Only pass --[no-]gc-sections if linker is GNU ld. by luqmana · Pull Request #85274 · rust-lang/rust

luqmana · 2021-05-14T00:09:38Z

Fixes a regression from #84468 where linking now fails with solaris linkers. LinkerFlavor::Gcc does not always mean GNU ld specifically. And in the case of at least the solaris ld in illumos, that flag is unrecognized and will cause the linking step to fail.

Even though removing the is_like_solaris branch from gc_sections in #84468 made sense as -z ignore/record are more analogous to the --[no-]-as-needed flags, it inadvertently caused solaris linkers to be passed the --gc-sections flag. So let's just change it to be more explicit about when we pass those flags.

LinkerFlavor::Gcc does not always mean GNU ld specifically. And in the case of at least the solaris ld in illumos, that flag is unrecognized and will cause the linking step to fail.

rust-highfive · 2021-05-14T00:09:41Z

r? @jackh726

(rust-highfive has picked a reviewer for you, use r? to override)

luqmana · 2021-05-14T00:15:13Z

r? @petrochenkov

petrochenkov · 2021-05-15T17:45:42Z

@bors r+

bors · 2021-05-15T17:45:44Z

📌 Commit 3fe1d7f has been approved by petrochenkov

bors · 2021-05-16T03:35:00Z

⌛ Testing commit 3fe1d7f with merge 5d28235010a9b782714887f7a7e91bf7dd8da0b4...

… also indicate linker_is_gnu.

luqmana · 2021-05-16T05:20:27Z

Test failed because the mingw targets do use gcc as the linker but the target spec never indicated linker_is_gnu so just updated that target spec.

petrochenkov · 2021-05-16T10:12:35Z

@luqmana
The windows_gnu_base.rs spec was right, the windows-gnu gcc/ld targets COFF so it's missing a bunch of options from the regular ELF-targeting gcc/ld's, that's why it wasn't marked as linker_is_gnu.

If it supports --gc-sections, then the condition should be self.sess.target.linker_is_gnu || self.sess.target.is_like_windows

luqmana · 2021-05-16T23:27:01Z

Hmm, then should the docs be updated since they say linker_is_gnu is:

Whether the linker support GNU-like arguments such as -O. Defaults to false.

The man pages for i686-w64-mingw32-ld even describe it as "The GNU linker". I also checked all the branches we use linker_is_gnu and all of them seem to work fine for mingw targets (tested by trying to pass them to i686-w64-mingw32-gcc -Wl,. So I guess the question is what option(s) in particular was preventing us from marking the mingw targets as such?

petrochenkov · 2021-05-17T21:55:21Z

So I guess the question is what option(s) in particular was preventing us from marking the mingw targets as such?

Here's what my investigation showed:

-pie errors with LLD, acts as compatibility noop with BFD
--as-needed errors with LLD, acts as compatibility noop with BFD
-znoexecstack errors with both LLD and BFD, but never hit in rustc due to the previous is_like_windows condition
--enable-new-dtags errors with both LLD and BFD, but never hit in rustc because has_rpath is false for windows-gnu targets

I'm ok with setting linker_is_gnu to true for windows-gnu targets, because the meaning of linker_is_gnu is vague enough in general, it's just some code need to be adjusted in the linker_is_gnu -> linker_is_gnu && !is_like_windows way for LLD to work.

luqmana · 2021-05-17T23:33:05Z

@petrochenkov Thanks for digging into that! Updated the PR to handle those now

petrochenkov · 2021-05-18T09:31:16Z

r=me with #85274 (comment) addressed.

petrochenkov · 2021-05-18T11:44:09Z

Thanks!
@bors r+

bors · 2021-05-18T11:44:10Z

📌 Commit ac5fd90 has been approved by petrochenkov

…r=petrochenkov Only pass --[no-]gc-sections if linker is GNU ld. Fixes a regression from rust-lang#84468 where linking now fails with solaris linkers. LinkerFlavor::Gcc does not always mean GNU ld specifically. And in the case of at least the solaris ld in illumos, that flag is unrecognized and will cause the linking step to fail. Even though removing the `is_like_solaris` branch from `gc_sections` in rust-lang#84468 made sense as `-z ignore/record` are more analogous to the `--[no-]-as-needed` flags, it inadvertently caused solaris linkers to be passed the `--gc-sections` flag. So let's just change it to be more explicit about when we pass those flags.

Rollup of 8 pull requests Successful merges: - rust-lang#83366 (Stabilize extended_key_value_attributes) - rust-lang#83767 (Fix v0 symbol mangling bug) - rust-lang#84883 (compiletest: "fix" FileCheck with --allow-unused-prefixes) - rust-lang#85274 (Only pass --[no-]gc-sections if linker is GNU ld.) - rust-lang#85297 (bootstrap: build cargo only if requested in tools) - rust-lang#85396 (rustdoc: restore header sizes) - rust-lang#85425 (Fix must_use on `Option::is_none`) - rust-lang#85438 (Fix escape handling) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

Rust used to pass `--gc-section` automatically. In rust-lang/rust#85274, this behavior was changed to only pass that flag for targets that use a GNU linker. So we have to add it manually, otherwise the bootloader grows too large to be loaded by our assembly stage.

Swap TargetOptions::linker_is_gnu default from false to true and update targets as appropriate. rust-lang#85274 gated the `--gc-sections` flag on targets that specified `linker_is_gnu` to stop us from passing it to incompatible linkers. But that had the unintended effect of the flag no longer being passed on targets for which it is valid and hence caused a regression in binary size. Given that most `ld`-style linkers are GNU compatible, this change flips our default for `linker_is_gnu` from false to true. That also means updating the targets that relied on the previous default: * Apple * Illumos * L4Re (not sure about this one) * MSVC * NvtPtx * Solaris Fixesrust-lang#85519

Only pass --[no-]gc-sections if linker is GNU ld.
3fe1d7f
LinkerFlavor::Gcc does not always mean GNU ld specifically. And in the case of at least the solaris ld in illumos, that flag is unrecognized and will cause the linking step to fail.

rust-highfive assigned jackh726 May 14, 2021

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label May 14, 2021

rust-highfive assigned petrochenkov and unassigned jackh726 May 14, 2021

bjorn3 reviewed May 14, 2021
View reviewed changes

Comment threadcompiler/rustc_codegen_ssa/src/back/linker.rs

bors 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 May 15, 2021

This comment has been minimized.
Sign in to view

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels May 16, 2021

Windows mingw targets use gcc as the linker so the target spec should…
45225d2
… also indicate linker_is_gnu.

petrochenkov added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 16, 2021

petrochenkov added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels May 17, 2021

petrochenkov added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 17, 2021

Adjust linker_is_gnu branches for cases that don't work on windows.
a862b1f

petrochenkov reviewed May 18, 2021
View reviewed changes

Comment threadcompiler/rustc_codegen_ssa/src/back/linker.rs Outdated

Undo unnecessary changes.
e0d5872

petrochenkov reviewed May 18, 2021
View reviewed changes

Comment threadcompiler/rustc_codegen_ssa/src/back/linker.rs Outdated

Don't pass -pie to linker on windows targets.
ac5fd90

bors 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-author Status: This is awaiting some action (such as code changes or more information) from the author. labels May 18, 2021

Dylan-DPC-zz mentioned this pull request May 18, 2021
Rollup of 9 pull requests #85452
Closed

jackh726 mentioned this pull request May 19, 2021
Rollup of 8 pull requests #85458
Merged

bors merged commit ec0e0d1 into rust-lang:masterMay 19, 2021

rustbot added this to the 1.54.0 milestone May 19, 2021

luqmana deleted the linker-is-gnu-gc-sections branch May 19, 2021 17:06

bjorn3 mentioned this pull request May 20, 2021
[Fixed] Bootloader v0.9.17 seems to be broken on Rust nightly-2021-05-20rust-osdev/bootloader#167
Closed

phil-opp mentioned this pull request May 20, 2021
Fix nightly regression by manually passing --gc-sectionsrust-osdev/bootloader#168
Merged

jonas-schievink mentioned this pull request May 20, 2021
28x binary size regression in recent nightly #85519
Closed

luqmana mentioned this pull request May 21, 2021
Swap TargetOptions::linker_is_gnu default from false to true and update targets as appropriate. #85531
Merged

phil-opp mentioned this pull request May 21, 2021
Likely possible to revert #168 in the next days rust-osdev/bootloader#171
Closed

Uh oh!

Uh oh!

Conversation

luqmana commented May 14, 2021

Uh oh!

rust-highfive commented May 14, 2021

Uh oh!

luqmana commented May 14, 2021

Uh oh!

Uh oh!

petrochenkov commented May 15, 2021

Uh oh!

bors commented May 15, 2021

Uh oh!

bors commented May 16, 2021

Uh oh!

This comment has been minimized.

This comment has been minimized.

luqmana commented May 16, 2021

Uh oh!

petrochenkov commented May 16, 2021

Uh oh!

luqmana commented May 16, 2021

Uh oh!

petrochenkov commented May 17, 2021

Uh oh!

luqmana commented May 17, 2021

Uh oh!

Uh oh!

petrochenkov commented May 18, 2021

Uh oh!

Uh oh!

petrochenkov commented May 18, 2021

Uh oh!

bors commented May 18, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants