Dummy PR to test coverage test changes in CI

[Zalathar]

The x86_64-gnu-llvm-15 job that normally runs on PRs doesn't include the profiler runtime, so most of the coverage tests don't get run on PRs.

(That's why #114875 didn't fail until an actual merge was attempted.)

r? @ghost

[Zalathar]

@rustbot experimental

[Zalathar]

@rustbot author

[Zalathar]

Success!

https://github.com/rust-lang/rust/actions/runs/5885191281

[Zalathar]

Now testing #114843 in CI on Linux/Windows/macOS, to ensure I don't get any nasty surprises after it's reviewed.

[Zalathar]

Success:

• x86_64-gnu
• x86_64-apple-1
• x86_64-msvc

[bors]

☔ The latest upstream changes (presumably #115183) made this pull request unmergeable. Please resolve the merge conflicts.

[Zalathar]

Testing #114843 on Linux/Windows/macOS (coverage-map and run-coverage only), to see if everything still works on the big 3 configurations.

[Zalathar]

Those Linux failures look like the same ones I saw on the aarch64-gnu build, so hopefully that's a platform issue and I don't have to worry about ARM Linux specifically.

I'm going to try another run of Windows only, to flush out any unknown issues on that platform.

[Zalathar]

Test failures on Windows seem to be the same as the Linux ones, which is surprising.

Perhaps there is some common property of the main Linux/Windows CI environments that is not shared by x86_64-gnu-llvm-15 or my local Aarch64 macOS environment.

[Zalathar]

I'll try a CI macOS build to double check all the major platforms.

[Zalathar]

Hmm, so the overall results so far seem to be:

Failure (probably the same failures, though I haven't fully verified this):

• aarch64-gnu
• x86_64-gnu
• x86_64-msvc
• x86_64-apple-1

Success:

• x86_64-gnu-llvm-15
• Local build on aarch64 macOS
• Local build on aarch64 macOS (--stage=2)

[Zalathar]

Trying another build with the default PR CI config (x86_64-gnu-llvm-15) just to confirm again that that one still works.

[Zalathar]

Testing some details of #119034 in advance.

[Zalathar]

As expected, the test does fail on Windows, so it's correct to // ignore-windows.

[bors]

☔ The latest upstream changes (presumably #119621) made this pull request unmergeable. Please resolve the merge conflicts.

[Zalathar]

Testing out a port of tests/run-make/coverage-llvmir to tests/codegen.

[bors]

☔ The latest upstream changes (presumably #120881) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #121904) made this pull request unmergeable. Please resolve the merge conflicts.

[Zalathar]

Testing whether I can remove //@ ignore-windows from tests/mir-opt/instrument_coverage.rs.

[Zalathar]

Looks like the test passed on Windows: https://github.com/rust-lang/rust/actions/runs/8295637584/job/22703073495?pr=114917

Might as well try mingw as well, since that supposedly has problems with the profiler runtime.

[Zalathar]

x86_64-mingw seems fine too: https://github.com/rust-lang/rust/actions/runs/8303600151/job/22727987597?pr=114917

[Zalathar]

OK, coverage-run tests are still broken on windows-gnu.

I think the reason tests/mir-opt/instrument_coverage.rs worked is that it doesn't actually run anything, so it doesn't care about having a broken profiler_runtime.

This suggests that maybe we should stop building the profiler runtime on windows-gnu, and compensate by migrating various existing tests from //@ needs-profiler-support to //@ compile-flags: -Zno-profiler-runtime as appropriate.

[bors]

☔ The latest upstream changes (presumably #122690) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #123451) made this pull request unmergeable. Please resolve the merge conflicts.