We should arguably have always been doing that. The state of libunwind
is quite sad, so this commit adds several XFAILs to make the CI pass.
We need to investigate why so many tests are not passing in some
configurations, but I'll defer that to folks who actually work on
libunwind for lack of bandwidth.
Details
- Reviewers
ldionne - Group Reviewers
Restricted Project Restricted Project - Commits
- rG60fe1f59d08b: [runtimes][ci] Run the tests for libunwind in the CI
Diff Detail
- Repository
- rG LLVM Github Monorepo
- Build Status
Buildable 126620 Build 183963: arc lint + arc unit
Event Timeline
Note that I suspect this won't work out-of-the-box cause the libunwind tests might be broken in a few ways (and that's why we haven't been running those tests from the start), however let's see what the CI tells us.
libcxx/utils/ci/run-buildbot | ||
---|---|---|
92 | Isn't -DLIBCXXABI_USE_LLVM_UNWINDER=Y needed also to use the in tree version of libunwind? |
Add a bunch of XFAILs. This is really bad, we have like 8 tests and more than half
of them are failing even in some common configurations. We'll need some serious
investigation.
I am pretty sure that all the tests used to pass on macOS a couple years ago, but apparently a lot of changes have landed without proper CI and testing, so now they're red.
Pinging some people who have touched libunwind lately: @vitalybuka @MaskRay @danielkiss. I'm not sure who's closest to being a libunwind code owner, but it would be awesome if we could get a grip over the project. I'd do it, but libc++ and libc++abi are already borderline too much. I'll add this CI to avoid introducing new regressions, at least that's one step forward.
Isn't -DLIBCXXABI_USE_LLVM_UNWINDER=Y needed also to use the in tree version of libunwind?