lix-project/lix
19
51
Fork 27
Code Issues 339 Code Review (Gerrit) Projects 1 Releases Wiki Activity

Increase of 7MB in the size of the Lix derivation on Linux #308

New issue
Open
opened 2024-05-12 13:53:14 +00:00 by raito · 19 comments
raito commented 2024-05-12 13:53:14 +00:00
Owner
Copy link

From https://github.com/NixOS/nixpkgs/pull/310194 we have the following size reports:

Darwin:
`aarch64-darwin`: `/nix/store/rb4s95dpl9s33zax58jp15dgg06a0m6h-lix-2.90-beta.1                        	 14.6M` (+1MB with Nix!)
`x86_64-darwin`: `/nix/store/awcgvimn965qz3qyiqrxpd0cfi381fhl-lix-2.90-beta.1                        	 14.9M` (+0.7MB with Nix!)

Linux:
`x86_64-linux`: `/nix/store/23paa6b9ijxq6lwxl5ad0bmly01rrj52-lix-2.90-beta.1                                	 18.7M` (+7MB with Nix)
`aarch64-linux`: `/nix/store/w4i6zfqh3i7kavn2hr4gyk645hbvfxz3-lix-2.90-beta.1                                 	 18.4M` (+8MB with Nix)

The Linux derivation size are fishy.

From https://github.com/NixOS/nixpkgs/pull/310194 we have the following size reports: ``` Darwin: `aarch64-darwin`: `/nix/store/rb4s95dpl9s33zax58jp15dgg06a0m6h-lix-2.90-beta.1 14.6M` (+1MB with Nix!) `x86_64-darwin`: `/nix/store/awcgvimn965qz3qyiqrxpd0cfi381fhl-lix-2.90-beta.1 14.9M` (+0.7MB with Nix!) Linux: `x86_64-linux`: `/nix/store/23paa6b9ijxq6lwxl5ad0bmly01rrj52-lix-2.90-beta.1 18.7M` (+7MB with Nix) `aarch64-linux`: `/nix/store/w4i6zfqh3i7kavn2hr4gyk645hbvfxz3-lix-2.90-beta.1 18.4M` (+8MB with Nix) ``` The Linux derivation size are fishy.
raito added the
Area/build-packaging
 label 2024-05-12 13:53:26 +00:00
delroth commented 2024-05-12 15:42:10 +00:00
Member
Copy link

I built the derivation with ninjaFlags = [ "-v" ]; and:

[10/456] g++ -Isrc/libutil/libnixutil.so.p -Isrc/libutil -I../src/libutil -I/nix/store/rh6sv1rpxavhhzprbaqpkqdm4sq39d19-boehm-gc-8.2.6-dev/include -I/nix/store/73ik8xsdqv0grj8z4rplfqhzb5ajd4s3-boost-1.81.0-dev/include -I/nix/store/lpsa98iwl9q7fb2hdiqhqcryyg2kvgnm-libcpuid-0.6.5/include/libcpuid -I/nix/store/j7c8ah9qzr65gkdid7p8r6qj309ngnr0-libseccomp-2.5.5-dev/include -I/nix/store/152md2dw0gf33ag9rqhf5zmh93pvkazx-libarchive-3.7.3-dev/include -I/nix/store/jhkpr5iyyx35c1sg52vdgjpvky03xi1n-brotli-1.1.0-dev/include -I/nix/store/5m7d3z1xcwy0gagkginzqvhcrxifzwqf-openssl-3.0.13-dev/include -I/nix/store/jwhvd9dhwhd0xpx00k7slv4b26n024r6-nlohmann_json-3.11.3/include -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -std=c++2a -include config.h -Wno-deprecated-declarations -Wimplicit-fallthrough -Werror=switch -Werror=switch-enum -D_GLIBCXX_ASSERTIONS=1 -fsanitize=signed-integer-overflow -fsanitize-undefined-trap-on-error -ffile-prefix-map=../src=src -fPIC -DBOOST_CONTEXT_DYN_LINK=1 -DBOOST_COROUTINES_DYN_LINK=1 -DBOOST_CONTAINER_DYN_LINK=1 -DBOOST_ALL_NO_LIB -MD -MQ src/libutil/libnixutil.so.p/position.cc.o -MF src/libutil/libnixutil.so.p/position.cc.o.d -o src/libutil/libnixutil.so.p/position.cc.o -c ../src/libutil/position.cc

... no optimization flags?! That would probably explain it.

I built the derivation with `ninjaFlags = [ "-v" ];` and: ``` [10/456] g++ -Isrc/libutil/libnixutil.so.p -Isrc/libutil -I../src/libutil -I/nix/store/rh6sv1rpxavhhzprbaqpkqdm4sq39d19-boehm-gc-8.2.6-dev/include -I/nix/store/73ik8xsdqv0grj8z4rplfqhzb5ajd4s3-boost-1.81.0-dev/include -I/nix/store/lpsa98iwl9q7fb2hdiqhqcryyg2kvgnm-libcpuid-0.6.5/include/libcpuid -I/nix/store/j7c8ah9qzr65gkdid7p8r6qj309ngnr0-libseccomp-2.5.5-dev/include -I/nix/store/152md2dw0gf33ag9rqhf5zmh93pvkazx-libarchive-3.7.3-dev/include -I/nix/store/jhkpr5iyyx35c1sg52vdgjpvky03xi1n-brotli-1.1.0-dev/include -I/nix/store/5m7d3z1xcwy0gagkginzqvhcrxifzwqf-openssl-3.0.13-dev/include -I/nix/store/jwhvd9dhwhd0xpx00k7slv4b26n024r6-nlohmann_json-3.11.3/include -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -std=c++2a -include config.h -Wno-deprecated-declarations -Wimplicit-fallthrough -Werror=switch -Werror=switch-enum -D_GLIBCXX_ASSERTIONS=1 -fsanitize=signed-integer-overflow -fsanitize-undefined-trap-on-error -ffile-prefix-map=../src=src -fPIC -DBOOST_CONTEXT_DYN_LINK=1 -DBOOST_COROUTINES_DYN_LINK=1 -DBOOST_CONTAINER_DYN_LINK=1 -DBOOST_ALL_NO_LIB -MD -MQ src/libutil/libnixutil.so.p/position.cc.o -MF src/libutil/libnixutil.so.p/position.cc.o.d -o src/libutil/libnixutil.so.p/position.cc.o -c ../src/libutil/position.cc ``` ... no optimization flags?! That would probably explain it.
qyriad commented 2024-05-12 15:46:21 +00:00
Owner
Copy link

…right, mesonConfigurePhase sets mesonBuildType/-Dbuild_type=plain by default…

…right, mesonConfigurePhase sets `mesonBuildType`/`-Dbuild_type=plain` by default…
delroth commented 2024-05-12 15:50:09 +00:00
Member
Copy link

With mesonBuildType = "release"; patched in to the nixpkgs expression (and checked that -O3 appears in the build flags):

/nix/store/3nijav5qh5960976z10bmwn99vc8ibik-lix-2.90-beta.1       91.8M
/nix/store/awshr3d8a94v2igyf16jh5p8bw9wf93s-nix-2.18.2    84.4M

So that alone doesn't fix it. Should probably still be fixed (and also in Lix's package.nix maybe?).

With `mesonBuildType = "release";` patched in to the nixpkgs expression (and checked that -O3 appears in the build flags): ``` /nix/store/3nijav5qh5960976z10bmwn99vc8ibik-lix-2.90-beta.1 91.8M /nix/store/awshr3d8a94v2igyf16jh5p8bw9wf93s-nix-2.18.2 84.4M ``` So that alone doesn't fix it. Should probably still be fixed (and also in Lix's `package.nix` maybe?).
delroth commented 2024-05-12 16:09:09 +00:00
Member
Copy link

Current hypothesis: this is because Nix >= 2.10 on GNU stdenvs uses -flto which Lix doesn't seem to use. This would explain why the discrepancy is only happening on the Linux builds and not the Darwin builds.

Current hypothesis: this is because Nix >= 2.10 on GNU stdenvs uses `-flto` which Lix doesn't seem to use. This would explain why the discrepancy is only happening on the Linux builds and not the Darwin builds.
delroth commented 2024-05-12 16:17:15 +00:00
Member
Copy link

with -Db_lto=true and mesonBuildType = "release";:

/nix/store/9lhnrg0rwjkk9si9fnpvyrbkkajw32ca-lix-2.90-beta.1       87.4M
/nix/store/awshr3d8a94v2igyf16jh5p8bw9wf93s-nix-2.18.2            84.4M

Getting there.

with `-Db_lto=true` and `mesonBuildType = "release";`: ``` /nix/store/9lhnrg0rwjkk9si9fnpvyrbkkajw32ca-lix-2.90-beta.1 87.4M /nix/store/awshr3d8a94v2igyf16jh5p8bw9wf93s-nix-2.18.2 84.4M ``` Getting there.
🎉 2
pennae commented 2024-05-12 16:28:16 +00:00
Owner
Copy link

Current hypothesis: this is because Nix >= 2.10 on GNU stdenvs uses -flto which Lix doesn't seem to use. This would explain why the discrepancy is only happening on the Linux builds and not the Darwin builds.

since you say this: lix should really enable lto in nixpkgs. lto makes quite a difference in eval performance (which is why we added it to nix in the first place 🫠)

> Current hypothesis: this is because Nix >= 2.10 on GNU stdenvs uses `-flto` which Lix doesn't seem to use. This would explain why the discrepancy is only happening on the Linux builds and not the Darwin builds. since you say this: lix should *really* enable lto in nixpkgs. lto makes quite a difference in eval performance (which is why we added it to nix in the first place 🫠)
delroth commented 2024-05-12 16:33:30 +00:00
Member
Copy link

Getting there.

The last 3MB are in very large part attributable to the Rust lix-doc being statically linked into libnixcmd. It's dragging in a good chunk of runtime, from what I can tell. Not exactly sure what can be done there since it's outside of my expertise - maybe it's dead code that could be eliminated with -Wl,--gc-sections and the equivalent of -ffunction-sections on the Rust side of things?

Could also treat as WAI since at least we have an explanation now.

> Getting there. The last 3MB are in very large part attributable to the Rust lix-doc being statically linked into libnixcmd. It's dragging in a good chunk of runtime, from what I can tell. Not exactly sure what can be done there since it's outside of my expertise - maybe it's dead code that could be eliminated with `-Wl,--gc-sections` and the equivalent of `-ffunction-sections` on the Rust side of things? Could also treat as WAI since at least we have an explanation now.
👍 1
raito commented 2024-05-12 16:38:41 +00:00
Author
Owner
Copy link

@delroth Perfect, that makes total sense to me.

@delroth Perfect, that makes total sense to me.
delroth commented 2024-05-12 16:54:52 +00:00
Member
Copy link

since you say this: lix should really enable lto in nixpkgs. lto makes quite a difference in eval performance (which is why we added it to nix in the first place 🫠)

Should package.nix and/or lix-project/nixos-module also enable it? AFAICT they currently don't, but I might be mistaken.

> since you say this: lix should *really* enable lto in nixpkgs. lto makes quite a difference in eval performance (which is why we added it to nix in the first place 🫠) Should `package.nix` and/or `lix-project/nixos-module` also enable it? AFAICT they currently don't, but I might be mistaken.
pennae commented 2024-05-12 17:00:38 +00:00
Owner
Copy link

package.nix probably shouldn't (but should give the option), the module may want to override this choice to give a close-to-nixos build result

`package.nix` probably shouldn't (but should give the option), the module may want to override this choice to give a close-to-nixos build result
jade commented 2024-05-12 19:14:01 +00:00
Owner
Copy link

now hold your horses though, because release is -O3 and I don't think we want that configuration to even be allowed after we found that gcc warning bug on O3. i think you want debugoptimized which is O2 and has debug info available.

now hold your horses though, because release is -O3 and I don't think we want that configuration to even be *allowed* after we found that gcc warning bug on O3. i think you want debugoptimized which is O2 and has debug info available.
raito commented 2024-05-12 19:43:16 +00:00
Author
Owner
Copy link

now hold your horses though, because release is -O3 and I don't think we want that configuration to even be allowed after we found that gcc warning bug on O3. i think you want debugoptimized which is O2 and has debug info available.

For the nixpkgs release?

> now hold your horses though, because release is -O3 and I don't think we want that configuration to even be *allowed* after we found that gcc warning bug on O3. i think you want debugoptimized which is O2 and has debug info available. For the nixpkgs release?
jade commented 2024-05-12 20:46:34 +00:00
Owner
Copy link

In general. I am not convinced that O3 is sound in gcc given that warning implies it thinks something can be uninit when it can't be.

In general. I am not convinced that O3 is sound in gcc given that warning implies it thinks something can be uninit when it can't be.
delroth commented 2024-05-12 20:48:56 +00:00
Member
Copy link

@jade do you have a ref to that warning you're talking about? I didn't see it on my side when building in release mode.

In general I'd say that building with build_type = release is a fairly intuitive thing to do, and users will try and do it. So if we think the set of flags used in that mode is wrong, they should probably be fixed in the Meson config. WDYT?

@jade do you have a ref to that warning you're talking about? I didn't see it on my side when building in `release` mode. In general I'd say that building with `build_type = release` is a fairly intuitive thing to do, and users will try and do it. So if we think the set of flags used in that mode is wrong, they should probably be fixed in the Meson config. WDYT?
jade commented 2024-05-12 20:49:58 +00:00
Owner
Copy link

yes that would be #46

yes that would be https://git.lix.systems/lix-project/lix/issues/46
👍 1
qyriad commented 2024-05-12 21:18:50 +00:00
Owner
Copy link

they should probably be fixed in the Meson config. WDYT?

note that meson doesn't allow you to force a build type or optimization level within meson.build; you can only default it if not specified (and the meson hook in Nixpkgs never uses the default). You can add the -O2 flags yourself, but Meson will probably emit a warning about it and/or apply the flags from its configured build type/opt level

> they should probably be fixed in the Meson config. WDYT? note that meson doesn't allow you to force a build type or optimization level within `meson.build`; you can only *default* it if not specified (and the meson hook in Nixpkgs never uses the default). You can add the `-O2` flags yourself, but Meson will probably emit a warning about it and/or apply the flags from its configured build type/opt level
jade commented 2024-05-13 01:30:25 +00:00
Owner
Copy link

can we detect people setting optimization to 3 and throw an error? that's my preference i think.

can we detect people setting optimization to 3 and throw an error? that's my preference i think.
delroth commented 2024-05-13 01:32:18 +00:00
Member
Copy link

Doesn't that seem premature when there is no evidence of harm from -O3? AFAICT all that has been reported so far is spurious warnings - all Lix tests are still passing.

Doesn't that seem premature when there is no evidence of harm from -O3? AFAICT all that has been reported so far is spurious warnings - all Lix tests are still passing.
jade commented 2024-05-13 01:56:52 +00:00
Owner
Copy link

Doesn't that seem premature when there is no evidence of harm from -O3? AFAICT all that has been reported so far is spurious warnings - all Lix tests are still passing.

The thing about the warning being reported is it implies there's an unsound optimization pass, or that that particular warning warns on legit code. And I'm mildly scared of it being the former case.

> Doesn't that seem premature when there is no evidence of harm from -O3? AFAICT all that has been reported so far is spurious warnings - all Lix tests are still passing. The thing about the warning being reported is it implies there's an unsound optimization pass, *or* that that particular warning warns on legit code. And I'm mildly scared of it being the former case.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
build-remote-ipc
sb/raito/phantom-referrers-gc
repl-overlay-example
release-2.91
reuse-compliance
release-2.90
git-series/activity-cleanup
sb/tom-hubrecht/primops.cc-v2
sb/bacchanalia/cache-strlen%topic=substr-perf
sb/arcuru/commit-msg
sb/qyriad/beta-1
sb/qyriad/2.90-beta.0
rbt/pre-commit-update
sb/rbt/test-branch
sb/rbt/justfile
sb/rbt/repl-overlays
sb/fuck/what
sb/rbt/pre-commit
sb/rbt/makeflags
sb/pennae/io-rewrite
sb/qyriad/maint-meson
sb/qyriad/maint/meson
sb/puck/meson-c-api
sb/pennae/parser-rewrite
sb/jade/burn-it-all
dependabot/github_actions/zeebe-io/backport-action-2.4.1
dependabot/github_actions/cachix/install-nix-action-25
dependabot/github_actions/actions/labeler-5
dependabot/github_actions/cachix/cachix-action-14
2.91.0
2.90.0
2.90.0-rc1
2.90-beta.1
2.90-beta.0
Labels
Clear labels   
Area/build-packaging

regarding the building or packaging of Lix itself

  
Area/cli

The command line interface of Lix

  
Area/evaluator

libexpr and anything else related to the evaluation of Nix expressions

  
Area/flakes

   
Area/language

language design, features, extensions, etc

  
Area/profiles

profiles: nix-env, nix profile, and installer flavors all alike

  
Area/protocol

The protocol, used for communicating between the client and the daemon

  
Area/releng

Release engineering

  
Area/remote-builds

remote builds

  
Area/repl

nix repl

  
Area/store

libstore and anything else relating to the Nix store

  
bug

it's a bug

  
Cross Compilation

   
devx

development experience issues

  
docs

requires or is about documentation changes

  
Downstream Dependents

regarding an external tool's usage with Lix (e.g. nix-eval-jobs)

  
E/easy

bug is relatively simple and could simply be fixed with some minor effort

  
E/hard

this issue should probably be solved with close involvement of an expert

  
E/help wanted

maintainers may not get to this soon, but we want it fixed

  
E/reproducible

This bug is reproducible and obvious.

  
E/requires rearchitecture

this bug cannot be fixed properly without fixing a known architectural flaw

  
imported

Issue imported from upstream nix (please do not remove this label, it will cause dupes if import is rerun)

  
Needs Langver

We can't get to this until we implement language versioning

  
OS/Linux

Affects specifically Lix on Linux

  
OS/macOS

Affects specifically Lix on macOS

  
performance

make lix go faster

  
regression

This is a regression since some previous version of Lix

  
release-blocker

Blocks public release

  
RFD

Request for discussion

  
stability

Fixing this makes the software more stable

  
Status
blocked
Blocked on something

  
Status
invalid
not our bug, duplicate, similar

  
Status
postponed
closed but should maybe look at later (generally with S-wontfix)

  
Status
wontfix
We chose not to change this (see issue comments for more detail).

  
testing

Tests that should be added

  
testing/flakey

Tests that are unreliable/flakey (not to be confused with flakes that test)

  
ux

makes the user experience better :)

No labels
Area/build-packaging
Area/cli
Area/evaluator
Area/flakes
Area/language
Area/profiles
Area/protocol
Area/releng
Area/remote-builds
Area/repl
Area/store
bug
Cross Compilation
devx
docs
Downstream Dependents
E/easy
E/hard
E/help wanted
E/reproducible
E/requires rearchitecture
imported
Needs Langver
OS/Linux
OS/macOS
performance
regression
release-blocker
RFD
stability
Status
blocked
Status
invalid
Status
postponed
Status
wontfix
testing
testing/flakey
ux
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
5 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: lix-project/lix#308
No description provided.
Delete branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?