[libcxx-commits] [libcxx] [libc++][test][msan] Fix bots after #67799 (PR #73152)

via libcxx-commits libcxx-commits at lists.llvm.org
Wed Nov 22 11:34:48 PST 2023


huixie90 wrote:

> > The concern I have with disabling these tests as done in this patch is that it'll disable it for our CI as well, where these tests are running fine. And there's a lot of value in having these tests run under msan and tsan due to their nature. I can see that the tests are clearly super slow on your bot, I am just not certain what's the best way to address this without removing a lot of coverage for everyone else.
> > Unrelated question: what additional coverage do your bots provide on top of the testing we already do with sanitizers? As a general guideline, libc++ doesn't support CI setups outside of our own pre-commit CI, because it leads to these kinds of difficulties where we don't control the setup.
> 
> I really think for project like this there is more value in testing on more different environments than keeping a particular test alive, as you will get feedback from other platforms yearly.
> 
> I myself rely appreciate when someone runs our sanitizers on bots with some platforms I have no access. Even I have no control over them, it's much easy resolve regressions soon after commit than after it gets into release.
> 
> About coverage:
> 
> 1. we have aarch64 asan/msan/hwasan.
> 2. clang is build from the tree one the same revision - we want to know when llvm changes causing sanitizers misbehave
> 3. Today our msan runs with default set flags, but usually we opt-in our bots into new features before we enabled them by default.
> 
> Regarding particular test, it claims clang is broken and XFAIL most of them any way. It's just XFAIL on timeout is not nice.

maybe msan on aarch64 is bit different? 

https://github.com/llvm/llvm-project/pull/73152


More information about the libcxx-commits mailing list