Make the api, shell and unit tests independent lit test suites. This allows us to specify different dependencies and skip rebuilding all the unit test (which is particularly expensive) when running check-lldb-api or check-lldb-shell. This does not change the autogenerated targets such as check-lldb-shell-driver or the top level check-lldb target, which all continue to work as before.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Just gonna post a comment to point out that this apparently adds all tests twice to check-all which makes the tests go kaputt (but I only messaged Jonas directly about that, hence the lack of detailed errors here).
Oh, I have a vague recollection of this happening when he tried to split the targets before.
Clean builds for all the test targets work now and this also looks good, so let's ship it.
(also thanks for working on that, the unit test builds where a big annoyance when testing API tests)
Thanks, the error message suggests the tests are run twice and the log shows every test failing twice. Let me try adding EXCLUDE_FROM_CHECK_ALL to add_lit_testsuites as well. It didn't change anything on macOS, but maybe it does on Windows. If that doesn't fix it I'll revert to turn the bot green again over the weekend.