ldionne wrote: Ack, thanks for the confirmation.
> If we wanted to send the report as a comment on GitHub, we can certainly look > into that as mentioned here > https://discourse.llvm.org/t/using-plugins-in-buildkite-ci-that-require-docker/82701/6?u=davidspickett. > The Buildkite specific parts of the script are easy to remove. I'd rather not do that. Commenting on the PR should be done very rarely, especially for CI failures, since it clutters the PR and Github already has a builtin system for presenting CI failures. Overall, I'm fine with this patch, however I wonder how much effort it is worth putting into the BuildKite infrastructure given that we plan on moving as much as possible to Github Actions. With Github actions, the `.ci/generate-buildkite-pipeline-premerge` script would probably go away entirely in favour of individual workflows or something along those lines, and then the issue of figuring out what failed would probably be a non-issue. I am especially unsure about adding functionality to fundamental utilities like `Lit` in that context: if we e.g. end up not needing to output test results in individual files a few weeks/months from now, then there may not be any users of `--use-unique-output-file-name` anymore and then we'll have increased the complexity of Lit for a temporary benefit. Just my .02. I really appreciate efforts like this to make the CI infra easier to use, I just wonder if the specific approach taken by this patch series is in line with the near future direction of the project w.r.t. CI. https://github.com/llvm/llvm-project/pull/113447 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
