Use the .drectve section for exporting symbols from dlls on Windows #142568

bjorn3 · 2025-06-16T09:24:31Z

While it would be reasonable to expect the Windows linker to handle linker args in the .drectve section identical to cli arguments, as it turns out exporting weak symbols only works when the /EXPORT is in the .drectve section, not when it is a linker argument or when a .DEF file is used.

Necessary for and extracted out of #142366.

Thanks @dpaoliello for figuring out this weird quirk of link.exe!

rustbot · 2025-06-16T09:24:35Z

r? @wesleywiser

rustbot has assigned @wesleywiser.
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

rustbot · 2025-06-16T09:24:36Z

Some changes occurred in compiler/rustc_codegen_ssa

cc @WaffleLapkin

compiler/rustc_codegen_ssa/src/back/linker.rs

bjorn3 · 2025-06-16T13:29:19Z

@bors2 try jobs=x86_64-msvc-1,x86_64-msvc-2

rust-bors · 2025-06-16T13:29:23Z

⌛ Trying commit 5e2efb6 with merge 7bae343…

To cancel the try build, run the command @bors2 try cancel.

Use the .drectve section for exporting symbols from dlls on Windows While it would be reasonable to expect the Windows linker to handle linker args in the .drectve section identical to cli arguments, as it turns out exporting weak symbols only works when the /EXPORT is in the .drectve section, not when it is a linker argument or when a .DEF file is used. Necessary for and extracted out of #142366. Thanks `@dpaoliello` for figuring out this weird quirk of link.exe! try-job: x86_64-msvc-1 try-job: x86_64-msvc-2

rust-bors · 2025-06-16T14:51:23Z

💔 Test failed

CI ❌

bjorn3 · 2025-06-16T15:29:16Z

@bors2 try jobs=x86_64-msvc-1,x86_64-msvc-2

rust-bors · 2025-06-16T15:29:20Z

⌛ Trying commit c353c7d with merge 3dadabc…

To cancel the try build, run the command @bors2 try cancel.

Use the .drectve section for exporting symbols from dlls on Windows While it would be reasonable to expect the Windows linker to handle linker args in the .drectve section identical to cli arguments, as it turns out exporting weak symbols only works when the /EXPORT is in the .drectve section, not when it is a linker argument or when a .DEF file is used. Necessary for and extracted out of #142366. Thanks `@dpaoliello` for figuring out this weird quirk of link.exe! try-job: x86_64-msvc-1 try-job: x86_64-msvc-2

dpaoliello

LGTM.

One overall design question: is it supported for folks to run the linker themselves AND specify their own def file? A couple of my changes broke the Chromium and Fuschia folks, so I'm a little paranoid now about changing anything to do with linking that may affect them.

dpaoliello · 2025-06-16T16:23:01Z

compiler/rustc_codegen_ssa/src/back/linker.rs

@@ -1116,10 +1122,6 @@ impl<'a> Linker for MsvcLinker<'a> {
            // straight to exports.
            writeln!(f, "LIBRARY")?;
            writeln!(f, "EXPORTS")?;
-            for symbol in symbols {


I don't believe that we need the def file any more, we should be able to get away with passing /DLL to the linker.

In the mingw docs I read something about the presence of a .DEF file disabling auto-export of certain symbols. I don't know if the same applies to MSVC, so I added it here just to be safe.

I don't think link.exe has "default exports" like the mingw linker does so this doesn't seem necessary but it also shouldn't hurt anything either.

Yes, this is an ld.bfd (and LLD in MinGW mode) feature to ease porting of Unix code, but it is not only .DEF that disables it. Any supported ld.bfd way to explicitly export symbols would suffice, and there is a CLI flag for that as well.

None of that matters here because link.exe doesn't export everything by default.

compiler/rustc_codegen_ssa/src/back/linker.rs

bjorn3 · 2025-06-16T16:58:23Z

One overall design question: is it supported for folks to run the linker themselves AND specify their own def file?

This PR on it's own has no effect at all on people running the linker themself. Both the .DEF file and the symbols.o file are only generated when rustc itself runs the linker.

As for a future PR that will use weak definitions, those symbols are not meant to be exported from a staticlib anyway. The only case that could break is if someone (Chromium) tries to link rlibs together into a dll and wants to export all symbols, which we don't officially support anyway. The Chromium people have had to adaot their build system for rustc changes anyway and I expect it to not all that hard to adapt to said future weak symbol usage too.

rust-bors · 2025-06-16T17:09:05Z

💔 Test failed

CI ❌

dpaoliello · 2025-06-16T17:32:23Z

💔 Test failed

CI ❌

The test executable failed with 0xc0000135 which is STATUS_DLL_NOT_FOUND

Looks like this is a test for shaking out unused functions: is the compiler supposed to do that, or is it relying on the linker?

bjorn3 · 2025-06-17T09:00:00Z

@bors2 try jobs=x86_64-msvc-1,x86_64-msvc-2

rust-bors · 2025-06-17T09:00:05Z

⌛ Trying commit 8c88ff4 with merge e97fbdf…

To cancel the try build, run the command @bors2 try cancel.

Use the .drectve section for exporting symbols from dlls on Windows While it would be reasonable to expect the Windows linker to handle linker args in the .drectve section identical to cli arguments, as it turns out exporting weak symbols only works when the /EXPORT is in the .drectve section, not when it is a linker argument or when a .DEF file is used. Necessary for and extracted out of #142366. Thanks `@dpaoliello` for figuring out this weird quirk of link.exe! try-job: x86_64-msvc-1 try-job: x86_64-msvc-2

rust-bors · 2025-06-17T11:30:12Z

☀️ Try build successful (CI)
Build commit: e97fbdf (e97fbdf63e618534ac0929eef0229de9fb477dad, parent: 55d436467c351b56253deeba209ae2553d1c243f)

bjorn3 · 2025-06-17T11:42:13Z

Squashed everything and reverted some now unnecessary changes.

@rustbot ready

wesleywiser · 2025-06-20T18:30:35Z

@bors r+

bors · 2025-06-20T18:30:38Z

📌 Commit e6aa201 has been approved by wesleywiser

It is now in the queue for this repository.

tgross35 · 2025-06-21T03:29:17Z

Should we be testing for weak symbol behavior somehow?

bjorn3 · 2025-06-21T12:30:06Z

Right. Will add a test and fix #142568 (comment) at the same time.

@bors r-

mati865 · 2025-06-21T13:54:55Z

While at it, have you confirmed that LLD behaves the same way? Or is it tested on the CI?

This is not a blocker, but it could cause a surprise later.

bjorn3 · 2025-06-21T16:59:21Z

Removed the needless .DEF file generation for MSVC and added a test with tests with both link.exe and lld.

@rustbot ready

While it would be reasonable to expect the Windows linker to handle linker args in the .drectve section identical to cli arguments, as it turns out exporting weak symbols only works when the /EXPORT is in the .drectve section, not when it is a linker argument or when a .DEF file is used.

rustbot assigned wesleywiser Jun 16, 2025

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jun 16, 2025

bjorn3 force-pushed the windows_symbols_o_export branch from e63c76b to 94ec88c Compare June 16, 2025 09:25

bjorn3 commented Jun 16, 2025

View reviewed changes

compiler/rustc_codegen_ssa/src/back/linker.rs Outdated Show resolved Hide resolved

bjorn3 commented Jun 16, 2025

View reviewed changes

compiler/rustc_codegen_ssa/src/back/linker.rs Outdated Show resolved Hide resolved

dpaoliello approved these changes Jun 16, 2025

View reviewed changes

bjorn3 force-pushed the windows_symbols_o_export branch from 8c88ff4 to fc88eae Compare June 17, 2025 09:01

bjorn3 force-pushed the windows_symbols_o_export branch from fc88eae to e6aa201 Compare June 17, 2025 11:41

This was referenced Jun 17, 2025

[WIP] Use weak aliases for the allocator shim #142366

Draft

[win][aarch64] Fix linking statics on Arm64EC, take 2 #142742

Open

wesleywiser approved these changes Jun 20, 2025

View reviewed changes

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 Jun 20, 2025

bors 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-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jun 21, 2025

bjorn3 force-pushed the windows_symbols_o_export branch 2 times, most recently from fb67d22 to 809b270 Compare June 21, 2025 16:59

rustbot 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 Jun 21, 2025

bjorn3 force-pushed the windows_symbols_o_export branch from 809b270 to d40bee8 Compare June 21, 2025 16:59

Use the .drectve section for exporting symbols from dlls on Windows #142568

Are you sure you want to change the base?

Use the .drectve section for exporting symbols from dlls on Windows #142568

Conversation

bjorn3 commented Jun 16, 2025

Uh oh!

rustbot commented Jun 16, 2025

Uh oh!

rustbot commented Jun 16, 2025

Uh oh!

Uh oh!

Uh oh!

bjorn3 commented Jun 16, 2025

Uh oh!

rust-bors bot commented Jun 16, 2025

Uh oh!

rust-bors bot commented Jun 16, 2025

Uh oh!

bjorn3 commented Jun 16, 2025

Uh oh!

rust-bors bot commented Jun 16, 2025

Uh oh!

dpaoliello left a comment

Choose a reason for hiding this comment

Uh oh!

dpaoliello Jun 16, 2025

Choose a reason for hiding this comment

Uh oh!

bjorn3 Jun 16, 2025

Choose a reason for hiding this comment

Uh oh!

wesleywiser Jun 20, 2025

Choose a reason for hiding this comment

Uh oh!

mati865 Jun 21, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

bjorn3 commented Jun 16, 2025

Uh oh!

rust-bors bot commented Jun 16, 2025

Uh oh!

dpaoliello commented Jun 16, 2025

Uh oh!

bjorn3 commented Jun 17, 2025

Uh oh!

rust-bors bot commented Jun 17, 2025

Uh oh!

rust-bors bot commented Jun 17, 2025

Uh oh!

bjorn3 commented Jun 17, 2025

Uh oh!

wesleywiser commented Jun 20, 2025

Uh oh!

bors commented Jun 20, 2025

Uh oh!

tgross35 commented Jun 21, 2025

Uh oh!

bjorn3 commented Jun 21, 2025

Uh oh!

mati865 commented Jun 21, 2025

Uh oh!

bjorn3 commented Jun 21, 2025

Uh oh!

Uh oh!