labath added a comment. celebration_balloons <https://reviews.llvm.org/file/data/dpeiswv2hxzncud4h55k/PHID-FILE-jd5ar6gmeeghghpcfmi3/meme-celebration_balloons.gif>
I like this a lot. Incidentally, I believe the main reason these tests don't work on non-darwin platforms is because their system libraries are not modularized. If you make the gmodules tests self-contained (which I would recommend, for the sake of reproducibility, anyway), then I think a lot of these tests could run elsewhere as well. In D134344#3805270 <https://reviews.llvm.org/D134344#3805270>, @Michael137 wrote: > An alternative would be to keep the `debug_info` category and simply flag it > as not to be run by default unless specified explicitly in the test. That way > we wouldn't need the `MAKE_GMODULES` in the Makefile and the special > `MAKE_DSYM` rule in https://reviews.llvm.org/D134345 Avoiding the MAKE_GMODULES repetition would definitely be nice, but I might try(*) do it slightly differently: - keep `gmodules` as a category, but not a *debug info* category. Among other things this enables running all gmodules tests with the `--category gmodules` flag. - teach the debug info replication to ignore tests with the gmodules category (just like it does for `@no_debug_info_test_case` tests). This step wouldn't be necessary if we made debug info replication opt-in instead of opt-out, as discussed on one of the previous patches (@JDevlieghere might remember which one it was) - teach `buildDefault` to build with gmodules enabled in case the test is annotated with the gmodules category. (*) I'm not entirely sure how this would work out, but I think it should be fine. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D134344/new/ https://reviews.llvm.org/D134344 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
