With Linux kernel >=6.6 & glibc 2.39 a `fchmodat2(2)` is available that
isn't filtered away by the libseccomp sandbox.
Being able to use this to bypass that restriction has surprising results
for some builds such as lxc[1]:
> With kernel ≥6.6 and glibc 2.39, lxc's install phase uses fchmodat2,
> which slips through 9b88e52846/src/libstore/build/local-derivation-goal.cc (L1650-L1663).
> The fixupPhase then uses fchmodat, which fails.
> With older kernel or glibc, setting the suid bit fails in the
> install phase, which is not treated as fatal, and then the
> fixup phase does not try to set it again.
Please note that there are still ways to bypass this sandbox[2] and this is
mostly a fix for the breaking builds.
This change works by creating a syscall filter for the `fchmodat2`
syscall (number 452 on most systems). The problem is that glibc 2.39
is needed to have the correct syscall number available via
`__NR_fchmodat2` / `__SNR_fchmodat2`, but this flake is still on
nixpkgs 23.11. To have this change everywhere and not dependent on the
glibc this package is built against, I added a header
"fchmodat2-compat.hh" that sets the syscall number based on the
architecture. On most platforms its 452 according to glibc with a few
exceptions:
$ rg --pcre2 'define __NR_fchmodat2 (?!452)'
sysdeps/unix/sysv/linux/x86_64/x32/arch-syscall.h
58:#define __NR_fchmodat2 1073742276
sysdeps/unix/sysv/linux/mips/mips64/n32/arch-syscall.h
67:#define __NR_fchmodat2 6452
sysdeps/unix/sysv/linux/mips/mips64/n64/arch-syscall.h
62:#define __NR_fchmodat2 5452
sysdeps/unix/sysv/linux/mips/mips32/arch-syscall.h
70:#define __NR_fchmodat2 4452
sysdeps/unix/sysv/linux/alpha/arch-syscall.h
59:#define __NR_fchmodat2 562
I added a small regression-test to the setuid integration-test that
attempts to set the suid bit on a file using the fchmodat2 syscall.
I confirmed that the test fails without the change in
local-derivation-goal.
Additionally, we require libseccomp 2.5.5 or greater now: as it turns
out, libseccomp maintains an internal syscall table and
validates each rule against it. This means that when using libseccomp
2.5.4 or older, one may pass `452` as syscall number against it, but
since it doesn't exist in the internal structure, `libseccomp` will refuse
to create a filter for that. This happens with nixpkgs-23.11, i.e. on
stable NixOS and when building Lix against the project's flake.
To work around that
* a backport of libseccomp 2.5.5 on upstream nixpkgs has been
scheduled[3].
* the package now uses libseccomp 2.5.5 on its own already. This is to
provide a quick fix since the correct fix for 23.11 is still a staging cycle
away.
We still need the compat header though since `SCMP_SYS(fchmodat2)`
internally transforms this into `__SNR_fchmodat2` which points to
`__NR_fchmodat2` from glibc 2.39, so it wouldn't build on glibc 2.38.
The updated syscall table from libseccomp 2.5.5 is NOT used for that
step, but used later, so we need both, our compat header and their
syscall table 🤷
Relevant PRs in CppNix:
* https://github.com/NixOS/nix/pull/10591
* https://github.com/NixOS/nix/pull/10501
[1] https://github.com/NixOS/nixpkgs/issues/300635#issuecomment-2031073804
[2] https://github.com/NixOS/nixpkgs/issues/300635#issuecomment-2030844251
[3] https://github.com/NixOS/nixpkgs/pull/306070
(cherry picked from commit ba6804518772e6afb403dd55478365d4b863c854)
Change-Id: I6921ab5a363188c6bff617750d00bb517276b7fe
Part of #7672
My main motivation is to be able to use `nix.checkConfig`[1]. This
doesn't work with Lix currently since the module uses `nix show-config`
if the Nix version is <2.20pre and `nix config show` otherwise. I think
this is the only instance where nixpkgs checks for which Nix commands
exist that affects us now, so I figured we could just perform the rename
here as well[2] and still provide the current version number[3].
I don't have a strong opinion on whether to deprecate `nix show-config`,
the warning is added there automatically.
(cherry picked from commit f300e11b056dea414d7d77bbc6e5a7dc5d9ddd41)
[1] https://nixos.org/manual/nixos/stable/options.html#opt-nix.checkConfig
[2] I should add that I don't use the "official" ways of installing Lix
because using the flake directly and callPackaging it seemed to fit
better into my workflow: I already have a little mess to make
sure Hydra from the flake uses the correct pkgs.nix and I didn't
want to complicate it further while keeping a single package-set I
can build in CI. Don't get me wrong, I think such a module for a
quick-start is very important, just giving context on why I bother
in the first place :)
[3] When we go public, I think it's worth considering to add support in
nixpkgs itself for Lix.
Change-Id: I47b4239b05cbeda3c370d2fa56ea768b768768ac
* changes:
docs: clarify how ^ works for -E/-f installables
docs: give translation examples from nix-build -E/-A to installables
docs: clarify how the different kinds of installables are selected
docs: guide to installables docs in installable commands' docs
Currently LocalDerivationGoal allows setting `__sandboxProfile`
to add sandbox parameters on Darwin when `sandbox=true`.
This was only supposed to have an effect when `sandbox=relaxed`
Change-Id: Ide44ee82d7e4d6b545285eab26547e7014817d3f
this is used in CA rewriting, replacement of placeholders in
derivations, generating scripts for devShells, and some more
places. in all of these transitive replacements are unsound,
and overlapping replacements would be as well. there even is
a test that transitive replacements do not happen (in the CA
RewriteSink suite), but none for overlapping replacements. a
minimally surprising binary rewriter surely would not do any
of these replacements, the only reason we have not seen this
break yet is probably that rewriteStrings is only called for
store paths and things that look like store paths (and those
should never overlap nor admit such transitive replacements)
Change-Id: I6fc29f939d5061d9f56c752624a823ece8437c07
* changes:
nix3-profile: remove check "name" attr in manifests
Add profile migration test
nix3-profile: make element names stable
getNameFromURL(): Support uppercase characters in attribute names
nix3-profile: remove indices
nix3-profile: allow using human-readable names to select packages
implement parsing human-readable names from URLs
As discussed in the maintainer meeting on 2024-01-29.
Mainly this is to avoid a situation where the name is parsed and
treated as a file name, mostly to protect users.
.-* and ..-* are also considered invalid because they might strip
on that separator to remove versions. Doesn't really work, but that's
what we decided, and I won't argue with it, because .-* probably
doesn't seem to have a real world application anyway.
We do still permit a 1-character name that's just "-", which still
poses a similar risk in such a situation. We can't start disallowing
trailing -, because a non-zero number of users will need it and we've
seen how annoying and painful such a change is.
What matters most is preventing a situation where . or .. can be
injected, and to just get this done.
(cherry picked from commit f1b4663805a9dbcb1ace64ec110092d17c9155e0)
Change-Id: I900a8509933cee662f888c3c76fa8986b0058839
We didn't even realize you *could* use this syntax with -E and -f, much
less that the attribute path could be *empty*.
Change-Id: Id1a6715609f3a76a5ce477bd43a7832effbbe07b
The installables syntax is not documented in any of the man pages or
docbook pages for any of those individual commands. And while these
commands really should at least peripherally individually document how
installables work, in the meantime we can at least direct people to the
right place.
This commit also clarifies the unexpected fact that `nix profile remove`
and `nix profile upgrade` do *not* take installables.
Change-Id: I3b1453cb197a613bbab639c66a466365c3592c6d
nix3-profile automatically migrates any profile its used on to its style
of profile -- the ones with manifest.json instead of manifest.nix. On
non-NixOS systems, Nix is conventionally installed to the profile at
/nix/var/nix/profiles/default, so if a user passed that to `--profile`
of `nix profile`, then it would break upgrade-nix from ever working
again, without recreating the profile.
This commit fixes that, and allows upgrade-nix to work on either kind of
profile.
Fixes#16.
Change-Id: I4c49b1beba93bb50e8f8a107edc451affe08c3f7
Notably, ProfileManifest and ProfileElement are useful generic
profile management code, and nix profile is not the only place in the
codebase where profiles are relevant.
This commit is in preparation for fixing upgrade-nix's interaction with
new-style profiles.
Change-Id: Iefc8bbd34b4bc6012175cb3d6e6a8207973bc792
This replaces the external sandbox-exec call with direct calls into
libsandbox. This API is technically deprecated and is missing some
prototypes, but all major browsers depend on it, so it is unlikely to
materially change without warning.
This commit also ensures the netrc file is only written if the
derivation is in fact meant to be able to access the internet.
This change commits a sin of not actually actively declaring its
dependency on macOS's libsandbox.dylib; this is due to the dylib
cache in macOS making that explicit dependency unnecessary. In the
future this might become a problem, so this commit marks our sins.
Co-authored-by: Artemis Tosini <lix@artem.ist>
Co-authored-by: Lunaphied <lunaphied@lunaphied.me>
Change-Id: Ia302141a53ce7b0327c1aad86a117b6645fe1189
That's expected by `build-remote` and makes sure that errors are
correctly forwarded to the user. For instance, let's say that the
host-key of `example.org` is unknown and
nix-build ../nixpkgs -A hello -j0 --builders 'ssh-ng://example.org'
is issued, then you get the following output:
cannot build on 'ssh-ng://example.org?&': error: failed to start SSH connection to 'example.org'
Failed to find a machine for remote build!
derivation: yh46gakxq3kchrbihwxvpn5bmadcw90b-hello-2.12.1.drv
required (system, features): (x86_64-linux, [])
2 available machines:
[...]
The relevant information (`Host key verification failed`) ends up in the
daemon's log, but that's not very obvious considering that the daemon
isn't very chatty normally.
This can be fixed - the same way as its done for legacy-ssh - by passing
fd 4 to the SSH wrapper. Now you'd get the following error:
cannot build on 'ssh-ng://example.org': error: failed to start SSH connection to 'example.org': Host key verification failed.
Failed to find a machine for remote build!
[...]
...and now it's clear what's wrong.
Please note that this is won't end up in the derivation's log.
For previous discussion about this change see
https://github.com/NixOS/nix/pull/7659.
Change-Id: I5790856dbf58e53ea3e63238b015ea06c347cf92
only decompress the response once all data has been received (in the
fully buffered case), or at least outside of the curl wrapper itself
(in the receive-to-sink case). unfortunately this means we will have
to duplicate decompression logic for these two cases for time being,
but once the curl wrapper has been rewritten to return a real future
or Source we can deduplicate this logic again. the curl wrapper will
have to turn into a proper Source first and use decompression source
logic which also does not currently exist—only decompression *sinks*
Change-Id: I66bc692f07d9b9e69fe10689ee73a2de8d65e35c
this is highly questionable. single-arg download calls will misbehave
with it set, and two-arg download calls will just overwrite it. being
an implementation detail this should not have been in the API at all.
Change-Id: I613772951ee03d8302366085f06a53601d13f132
this lets each implementation of FileTransfer (of which currently only
the one exists at all) implement appropriate handling for its internal
behaviours that are not otherwise exposed. in curl this lets us switch
the buffer-full handling method from "block the entire curl thread" to
"pause just the one transfer", move the non-libcurl body decompression
out of the actual curl wrapper (which will let us eventually morph the
curl wrapper intto an actual source of Sources), and some other things
Change-Id: Id6d3593cde6b4915aab3e90a43b175c103cc3f18
Previously, the garbage collector found runtime roots on Darwin by
shelling out to `lsof -n -w -F n` then parsing the result.
However, this requires an lsof binary and can be extremely slow.
The official Apple lsof returns in a reasonable amount of time,
about 250ms in my tests, but the lsof packaged in nixpkgs is quite slow,
taking about 40 seconds to run the command.
Using libproc directly is about the same speed as Apple lsof,
and allows us to reënable several tests that were disabled on Darwin.
Change-Id: Ifa0adda7984e13c15535693baba835aae79a3577
just accumulate error data into result.data as we would for successful
transfers without a dataCallback. errorSink and data would contain the
same data in error cases anyway, so splitting them is not very useful.
Change-Id: I00e449866454389ac6a564ab411c903fd357dabf
This creates new subclasses of LocalStore for each OS to include
platform-specific functionality. Currently this just includes garbage
collector roots but it could be extended to sandboxing as well.
In order to make sure that the generic LocalStore is not accidentally
constructed, its constructor is protected. A Fallback is provided which
implements no functionality except constructors.
Change-Id: I836a28e90b68309873f75afb83e0f1b2e2c89fb3
This should fix cross compilation in the base case, but this is
difficult to test as cross compilation is broken in many different
places right now. This should bring Meson back up to cross parity with
the Make buildsystem though.
Change-Id: If09be8142d1fc975a82b994143ff35be1297dad8
don't reimplement header parsing. this was only really needed due to the
ancient github bug we no longer care about, everything else we have done
in custom code can also be done using curl itself. doing this also fixes
possible sources of header smuggling (because the header function didn't
unfold headers and we'd trim them before parsing, which would've made us
read contents of one header as a fully formed header in itself). this is
a slight behavior change because we now honor only the first instance of
a given header where previous behavior was to honor either the last or a
combination of all of them (accept-ranges was logical-or'd by accident).
Change-Id: I93cb93ddb91ab98c8991f846014926f6ef039fdb
this was a workaround for a *github* bug that happend *in 2015*.
not only is github no longer buggy, it shouldn't have been nix's
responsibility to work around these bugs like this to begin with
while we're at it we'll also remove another workaround—again for
github specifically and again for etag handling—from 2021 that's
also not needed any more. future workarounds for serverside bugs
should probably come with an expiration date that mutates into a
build warning after a while, otherwise this *will* happen again.
Change-Id: I74f739ae3e36d40350f78bebcb5869aa8cc9adcd
the previous solution to the wakeup problem (adding a pipe and passing
it as an additional fd to curl_multi_wait) worked, but there have been
builtin alternatives for this since 2020. not only do these save code,
they're also a lot more likely to work natively on windows when needed
Change-Id: Iab751b900997110a8d15de45ea3ab0c42f7e5973
the oldest version checked for here is 7.47, which was released in
2016. it's probably safe to say that we do not need these any more
Change-Id: I003411f6b2ce6d56f7ca337390df3ea86bd59a99
With Nix 2.3, it was possible to pass a subpath of a store path to
exportReferencesGraph:
with import <nixpkgs> {};
let
hello = writeShellScriptBin "hello" ''
echo ${toString builtins.currentTime}
'';
in
writeClosure [ "${hello}/bin/hello" ]
This regressed with Nix 2.4, with a very confusing error message, that
presumably indicates it was unintentional:
error: path '/nix/store/3gl7kgjr4pwf03f0x70dgx9ln3bhl7zc-hello/bin/hello' is not in the Nix store
(cherry picked from commit 0774e8ba33c060f56bad3ff696796028249e915a)
Change-Id: I00920fb33077b831a1bb4a1b68d515ba8c3c2a69
The statically embedded busybox is not required for Lix to work, but
package.nix explicitly sets this, which was accidentally being ignored.
Change-Id: Ieeff830ac7d1f5fabe84d1a6cfd82f13d79035bf
Either the contents of `line` could cause format errors, or this usage
is Technically safe. However, I trust nothing, especially with
boost::format.
Change-Id: I07933b20bde3b305a6e5d61c2a7bab6ecb042ad9
Previously if isStorePath() was called on anything other than a
top-level /nix/store/some-path, it would throw a BadStorePath exception.
This commit duplicates the absolutely trivial check, into
maybeParseStorePath(), and leaves exception throwing to
parseStorePath(), the function that assumes you're already giving a
valid path instead of the one whose purpose is to check if its valid or
not...
Change-Id: I8dda548f0f88d14ca8c3ee927d64e0ec0681fc7b
Saves us a bunch of thinking about how to handle symlinks, and prevents
the DNS config from changing on the fly under the build, which may or may
not be a good thing?
Change-Id: I071e6ae7e220884690b788d94f480866f428db71
* changes:
meson: fix log-dir
manual: build docs with dummy envs
libcmd: install generated headers as well
docs: redo content generation for mdbook and manual
manpages can be rendered using the markdown output of mdbook, the rest
of the manual can generated out of the main doc/manual source tree. we
still use lowdown to actually render manpages instead of eg mdbook-man
because lowdown does generate reasonably good manpages (though that is
also somewhat debatable, but they're a lot better than mdbook-man).
doing this not only lets us drastically simplify the lowdown pipeline,
but also remove all custom {{#include}} handling since now mdbook does
all of it, even for the manpage builds. even the lowdown wrapper isn't
entirely necessary because lowdown can take all wrapper arguments with
command line flags rather than bits of input file content.
This also implements running mdbook in Meson, in order to generate the
manpages. The mdbook outputs are also installed in the usual location.
Co-authored-by: Qyriad <qyriad@qyriad.me>
Change-Id: I60193f9fd0f15d48872f071af35855cda2a0f40b
this should be a link, not an anchor. it should also point to the
`gloss-store` element, not the `#gloss-store` element.
Change-Id: I1f2803093179549637e10f917ad73399a419131b
Previously, errors while printing values in `nix repl` would be printed
in `«error: ...»` brackets rather than displayed normally:
```
nix-repl> legacyPackages.aarch64-darwin.pythonPackages.APScheduler
«error: Package ‘python-2.7.18.7’ in /nix/store/6s0m1qc31zw3l3kq0q4wd5cp3lqpkq0q-source/pkgs/development/interpreters/python/cpython/2.7/default.nix:335 is marked as insecure, refusing to evaluate.»
```
Now, errors will be displayed normally if they're emitted at the
top-level of an expression:
```
nix-repl> legacyPackages.aarch64-darwin.pythonPackages.APScheduler
error:
… in the condition of the assert statement
at /nix/store/6s0m1qc31zw3l3kq0q4wd5cp3lqpkq0q-source/lib/customisation.nix:268:17:
267| in commonAttrs // {
268| drvPath = assert condition; drv.drvPath;
| ^
269| outPath = assert condition; drv.outPath;
… in the left operand of the OR (||) operator
at /nix/store/6s0m1qc31zw3l3kq0q4wd5cp3lqpkq0q-source/pkgs/development/interpreters/python/passthrufun.nix:28:45:
27| if lib.isDerivation value then
28| lib.extendDerivation (valid value || throw "${name} should use `buildPythonPackage` or `toPythonModule` if it is to be part of the Python packages set.") {} value
| ^
29| else
(stack trace truncated; use '--show-trace' to show the full trace)
error: Package ‘python-2.7.18.7’ in /nix/store/6s0m1qc31zw3l3kq0q4wd5cp3lqpkq0q-source/pkgs/development/interpreters/python/cpython/2.7/default.nix:335 is marked as insecure, refusing to evaluate.
```
Errors emitted in nested structures (like e.g. when printing `nixpkgs`)
will still be printed in brackets.
Change-Id: I25aeddf08c017582718cb9772a677bf51b9fc2ad
Instead of $sysconfdir.
Fixes#231, but there's more to do in following commits to make
Meson-built Lix actually look in /etc/nix.
Change-Id: Ia8d627070f405843add46e05cff5134b76b8eb48
Adds a `repl-overlays` option, which specifies files that can overlay
and modify the top-level bindings in `nix repl`. For example, with the
following contents in `~/.config/nix/repl.nix`:
info: final: prev: let
optionalAttrs = predicate: attrs:
if predicate
then attrs
else {};
in
optionalAttrs (prev ? legacyPackages && prev.legacyPackages ? ${info.currentSystem})
{
pkgs = prev.legacyPackages.${info.currentSystem};
}
We can run `nix repl` and use `pkgs` to refer to `legacyPackages.${currentSystem}`:
$ nix repl --repl-overlays ~/.config/nix/repl.nix nixpkgs
Lix 2.90.0
Type :? for help.
Loading installable 'flake:nixpkgs#'...
Added 5 variables.
Loading 'repl-overlays'...
Added 6 variables.
nix-repl> pkgs.bash
«derivation /nix/store/g08b5vkwwh0j8ic9rkmd8mpj878rk62z-bash-5.2p26.drv»
Change-Id: Ic12e0f2f210b2f46e920c33088dfe1083f42391a
This probably snuck in in a refactor using truthiness or so. The
trustedness flag was having the optional fullness checked, rather than
the actual contained trust level.
Also adds some tests.
```
m1@6876551b-255d-4cb0-af02-8a4f17b27e2e ~ % nix store ping
warning: 'nix store ping' is a deprecated alias for 'nix store info'
Store URL: daemon
Version: 2.20.4
Trusted: 0
m1@6876551b-255d-4cb0-af02-8a4f17b27e2e ~ % nix doctor
warning: 'doctor' is a deprecated alias for 'config check'
[PASS] PATH contains only one nix version.
[PASS] All profiles are gcroots.
[PASS] Client protocol matches store protocol.
[INFO] You are trusted by store uri: daemon
```
Fixes: lix-project/lix#232
Change-Id: I21576e2a0a755036edf8814133345987617ba3d0
package.nix previously needed this callPackage'd externally, which
didn't make a lot of sense to us since this is an internal dependency.
Thus we changed it to make it more self contained.
Change-Id: I4935bc0bc80e1a132bc9b1519e917791da95037c
code blocks, if not surrounded by empty lines, have the language
tags (in these cases, always `nix`) show up in the output of :doc.
for example:
nix-repl> :doc builtins.parseFlakeRef
Synopsis: builtins.parseFlakeRef flake-ref
Parse a flake reference, and return its exploded form.
For example: nix builtins.parseFlakeRef
"github:NixOS/nixpkgs/23.05?dir=lib" evaluates to: nix { dir =
"lib"; owner = "NixOS"; ref = "23.05"; repo = "nixpkgs"; type =
"github"; }
is now instead:
nix-repl> :doc builtins.parseFlakeRef
Synopsis: builtins.parseFlakeRef flake-ref
Parse a flake reference, and return its exploded form.
For example:
| builtins.parseFlakeRef "github:NixOS/nixpkgs/23.05?dir=lib"
evaluates to:
| { dir = "lib"; owner = "NixOS"; ref = "23.05"; repo = "nixpkgs"; type = "github"; }
(closes#225)
Change-Id: I0741aeb1006a5376bb2f663d202c7a4da7e38cce
This is terrible UX, and frankly an eval failure should be a cache
invalidation anyway.
This removes the CachedEvalError type entirely.
Fixes#223.
Change-Id: I91f8003eabd0ea45003024e96d1de3c7ae8e49d8
This reverts commit 491caad6f6.
this is not actually legal for nix! throwing exceptions in destructors
is fine, but the way nix is set up we'll end up throwing the exception
we received from the remote *twice* in some cases, and such cases will
cause an immediate terminate without active exception.
Change-Id: I74c46b9f26fd791086e4193ec60eb1deb9a5bb2a
throwing exceptions is fine, but throwing exceptions during exception
handling is hard enough to do correctly that we should just forbid it
entirely out of an overabundance of caution. in cases where terminate
is the correct answer the users of Finally must call it manually now.
Change-Id: Ia51a2cb4a0638500550bfabc89cf01a6d8098983
For a long time `nix repl` has supported displaying documentation set on
builtins, however, it has long been convention to use Markdown comments
on Nix functions themselves for documentation. This exposes that
information to `nix repl` users in a nice and formatted way.
NixOS/rfcs#145 doc-comments are primarily what this feature is intended
to consume, however, support for lambda documentation in the repl is
experimental. We do our best effort to support the RFC here.
These changes are based on [the nix-doc library](https://github.com/lf-/nix-doc) and
are licensed under the terms described in the relevant source files.
Change-Id: Ic6fe947d39a22540705d890737e336c4720b0a22
setting this only on exceptions caused by actual fd access is not
sufficient to diagnose all errors (such as SerialisationError) in
some cases. this usually does not have any negative effects since
those errors will end up killing the process in another way. this
is not a reliable assumption though and we should be using proper
error handling (and closing connections more often, preferring to
close over keeping something open that might be in a weird state)
Change-Id: I1b792cd7ad8ba9ff0f6bd174945ab2575ff2208e
not needed yet, but returning a resource from the exception handling
path that has ownership of a handle is currently not well-supported.
we could also add a default constructor to Handle, but then we would
also need to change the pool reference to a pointer. eventually that
should be done since now resources can be swapped between pools with
clever moves, but since that's not a problem yet we won't do it now.
Change-Id: I26eb06581f7be34569e9e67a33da736128d167af
the duplication of exception handling was added without justification,
so we can only assume that it was done like this because Finally could
not throw exceptions safely. since this has now been rectified we will
deduplicate this handler code again.
Change-Id: I40721f3378c0fd9f34e2914a16d383f6e2713b40
this is supposed to act like a finally block does in other languages. a
finally block should be able to throw exceptions of its own rather than
just crashing the entire program when it throws it own exceptions. even
in the rare case of a finally throwing an unexpected exception it might
be better to report the exception from Finally instead of the original,
at least that can keep our program running instead of letting it crash.
Change-Id: Id42011e46b1df369152b4564938c0e93fa1acf32
usage of this flag previously kept connections open much longer than
necessary, and at the same time obscured that a connection was being
dropped when it *was* set. new variable names clarify this somewhat.
Change-Id: I11f6f08f37a5e4dc04ea6c6036ea589154b121c6
it was used incorrectly (not swapped on handle move), only used in one
place (that is now handled with exception handling detection in Handle
itself), and if ever reintroduced should be replaced with a different,
more understandable mechanism (like an explicit dropAsInvalid method).
Change-Id: Ie3e5d5cfa81d335429cb2ee5c3ad85c74a9df17b
this was never actually used, and bad design in the first place—why
should a bad resource be put back into the idle pool? just drop it.
Change-Id: Idab8774bee19dadae0209d404c4fb86dd4aeba1e
if a scope owning a resource does not gracefully drop that resource
while handling exceptions from deeper down the call stack we should
assume the resource is invalid state and drop it. currently it *is*
true that such cases do not cause resources to be freed, but thanks
to validator misuses this has so far not caused any larger problem.
Change-Id: Ie4f91bcd60a64d05c5ff9d22cc97954816d13b97
The big ones here are `trim-trailing-whitespace` and `end-of-file-fixer`
(which makes sure that every file ends with exactly one newline
character).
Change-Id: Idca73b640883188f068f9903e013cf0d82aa1123
This does involve making a large number of destructors able to throw,
because we had to change it high in the class hierarchy. Oh well.
Change-Id: Ib62d3d6895b755f20322bb8acc9bf43daf0174b2
* some things that can throw are marked noexcept
yet the linter seems to think not. Maybe they can't throw in practice.
I would rather not have the UB possibility in pretty obvious cold
paths.
* various default-case-missing complaints
* a fair pile of casts from integer to character, which are in fact
deliberate.
* an instance of <https://clang.llvm.org/extra/clang-tidy/checks/bugprone/move-forwarding-reference.html>
* bugprone-not-null-terminated-result on handing a string to curl in
chunks of bytes. our usage is fine.
* reassigning a unique_ptr by CRIMES instead of using release(), then
using release() and ignoring the result. wild. let's use release() for
its intended purpose.
Change-Id: Ic3e7affef12383576213a8a7c8145c27e662513d
We got confused what formals did and had to briefly figure it out. We
should just have docs, so these are some.
Change-Id: If3e794a401e69d022785cbfa0b0c2e2284f41f58
the autoconf build system defaults to /nix/var, not /nix/var/nix. the
latter is only used in libstore, so we'll move the extra segment there.
Change-Id: Idfbc988ee302355982abdcd51d6d7b5d5d661c0d
Without this, the Meson setup won't bail out if nlohmann_json is
missing, leading to subpar DX (and maybe worse, but I'm not entirely
sure).
Change-Id: I5913111060226b540dcf003257c99a08e84da0de
one headers (args/root.hh) was simply missing, and the generated headers
were not installed. not all of them *should* be installed either, only a
select few (and sadly this needs a custom target for each one, it seems)
Change-Id: I37b25517895d0e5e521abc1202fa65624de57ed1
Functional tests can be run with
`meson test -C build --suite installcheck`.
Notably, functional tests must be run *after* running `meson install`
(Lix's derivation runs the installcheck suite in installCheckPhase so it
does this correctly), due to some quirks between Meson and the testing
system.
As far as I can tell the functional tests are meant to be run after
installing anyway, but unfortunately I can't transparently make
`meson test --suite installcheck` depend on the install targets.
The script that runs the functional tests, meson/run-test.py, checks
that `meson install` has happened and fails fast with a (hopefully)
helpful error message if any of the functional tests are run before
installing.
TODO: this change needs reflection in developer documentation
Change-Id: I8dcb5fdfc0b6cb17580973d24ad930abd57018f6
This was achieved by running maintainers/buildtime_report.sh on the
build directory of a meson build, then asking "why the heck is json
eating our build times", and strategically moving the json using bits
out of widely included headers.
It turns out that putting literally any metrics whatsoever into the
build had immediate and predictable results.
Results are 1382.5s frontend time -> 1175.4s frontend time, back end
time approximately invariant.
Related: lix-project/lix#159
Change-Id: I7edea95c8536203325c8bb4dae5f32d727a21b2d
Once this commit lands, we are even more visible in analytics FWIW.
Change-Id: Id7e0c162315d0f191edbea9cb5fb82ce363704b9
Signed-off-by: Raito Bezarius <raito@lix.systems>
HintFmt(string) invokes the HintFmt("%s", literal) constructor,
which is not what we want here. Add a constructor with a proper name
and call that.
Next step: rename all the other ones to HintFmt::literal(string).
Fixes: lix-project/lix#178
Change-Id: If52d2eb8864ceb8663e05992e9d1fffef573d6b8
Committing a lock file using markFileChanged() required the input to
be writable by the caller in the local filesystem (using the path
returned by getSourcePath()). putFile() abstracts over this.
(cherry picked from commit 95d657c8b3ae4282e24628ba7426edb90c8f3942)
Change-Id: Ie081c5d9eb4e923b229191c5e23ece85145557ff
As I complained in
https://github.com/NixOS/nix/pull/6784#issuecomment-1421777030 (a
comment on the wrong PR, sorry again!), #6693 introduced a second
completions mechanism to fix a bug. Having two completion mechanisms
isn't so nice.
As @thufschmitt also pointed out, it was a bummer to go from `FlakeRef`
to `std::string` when collecting flake refs. Now it is `FlakeRefs`
again.
The underlying issue that sought to work around was that completion of
arguments not at the end can still benefit from the information from
latter arguments.
To fix this better, we rip out that change and simply defer all
completion processing until after all the (regular, already-complete)
arguments have been passed.
In addition, I noticed the original completion logic used some global
variables. I do not like global variables, because even if they save
lines of code, they also obfuscate the architecture of the code.
I got rid of them moved them to a new `RootArgs` class, which now has
`parseCmdline` instead of `Args`. The idea is that we have many argument
parsers from subcommands and what-not, but only one root args that owns
the other per actual parsing invocation. The state that was global is
now part of the root args instead.
This did, admittedly, add a bunch of new code. And I do feel bad about
that. So I went and added a lot of API docs to try to at least make the
current state of things clear to the next person.
--
This is needed for RFC 134 (tracking issue #7868). It was very hard to
modularize `Installable` parsing when there were two completion
arguments. I wouldn't go as far as to say it is *easy* now, but at least
it is less hard (and the completions test finally passed).
Co-authored-by: Valentin Gagarin <valentin.gagarin@tweag.io>
Change-Id: If18cd5be78da4a70635e3fdcac6326dbfeea71a5
(cherry picked from commit 67eb37c1d0de28160cd25376e51d1ec1b1c8305b)
An attrPath prefix of "." indicates no need to try default attrPath prefixes. For example `nixpkgs#legacyPackages.x86_64-linux.ERROR` searches through
```
trying flake output attribute 'packages.x86_64-linux.legacyPackages.x86_64-linux.ERROR'
using cached attrset attribute ''
trying flake output attribute 'legacyPackages.x86_64-linux.legacyPackages.x86_64-linux.ERROR'
using cached attrset attribute 'legacyPackages.x86_64-linux'
trying flake output attribute 'legacyPackages.x86_64-linux.ERROR'
using cached attrset attribute 'legacyPackages.x86_64-linux'
```
And there is no way to specify that one does not want the automatic
search behavior. Now one can specify
`nixpkgs#.legacyPackages.x86_64-linux.ERROR` to only refer to the rooted
attribute path without any default injection of attribute search path or
system.
Change-Id: Iac1334e1470137b7ce11dcf845513810230638ec
(cherry picked from commit d4aed18883b361133607296fb6cd789c47427a38)
protocol versions are sent as u64. on the peer we read them as uint64,
check that the upper half is 0, and throw an exception if not. we then
read an arbitrary amount of data from the peer and dump it to the user
terminal. this is a little bit ridiculous, can never happen in correct
implementation, and is severly untested. let us just drop it entirely.
Change-Id: Ibd2f53a765341ed6439d40d9d1eac11e79c6b5e3
this is not needed and introduces a bunch of memset calls, making up for
3% of valgrind cycle estimation *alone*. real-world impact is a lot
lower on our test machine, but we suspect that less powerful machines
would see an impact from dropping this.
Change-Id: Iad10e9d556e64fdeb0bee0059a4e52520058d11e
If the state SQLite database is configured to use a write-ahead-log, it
creates WAL files in the state directory.
When the state SQLite database is closed by the `nix-daemon` after
builds, those files are removed.
When an unprivileged user would like to open _in read only_ that
database, they cannot do so because they would need to create those WAL
files and they do not have the permission to do so.
For this, SQLite offers a "persistent WAL" feature [1] to leave the WAL
files around, even after closing the database.
This CL enable the persistent WAL mode.
Fixes: https://github.com/NixOS/nix/issues/10300
[1]: https://www.sqlite.org/wal.html
Change-Id: Id8ae534d7d2290457af28782e5215222ae051fe5
Signed-off-by: Raito Bezarius <raito@lix.systems>
This commit adds several meson.build, which successfully build and
install Lix executables, libraries, and headers. Meson does not yet
build docs, Perl bindings, or run tests, which will be added in
following commits. As such, this commit does not remove the existing
build system, or make it the default, and also as such, this commit has
several FIXMEs and TODOs as notes for what should be done before the
existing autoconf + make buildsystem can be removed and Meson made the
default. This commit does not modify any source files.
A Meson-enabled build is also added as a Hydra job, and to
`nix flake check`.
Change-Id: I667c8685b13b7bab91e281053f807a11616ae3d4
this was mostly an inconvenience for error reporting, but fully broke
the debugger (because the debugger does *a lot* of eager position
resolution). copying the line offsets into a local and filling that
local when empty without also storing the calculated offsets back does
kind of ... not cache anything.
fixes lix-project/lix#165
Change-Id: Iccb0ba193ce2f15c832978daecf7b9bebbbe8585
within lix itself this problem is caught by the test suite. outside of
lix itself three cases can be had: either the problem is fully inside
lix libs, fully inside user code, or it exists at the boundary. the
first is caught by the test suite, the second isn't caught at all, and
the third is something lix should not be responsible for.
Change-Id: I95aa35d8cb6f0ef5816a2941c467bc0c15916063
* changes:
Release notes for builtins.nixVersion change
un-nixes ur lix, a little
issue importer: list issues that are *not* closed when finding existing issues
I didn't really go attack the docs because we need to pull a bunch of
PRs. I went looking for strings in the code that called lix nix.
Change-Id: I2138bb4dd239096bc530946b281db7f875195b39
add a reset() method to close the wrapped fd instead of assigning magic
constants. also make the from-fd constructor explicit so you can't
accidentally assign the *wrong* magic constant, or even an unrelated
integer that also just happens to be an fd by pure chance.
Change-Id: I51311b0f6e040240886b5103d39d1794a6acc325
static env association is from expr to its enclosing scope, but let
exprs set their association to their *inner* scope. this skips one level
of envs and will cause segfaults if the parent is a with expr.
fixes#145
Change-Id: I1d22146110f071ede21b4eed7ed34b5850ef2ef3
not doing this exposes the binding name order to the annoying
interference of parse order on symbol order, which wouldn't be so bad if
it didn't make the tests less reliable and, importantly, dependent on
linker behavior (due to primop initialization being done in static
initializer, and the order of static initializers being defined only
within a single translation unit).
fixes#143
Change-Id: I3cf417893fbcf19e9ad3ff8986deb7cbcf3ca511
we now keep not a table of all positions, but a table of all origins and
their sizes. position indices are now direct pointers into the virtual
concatenation of all parsed contents. this slightly reduces memory usage
and time spent in the parser, at the cost of not being able to report
positions if the total input size exceeds 4GiB. this limit is not unique
to nix though, rustc and clang also limit their input to 4GiB (although
at least clang refuses to process inputs that are larger, we will not).
this new 4GiB limit probably will not cause any problems for quite a
while, all of nixpkgs together is less than 100MiB in size and already
needs over 700MiB of memory and multiple seconds just to parse. 4GiB
worth of input will easily take multiple minutes and over 30GiB of
memory without even evaluating anything. if problems *do* arise we can
probably recover the old table-based system by adding some tracking to
Pos::Origin (or increasing the size of PosIdx outright), but for time
being this looks like more complexity than it's worth.
since we now need to read the entire input again to determine the
line/column of a position we'll make unsafeGetAttrPos slightly lazy:
mostly the set it returns is only used to determine the file of origin
of an attribute, not its exact location. the thunks do not add
measurable runtime overhead.
notably this change is necessary to allow changing the parser since
apparently nothing supports nix's very idiosyncratic line ending choice
of "anything goes", making it very hard to calculate line/column
positions in the parser (while byte offsets are very easy).
(cherry picked from commit 5d9fdab3de0ee17c71369ad05806b9ea06dfceda)
Change-Id: Ie0b2430cb120c09097afa8c0101884d94f4bbf34
this needs a string comparison because there seems to be no other way to
get that information out of bison. usually the location info is going to
be correct (pointing at a bad token), but since EOF isn't a token as
such it'll be wrong in that this case.
this hasn't shown up much so far because a single line ending *is* a
token, so any file formatted in the usual manner (ie, ending in a line
ending) would have its EOF position reported correctly.
(cherry picked from commit 855fd5a1bb781e4f722c1d757ba43e866d370132)
Change-Id: I120c56a962f4286b1ae3b71da7b71ce8ec3e0535
the parser treats a plain \r as a newline, error reports do not. this
can lead to interesting divergences if anything makes use of this
feature, with error reports pointing to wrong locations in the input (or
even outside the input altogether).
(cherry picked from commit 2be6b143289e5479cc4a2667bb84e879116c2447)
Change-Id: Ieb7f7655bac8cb0cf5734c60bd41723388f2973c
previously we reported the error at the beginning of the binding
block (for plain inherits) or the beginning of the attr list (for
inherit-from), effectively hiding where exactly the error happened.
this also carries over to runtime positions of attributes in sets as
reported by unsafeGetAttrPos. we're not worried about this changing
observable eval behavior because it *is* marked unsafe, and the new
behavior is much more useful.
(cherry picked from commit 1edd6fada53553b89847ac3981ac28025857ca02)
Change-Id: I2f50eb9f3dc3977db4eb3e3da96f1cb37ccd5174
we already normalize attr order to lexicographic, doing the same for
formals makes sense. doubly so because the order of formals would
otherwise depend on the context of the expression, which is not quite as
useful as one might expect.
(cherry picked from commit 4147ecfb1c51f3fe3b4adcbd4e753fd487dab645)
Change-Id: I3fd0dbdef3ac7447a3a03ff20bb514a0d0f23fb1
the parser modifies its inputs, which means that sharing them between
the error context reporting system and the parser itself can confuse the
reporting system. usually this led to early truncation of error context
reports which, while not dangerous, can be quite confusing.
(cherry picked from commit d384ecd553aa997270b79ee98d02f7cf7e1849e6)
Change-Id: I677646b5675b12b2faa787943646aa36dc6e6ee3
vfork confers a large performance advantage over fork, measured locally
at 16µs per vfork agains 90µs per fork. however nix *almost always*
follows a vfork up with an execve-family call, melting the performance
advantage from 6x to only 15%. in most of those cases it's doing things
that are undefined behavior (like manipulating the heap, or even
throwing exceptions and trashing the parent process stack).
most notably the one place that could benefit from the vfork performance
improvement is linux derivation sandbox setup—which doesn't use vfork.
Change-Id: I2037b7384d5a4ca24da219a569e1b1f39531410e
These now have equivalents in the standard lib in C++20. This change was
performed with a custom clang-tidy check which I will submit later.
Executed like so:
ninja -C build && run-clang-tidy -checks='-*,nix-*' -load=build/libnix-clang-tidy.so -p .. -fix ../tests | tee -a clang-tidy-result
Change-Id: I62679e315ff9e7ce72a40b91b79c3e9fc01b27e9
This builtin is only going to cause us problems because we are not Nix,
so let's just falsify being in the 2.18 series, since that is the
closest target that has any meaning.
In future we might want to have a better feature detection mechanism,
for when we actually add stuff to some builtin's attr set argument. But
builtins.nixVersion is just going to be hopelessly broken and it should
be stubbed out.
Fixes lix-project/lix#144
Change-Id: Id7390b32a29c6147f2977737d81846320de5d67e
diagnose attr duplication at the path the duplication was detected, not
at the path the current attribute wanted to place. doing the latter is
only correct if a leaf attribute was duplicated, not if an attrpath was
set to a non-attrset in one binding and a (potentially implied) attrset
in another binding.
fixes#124
Change-Id: Ic4aa9cc12a9874d4e7897c6f64408f10aa36fc82
It happens with some frequency that plugins that might be unimportant to
the evaluation at hand mismatch with the nix version, leading to
spurious load failures. Let's make these non fatal.
Change-Id: Iba10e951d171725ccf1a121bcd9be1e1d6ad69eb
using the total-attrs-printed and total-list-items-printed counters to
calculate how many attrs were elided only works properly if no nesting
is involved. once things do nest the global counter can exceed the size
of the currently printed object, leading to unsigned wrapping and great
overestimation of elided counts. counting locally in addition to global
counts fixes this.
these are functional tests because creating these objects requires the
evaluator to not be a huge amount of code, and we also want defaults to
be tested for cli usage.
fixes#14
Change-Id: Icb9a0cb21b2f4bacbc5e9dcdd8c0b9055b4088a7
This is definitely not a stable thing, but it does feel slightly crimes
to put it as an experimental feature. Shrug, up for bikeshedding.
Change-Id: I6ef176e3dee6fb1cac9c0a7a60d553a2c63ea728
This solves the problem of collections of boxed subclasses with virtual
dispatch, which should still be treated as values, since the
indirection is only there due to the virtual dispatch.
Change-Id: I368daedd3f31298e99c6e56a15606337a55494c6
it's no longer widely used and has a rather confusing meaning now that
inherit-from is handled very differently.
(cherry picked from commit 1cd87b7042d14aae1fafa47b1c28db4c5bd20de7)
Change-Id: I90bbebddf06762960d8ca4f621cf042ce8ae83f9
desugaring inherit-from to syntactic duplication of the source expr also
duplicates side effects of the source expr (such as trace calls) and
expensive computations (such as derivationStrict).
(cherry picked from commit cefd0302b55b3360dbca59cfcb4bf6a750d6cdcf)
Change-Id: Iff519f991adef2e51683ba2c552d37a3df7a179e
deduplication does not currently work fully, showing derivations
multiple times if they have different underlying values. this can happen
by selecting the same derivation twice for two different attributes of a
set, using inherit-from (which reduces to the previous), importing
nixpkgs twice, or any other number of things.
since users already have to deal with duplicates for this reason it
won't hurt to add *more* duplicates. the alternative would be to
deduplicate fully, which would drop derivations that are currently
returned and those pose a regression risk.
Change-Id: I64b397351237e10375d270f1bddecb71f62aa131
for plain inherits this is really just a stylistic choice, but for
inherit-from it actually fixes an exponential size increase problem
during expr printing (as may happen during assertion failure reporting,
on during duplicate attr detection in the parser)
(cherry picked from commit ecf8b12d60ad2929f9998666cf0966475b91e291)
Change-Id: Ie55f0cb01a37e766414c31f8d40f51c2c7d106b0
this also has the effect of sorting let bindings lexicographically
rather than by symbol creation order as was previously done, giving a
better canonicalization in the process.
(cherry picked from commit 6c08fba533ef31cad2bdc03ba72ecf58dc8ee5a0)
Change-Id: Ia887f629305645bb8a165fbbc0d32e620912595a
in place of inherited() — not quite useful yet since we don't
distinguish plain and inheritFrom attr kinds so far.
(cherry picked from commit 1f542adb3e18e7078e6a589182a53a47d971748a)
Change-Id: If948c9d43e875de18f213a73a06a36f7c335b536
Do not skip any stack frames when `--show-trace` is given
(cherry picked from commit 0b47783d0a879875d558f0b56e49584f25ceb2d0)
Change-Id: Ia0f18266dbcf97543110110c655c219c7a3e3270
Enter debugger on `builtins.trace` with an option
(cherry picked from commit 774e7ca5847ebc392eac2a124a8f12b24da4f65a)
Change-Id: If01e2110b3a128e639b05143227e365227d149f1
Pretty-print values in the REPL by printing each item in a list or
attrset on a separate line. When possible, single-item lists and
attrsets are printed on one line, as long as they don't contain a nested
list, attrset, or thunk.
Before:
```
{ attrs = { a = { b = { c = { }; }; }; }; list = [ 1 ]; list' = [ 1 2 3 ]; }
```
After:
```
{
attrs = {
a = {
b = {
c = { };
};
};
};
list = [ 1 ];
list' = [
1
2
3
];
}
```
(cherry picked from commit c0a15fb7d03dfb8f53bc6726c414bc88aa362592)
Change-Id: Ia2b41849165a5ddb63f7a8c272a2476b3e4292df
While preparing PRs like #9753, I've had to change error messages in
dozens of code paths. It would be nice if instead of
EvalError("expected 'boolean' but found '%1%'", showType(v))
we could write
TypeError(v, "boolean")
or similar. Then, changing the error message could be a mechanical
refactor with the compiler pointing out places the constructor needs to
be changed, rather than the error-prone process of grepping through the
codebase. Structured errors would also help prevent the "same" error
from having multiple slightly different messages, and could be a first
step towards error codes / an error index.
This PR reworks the exception infrastructure in `libexpr` to
support exception types with different constructor signatures than
`BaseError`. Actually refactoring the exceptions to use structured data
will come in a future PR (this one is big enough already, as it has to
touch every exception in `libexpr`).
The core design is in `eval-error.hh`. Generally, errors like this:
state.error("'%s' is not a string", getAttrPathStr())
.debugThrow<TypeError>()
are transformed like this:
state.error<TypeError>("'%s' is not a string", getAttrPathStr())
.debugThrow()
The type annotation has moved from `ErrorBuilder::debugThrow` to
`EvalState::error`.
(cherry picked from commit c6a89c1a1659b31694c0fbcd21d78a6dd521c732)
Change-Id: Iced91ba4e00ca9e801518071fb43798936cbd05a
Maximilian Bosch
Qyriad
Artemis Tosini
eldritch horrors
Robert Hensing
puck
Ilya K
Alyssa Ross
Rebecca Turner
jade
terru -
Lunaphied
Jade Lovelace
Winter Cute
raito