sodium.h>
+
 namespace nix {
 
 static void checkName(std::string_view path, std::string_view name)
@@ -41,6 +43,13 @@ bool StorePath::isDerivation() const
 
 StorePath StorePath::dummy("ffffffffffffffffffffffffffffffff-x");
 
+StorePath StorePath::random(std::string_view name)
+{
+    Hash hash(htSHA1);
+    randombytes_buf(hash.hash, hash.hashSize);
+    return StorePath(hash, name);
+}
+
 StorePath Store::parseStorePath(std::string_view path) const
 {
     auto p = canonPath(std::string(path));

src/libstore/path.hh

@@ -58,6 +58,8 @@ public:
     }
 
     static StorePath dummy;
+
+    static StorePath random(std::string_view name);
 };
 
 typedef std::set<StorePath> StorePathSet;

src/libstore/remote-store.hh

@@ -5,6 +5,7 @@
 
 #include "store-api.hh"
 #include "gc-store.hh"
+#include "log-store.hh"
 
 
 namespace nix {
@@ -30,7 +31,10 @@ struct RemoteStoreConfig : virtual StoreConfig
 
 /* FIXME: RemoteStore is a misnomer - should be something like
    DaemonStore. */
-class RemoteStore : public virtual RemoteStoreConfig, public virtual Store, public virtual GcStore
+class RemoteStore : public virtual RemoteStoreConfig,
+    public virtual Store,
+    public virtual GcStore,
+    public virtual LogStore
 {
 public:
 

src/libstore/repair-flag.hh

@@ -0,0 +1,7 @@
+#pragma once
+
+namespace nix {
+
+enum RepairFlag : bool { NoRepair = false, Repair = true };
+
+}

src/libstore/sqlite.cc

@@ -8,22 +8,32 @@
 
 namespace nix {
 
-[[noreturn]] void throwSQLiteError(sqlite3 * db, const FormatOrString & fs)
+SQLiteError::SQLiteError(const char *path, int errNo, int extendedErrNo, hintformat && hf)
+  : Error(""), path(path), errNo(errNo), extendedErrNo(extendedErrNo)
+{
+    err.msg = hintfmt("%s: %s (in '%s')",
+        normaltxt(hf.str()),
+        sqlite3_errstr(extendedErrNo),
+        path ? path : "(in-memory)");
+}
+
+[[noreturn]] void SQLiteError::throw_(sqlite3 * db, hintformat && hf)
 {
     int err = sqlite3_errcode(db);
     int exterr = sqlite3_extended_errcode(db);
 
     auto path = sqlite3_db_filename(db, nullptr);
-    if (!path) path = "(in-memory)";
 
     if (err == SQLITE_BUSY || err == SQLITE_PROTOCOL) {
-        throw SQLiteBusy(
+        auto exp = SQLiteBusy(path, err, exterr, std::move(hf));
+        exp.err.msg = hintfmt(
             err == SQLITE_PROTOCOL
-            ? fmt("SQLite database '%s' is busy (SQLITE_PROTOCOL)", path)
+                ? "SQLite database '%s' is busy (SQLITE_PROTOCOL)"
-            : fmt("SQLite database '%s' is busy", path));
+                : "SQLite database '%s' is busy",
-    }
+            path ? path : "(in-memory)");
-    else
+        throw exp;
-        throw SQLiteError("%s: %s (in '%s')", fs.s, sqlite3_errstr(exterr), path);
+    } else
+        throw SQLiteError(path, err, exterr, std::move(hf));
 }
 
 SQLite::SQLite(const Path & path, bool create)
@@ -37,7 +47,7 @@ SQLite::SQLite(const Path & path, bool create)
         throw Error("cannot open SQLite database '%s'", path);
 
     if (sqlite3_busy_timeout(db, 60 * 60 * 1000) != SQLITE_OK)
-        throwSQLiteError(db, "setting timeout");
+        SQLiteError::throw_(db, "setting timeout");
 
     exec("pragma foreign_keys = 1");
 }
@@ -46,7 +56,7 @@ SQLite::~SQLite()
 {
     try {
         if (db && sqlite3_close(db) != SQLITE_OK)
-            throwSQLiteError(db, "closing database");
+            SQLiteError::throw_(db, "closing database");
     } catch (...) {
         ignoreException();
     }
@@ -62,7 +72,7 @@ void SQLite::exec(const std::string & stmt)
 {
     retrySQLite<void>([&]() {
         if (sqlite3_exec(db, stmt.c_str(), 0, 0, 0) != SQLITE_OK)
-            throwSQLiteError(db, format("executing SQLite statement '%s'") % stmt);
+            SQLiteError::throw_(db, "executing SQLite statement '%s'", stmt);
     });
 }
 
@@ -76,7 +86,7 @@ void SQLiteStmt::create(sqlite3 * db, const std::string & sql)
     checkInterrupt();
     assert(!stmt);
     if (sqlite3_prepare_v2(db, sql.c_str(), -1, &stmt, 0) != SQLITE_OK)
-        throwSQLiteError(db, fmt("creating statement '%s'", sql));
+        SQLiteError::throw_(db, "creating statement '%s'", sql);
     this->db = db;
     this->sql = sql;
 }
@@ -85,7 +95,7 @@ SQLiteStmt::~SQLiteStmt()
 {
     try {
         if (stmt && sqlite3_finalize(stmt) != SQLITE_OK)
-            throwSQLiteError(db, fmt("finalizing statement '%s'", sql));
+            SQLiteError::throw_(db, "finalizing statement '%s'", sql);
     } catch (...) {
         ignoreException();
     }
@@ -109,7 +119,7 @@ SQLiteStmt::Use & SQLiteStmt::Use::operator () (std::string_view value, bool not
 {
     if (notNull) {
         if (sqlite3_bind_text(stmt, curArg++, value.data(), -1, SQLITE_TRANSIENT) != SQLITE_OK)
-            throwSQLiteError(stmt.db, "binding argument");
+            SQLiteError::throw_(stmt.db, "binding argument");
     } else
         bind();
     return *this;
@@ -119,7 +129,7 @@ SQLiteStmt::Use & SQLiteStmt::Use::operator () (const unsigned char * data, size
 {
     if (notNull) {
         if (sqlite3_bind_blob(stmt, curArg++, data, len, SQLITE_TRANSIENT) != SQLITE_OK)
-            throwSQLiteError(stmt.db, "binding argument");
+            SQLiteError::throw_(stmt.db, "binding argument");
     } else
         bind();
     return *this;
@@ -129,7 +139,7 @@ SQLiteStmt::Use & SQLiteStmt::Use::operator () (int64_t value, bool notNull)
 {
     if (notNull) {
         if (sqlite3_bind_int64(stmt, curArg++, value) != SQLITE_OK)
-            throwSQLiteError(stmt.db, "binding argument");
+            SQLiteError::throw_(stmt.db, "binding argument");
     } else
         bind();
     return *this;
@@ -138,7 +148,7 @@ SQLiteStmt::Use & SQLiteStmt::Use::operator () (int64_t value, bool notNull)
 SQLiteStmt::Use & SQLiteStmt::Use::bind()
 {
     if (sqlite3_bind_null(stmt, curArg++) != SQLITE_OK)
-        throwSQLiteError(stmt.db, "binding argument");
+        SQLiteError::throw_(stmt.db, "binding argument");
     return *this;
 }
 
@@ -152,14 +162,14 @@ void SQLiteStmt::Use::exec()
     int r = step();
     assert(r != SQLITE_ROW);
     if (r != SQLITE_DONE)
-        throwSQLiteError(stmt.db, fmt("executing SQLite statement '%s'", sqlite3_expanded_sql(stmt.stmt)));
+        SQLiteError::throw_(stmt.db, fmt("executing SQLite statement '%s'", sqlite3_expanded_sql(stmt.stmt)));
 }
 
 bool SQLiteStmt::Use::next()
 {
     int r = step();
     if (r != SQLITE_DONE && r != SQLITE_ROW)
-        throwSQLiteError(stmt.db, fmt("executing SQLite query '%s'", sqlite3_expanded_sql(stmt.stmt)));
+        SQLiteError::throw_(stmt.db, fmt("executing SQLite query '%s'", sqlite3_expanded_sql(stmt.stmt)));
     return r == SQLITE_ROW;
 }
 
@@ -185,14 +195,14 @@ SQLiteTxn::SQLiteTxn(sqlite3 * db)
 {
     this->db = db;
     if (sqlite3_exec(db, "begin;", 0, 0, 0) != SQLITE_OK)
-        throwSQLiteError(db, "starting transaction");
+        SQLiteError::throw_(db, "starting transaction");
     active = true;
 }
 
 void SQLiteTxn::commit()
 {
     if (sqlite3_exec(db, "commit;", 0, 0, 0) != SQLITE_OK)
-        throwSQLiteError(db, "committing transaction");
+        SQLiteError::throw_(db, "committing transaction");
     active = false;
 }
 
@@ -200,7 +210,7 @@ SQLiteTxn::~SQLiteTxn()
 {
     try {
         if (active && sqlite3_exec(db, "rollback;", 0, 0, 0) != SQLITE_OK)
-            throwSQLiteError(db, "aborting transaction");
+            SQLiteError::throw_(db, "aborting transaction");
     } catch (...) {
         ignoreException();
     }
@@ -215,7 +225,6 @@ void handleSQLiteBusy(const SQLiteBusy & e)
     if (now > lastWarned + 10) {
         lastWarned = now;
         logWarning({
-            .name = "Sqlite busy",
             .msg = hintfmt(e.what())
         });
     }