Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends - #159999

Kobzol · 2026-07-27T08:44:44Z

This was quite the detective hunt. Antoyo reached out to me that when using the GCC codegen backend by default (--set rust.codegen-backends=["gcc"]), the stdlib is invalidated after every rebuild.

I did some digging, and found out this:

The Builder::cargo step manually invalidates the target directory if any of the used codegen backends is dirty, because apparently the used codegen backend is not tracked by -Zbinary-dep-depinfo yet.
It does so by getting all paths from <sysroot>/lib/rustlib/<target>/codegen-backends via the Builder::codegen_backends function, and if any such path is newer than the output directory, the output directory is deleted.
The intention was for this function to return codegen backends, i.e. .so/.dll files. However, since the GCC codegen backend has additional dependencies (libgccjit), the codegen-backends contains a lib subdirectory. And bootstrap was checking the mtime of this directory, which was always considered dirty.

So I changed the function to only return the actual codegen backends, which fixes the issue. In theory, we should also be checking all the nested contents of the lib directory, i.e. the libgccjit dylibs, but I think that if they change, the GCC codegen backend itself should also change, in the vast majority of cases, so I consider this to be a worthy fix in the meantime.

CC @antoyo @bjorn3

…ackends

rustbot · 2026-07-27T08:44:49Z

r? @Mark-Simulacrum

rustbot has assigned @Mark-Simulacrum.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: bootstrap
bootstrap expanded to 6 candidates
Random selection from Mark-Simulacrum, clubby789, jieyouxu

Kobzol · 2026-07-30T12:32:22Z

r? @bjorn3

bjorn3 · 2026-07-31T09:18:33Z

@bors r+

rust-bors · 2026-07-31T09:18:36Z

📌 Commit 216ee73 has been approved by bjorn3

It is now in the queue for this repository.

@antoyo

…nvalidation, r=bjorn3 Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends This was quite the detective hunt. Antoyo reached out to me that when using the GCC codegen backend by default (`--set rust.codegen-backends=["gcc"]`), the stdlib is invalidated after every rebuild. I did some digging, and found out this: - The `Builder::cargo` step manually invalidates the target directory if any of the used codegen backends is dirty, because apparently the used codegen backend is not tracked by `-Zbinary-dep-depinfo` yet. - It does so by getting all paths from `<sysroot>/lib/rustlib/<target>/codegen-backends` via the `Builder::codegen_backends` function, and if any such path is newer than the output directory, the output directory is deleted. - The intention was for this function to return codegen backends, i.e. `.so/.dll` files. However, since the GCC codegen backend has additional dependencies (libgccjit), the `codegen-backends` contains a `lib` subdirectory. And bootstrap was checking the mtime of this *directory*, which was always considered dirty. So I changed the function to only return the actual codegen backends, which fixes the issue. In theory, we should also be checking all the nested contents of the `lib` directory, i.e. the libgccjit dylibs, but I think that if they change, the GCC codegen backend itself should also change, in the vast majority of cases, so I consider this to be a worthy fix in the meantime. CC @antoyo @bjorn3

@antoyo

…nvalidation, r=bjorn3 Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends This was quite the detective hunt. Antoyo reached out to me that when using the GCC codegen backend by default (`--set rust.codegen-backends=["gcc"]`), the stdlib is invalidated after every rebuild. I did some digging, and found out this: - The `Builder::cargo` step manually invalidates the target directory if any of the used codegen backends is dirty, because apparently the used codegen backend is not tracked by `-Zbinary-dep-depinfo` yet. - It does so by getting all paths from `<sysroot>/lib/rustlib/<target>/codegen-backends` via the `Builder::codegen_backends` function, and if any such path is newer than the output directory, the output directory is deleted. - The intention was for this function to return codegen backends, i.e. `.so/.dll` files. However, since the GCC codegen backend has additional dependencies (libgccjit), the `codegen-backends` contains a `lib` subdirectory. And bootstrap was checking the mtime of this *directory*, which was always considered dirty. So I changed the function to only return the actual codegen backends, which fixes the issue. In theory, we should also be checking all the nested contents of the `lib` directory, i.e. the libgccjit dylibs, but I think that if they change, the GCC codegen backend itself should also change, in the vast majority of cases, so I consider this to be a worthy fix in the meantime. CC @antoyo @bjorn3

@antoyo

…nvalidation, r=bjorn3 Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends This was quite the detective hunt. Antoyo reached out to me that when using the GCC codegen backend by default (`--set rust.codegen-backends=["gcc"]`), the stdlib is invalidated after every rebuild. I did some digging, and found out this: - The `Builder::cargo` step manually invalidates the target directory if any of the used codegen backends is dirty, because apparently the used codegen backend is not tracked by `-Zbinary-dep-depinfo` yet. - It does so by getting all paths from `<sysroot>/lib/rustlib/<target>/codegen-backends` via the `Builder::codegen_backends` function, and if any such path is newer than the output directory, the output directory is deleted. - The intention was for this function to return codegen backends, i.e. `.so/.dll` files. However, since the GCC codegen backend has additional dependencies (libgccjit), the `codegen-backends` contains a `lib` subdirectory. And bootstrap was checking the mtime of this *directory*, which was always considered dirty. So I changed the function to only return the actual codegen backends, which fixes the issue. In theory, we should also be checking all the nested contents of the `lib` directory, i.e. the libgccjit dylibs, but I think that if they change, the GCC codegen backend itself should also change, in the vast majority of cases, so I consider this to be a worthy fix in the meantime. CC @antoyo @bjorn3

…uwer Rollup of 21 pull requests Successful merges: - #160100 (Add "system" option to `override-allocator` directive) - #159675 (rustc: Support `--jobs` options for limiting parallelism in various parts of the compiler) - #159999 (Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends) - #160233 (Bubble bad path error while parsing field to avoid unecessary second error) - #160272 (rustc_metadata: Move native library search code to `rustc_codegen_ssa`) - #154202 (rustfmt: Format `cfg_select!`) - #158835 (rustc_passes: lint unused `#[path]` attributes on inline modules) - #159520 (Suggest `Vec<T>` instead of `[T]`) - #160034 (Move "macro only" check for `#[allow_internal_unsafe/unstable]` to attribute parser) - #160066 (rustc_middle: lint attribute cleanups) - #160085 (Remove various superfluous lint attributes) - #160113 (Coalesce `rustc_on_unimplemented` attributes and lint malformed filters) - #160119 (fix query cycle in `coroutine_hidden_types` for the next solver) - #160147 (tests: Remove `-Zthreads` options from tests in `ui/parallel-rustc`) - #160157 (Remove outdated comments from `va_list.rs`) - #160159 (More accurately check for interior mutability in `invalid_reference_casting` lint) - #160208 (rustdoc: Fix crash when trying to list attributes on an opaque type) - #160244 (Rename splat to rustc_splat in error messages) - #160246 (dont fire `unused_mut` on `&pin mut self`) - #160247 (Configure backport nominations for rustfmt) - #160274 (renovate: update lock files weekly)

…uwer Rollup of 22 pull requests Successful merges: - #160100 (Add "system" option to `override-allocator` directive) - #160220 (Refactor: shrink region ext traits) - #159675 (rustc: Support `--jobs` options for limiting parallelism in various parts of the compiler) - #159999 (Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends) - #160233 (Bubble bad path error while parsing field to avoid unecessary second error) - #160272 (rustc_metadata: Move native library search code to `rustc_codegen_ssa`) - #154202 (rustfmt: Format `cfg_select!`) - #159520 (Suggest `Vec<T>` instead of `[T]`) - #159710 (Add rustdoc/cargo PGO profiles to reproducible artifacts) - #160034 (Move "macro only" check for `#[allow_internal_unsafe/unstable]` to attribute parser) - #160066 (rustc_middle: lint attribute cleanups) - #160085 (Remove various superfluous lint attributes) - #160113 (Coalesce `rustc_on_unimplemented` attributes and lint malformed filters) - #160119 (fix query cycle in `coroutine_hidden_types` for the next solver) - #160147 (tests: Remove `-Zthreads` options from tests in `ui/parallel-rustc`) - #160157 (Remove outdated comments from `va_list.rs`) - #160159 (More accurately check for interior mutability in `invalid_reference_casting` lint) - #160208 (rustdoc: Fix crash when trying to list attributes on an opaque type) - #160244 (Rename splat to rustc_splat in error messages) - #160246 (dont fire `unused_mut` on `&pin mut self`) - #160247 (Configure backport nominations for rustfmt) - #160274 (renovate: update lock files weekly)

@antoyo

Rollup merge of #159999 - Kobzol:bootstrap-codegen-backend-invalidation, r=bjorn3 Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends This was quite the detective hunt. Antoyo reached out to me that when using the GCC codegen backend by default (`--set rust.codegen-backends=["gcc"]`), the stdlib is invalidated after every rebuild. I did some digging, and found out this: - The `Builder::cargo` step manually invalidates the target directory if any of the used codegen backends is dirty, because apparently the used codegen backend is not tracked by `-Zbinary-dep-depinfo` yet. - It does so by getting all paths from `<sysroot>/lib/rustlib/<target>/codegen-backends` via the `Builder::codegen_backends` function, and if any such path is newer than the output directory, the output directory is deleted. - The intention was for this function to return codegen backends, i.e. `.so/.dll` files. However, since the GCC codegen backend has additional dependencies (libgccjit), the `codegen-backends` contains a `lib` subdirectory. And bootstrap was checking the mtime of this *directory*, which was always considered dirty. So I changed the function to only return the actual codegen backends, which fixes the issue. In theory, we should also be checking all the nested contents of the `lib` directory, i.e. the libgccjit dylibs, but I think that if they change, the GCC codegen backend itself should also change, in the vast majority of cases, so I consider this to be a worthy fix in the meantime. CC @antoyo @bjorn3

…uwer Rollup of 22 pull requests Successful merges: - rust-lang/rust#160100 (Add "system" option to `override-allocator` directive) - rust-lang/rust#160220 (Refactor: shrink region ext traits) - rust-lang/rust#159675 (rustc: Support `--jobs` options for limiting parallelism in various parts of the compiler) - rust-lang/rust#159999 (Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends) - rust-lang/rust#160233 (Bubble bad path error while parsing field to avoid unecessary second error) - rust-lang/rust#160272 (rustc_metadata: Move native library search code to `rustc_codegen_ssa`) - rust-lang/rust#154202 (rustfmt: Format `cfg_select!`) - rust-lang/rust#159520 (Suggest `Vec<T>` instead of `[T]`) - rust-lang/rust#159710 (Add rustdoc/cargo PGO profiles to reproducible artifacts) - rust-lang/rust#160034 (Move "macro only" check for `#[allow_internal_unsafe/unstable]` to attribute parser) - rust-lang/rust#160066 (rustc_middle: lint attribute cleanups) - rust-lang/rust#160085 (Remove various superfluous lint attributes) - rust-lang/rust#160113 (Coalesce `rustc_on_unimplemented` attributes and lint malformed filters) - rust-lang/rust#160119 (fix query cycle in `coroutine_hidden_types` for the next solver) - rust-lang/rust#160147 (tests: Remove `-Zthreads` options from tests in `ui/parallel-rustc`) - rust-lang/rust#160157 (Remove outdated comments from `va_list.rs`) - rust-lang/rust#160159 (More accurately check for interior mutability in `invalid_reference_casting` lint) - rust-lang/rust#160208 (rustdoc: Fix crash when trying to list attributes on an opaque type) - rust-lang/rust#160244 (Rename splat to rustc_splat in error messages) - rust-lang/rust#160246 (dont fire `unused_mut` on `&pin mut self`) - rust-lang/rust#160247 (Configure backport nominations for rustfmt) - rust-lang/rust#160274 (renovate: update lock files weekly)

Fix invalidation of stdlib in bootstrap when using non-LLVM codegen b…

216ee73

…ackends

rustbot assigned Mark-Simulacrum Jul 27, 2026

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-bootstrap Relevant to the bootstrap subteam: Rust's build system (x.py and src/bootstrap) labels Jul 27, 2026

rustbot assigned bjorn3 and unassigned Mark-Simulacrum Jul 30, 2026

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jul 31, 2026

JonathanBrouwer mentioned this pull request Jul 31, 2026

Rollup of 11 pull requests #160273

Closed

JonathanBrouwer mentioned this pull request Jul 31, 2026

Rollup of 17 pull requests #160278

Closed

JonathanBrouwer mentioned this pull request Jul 31, 2026

Rollup of 21 pull requests #160283

Closed

JonathanBrouwer mentioned this pull request Jul 31, 2026

Rollup of 22 pull requests #160284

Merged

rust-bors Bot merged commit 1b7af29 into rust-lang:main Jul 31, 2026
13 checks passed

rustbot added this to the 1.99.0 milestone Jul 31, 2026

Kobzol deleted the bootstrap-codegen-backend-invalidation branch August 1, 2026 06:34

Navigation Menu

Uh oh!

Uh oh!

Fix invalidation of stdlib in bootstrap when using non-LLVM codegen backends - #159999