This is an archive of the discontinued LLVM Phabricator instance.

Revert "[ThinLTO][test] Add visibility related tests"
ClosedPublic

Authored by bgraur on Dec 31 2020, 3:36 AM.

Details

Summary

Both newly added tests fail in Release.

This reverts commit 52aa4e210744361a5ed6dc50fef78ed91706e508.

Diff Detail

Event Timeline

bgraur created this revision.Dec 31 2020, 3:36 AM
bgraur requested review of this revision.Dec 31 2020, 3:36 AM
Herald added a project: Restricted Project. · View Herald TranscriptDec 31 2020, 3:36 AM
SureYeaah accepted this revision.Dec 31 2020, 3:37 AM
This revision is now accepted and ready to land.Dec 31 2020, 3:37 AM
This revision was landed with ongoing or failed builds.Dec 31 2020, 3:45 AM
This revision was automatically updated to reflect the committed changes.

How did it break tests? It tested well on Linux/macOS/Windows.

I am pretty sure this is a misfire.

This was due to a -DLLVM_ENABLE_ASSERTIONS=off vs -DLLVM_ENABLE_ASSERTIONS=on difference. llvm-lto2 needs -enable-import-metadata

Ideally, this roll-forward should also have included this fix.
The test failure was 100% reproducible.

Ideally, this roll-forward should also have included this fix.
The test failure was 100% reproducible.

I have checked some Linux/macOS/Windows built bots and even tested on a ppc64le machine. The description of this commit isn't clear about what happened.

The descritption is: "Both newly added tests fail in Release."

That means:
1/ Build llvm in Release configuration.
2/ Execute the tests and they will fail.

Checking any build bots won't show any failure because the problem is a test failure, not a build breakage.

Most build bots are configured with -DLLVM_ENABLE_ASSERTIONS=ON. Anyhow I figured out the root cause and make the behavior not dependent on -DLLVM_ENABLE_ASSERTIONS= to prevent future issues like this.

lebedev.ri added a subscriber: lebedev.ri.EditedDec 31 2020, 10:11 AM

The descritption is: "Both newly added tests fail in Release."

That means:
1/ Build llvm in Release configuration.
2/ Execute the tests and they will fail.

Checking any build bots won't show any failure because the problem is a test failure, not a build breakage.

Are you suggesting that buildbots don't actually run tests?

Not sure how the build bots are configured.

What is sure is that if you run the tests using the llvm tools these tests will always fail in Release.

So, the conclusion is one of the following:
1/ the bots didn't run your tests.
2/ the bots have a different run configuration which hides the failures.

llvm/test/ThinLTO/X86/visibility-elf.ll