We have been noticing issues with the lldb bots on builds using versions below clang 14 and dwarf 2, so to make sure we can get clean builds for a while, we are disabling the tests for those versions
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
From a pragmatic point of view this LGTM. It would be great if someone with an interest in lldb server could look into how to make this test more reliable. Thanks!
What's the failure mode?
I find it highly unlikely that compiler or dwarf info version has anything to do with the failures. lldb-server tests don't read debug info at all and they might just as well be compiled with -g0.
The only unusual thing I can think of which makes these tests different from all other lldb-server tests is that there is so many of them in the same file. Are you sure you're not just running into some per-test timeout limit?
Also, if you're running these tests on some kind of a matrix bot, maybe you can just skip all lldb-server tests completely (dotest.py --skip-categories llgs). Given that they don't use debug info, they definitely don't provide any useful signal in these contexts.