This is an archive of the discontinued LLVM Phabricator instance.

[WebAssembly][NFC] simplify SjLj inline assembly test
ClosedPublic

Authored by quantum on Jul 2 2019, 6:22 PM.

Details

Summary

Per feedback in D64115, simplify the test.

hidden is left in though, because every test in the file has it.

Diff Detail

Repository
rL LLVM

Event Timeline

quantum created this revision.Jul 2 2019, 6:22 PM
aheejin accepted this revision.Jul 2 2019, 6:27 PM

Hmm right, every other function has hidden... while you're on a roll, can you remove all hiddens here? :) Thanks!

This revision is now accepted and ready to land.Jul 2 2019, 6:27 PM
aheejin added inline comments.Jul 2 2019, 6:29 PM
llvm/test/CodeGen/WebAssembly/lower-em-sjlj.ll
198 ↗(On Diff #207686)

Oh one more thing. I think it's better to do
CHECK-NOT: invoke void blah blah
so that what this test is looking for. And maybe add a one-line comment that inline asms should NOT be transformed into invoke wrappers.

aheejin added inline comments.Jul 2 2019, 6:35 PM
llvm/test/CodeGen/WebAssembly/lower-em-sjlj.ll
198 ↗(On Diff #207686)

so that what this test is looking for -> so that it is clear what this is looking for

quantum marked an inline comment as done.Jul 3 2019, 10:31 AM
quantum added inline comments.
llvm/test/CodeGen/WebAssembly/lower-em-sjlj.ll
198 ↗(On Diff #207686)

If it generated an invoke wrapper, opt would exit with a failure. I'll put the CHECK-NOT in anyways to make it more explicit.

quantum updated this revision to Diff 207838.Jul 3 2019, 10:34 AM

Add comments and CHECK-NOT: __invoke_void

This revision was automatically updated to reflect the committed changes.

Hmm right, every other function has hidden... while you're on a roll, can you remove all hiddens here? :) Thanks!

Friendly ping. Next time, if you don't want to make some changes I suggested, could you please not just ignore it but rather let me know why you don't want that? This is really nothing, so I'm not asking you to upload this in a separate patch or anything so nevermind about this, but I noticed this pattern a few times now, so I'd just mention this. Thanks for the bugfix again.

@aheejin Sorry, did not see that comment.