arsenm added a comment. I'm growing to dislike the new behavior even more as I use it. I don't think restricting the dump to CHECK-LABEL prefixes will even help. The amount of context printed previously was good enough for development use. If I ever can't figure it out from the diff, I want to look at the output completely separate from the terminal. This is adding a lot of noise that doesn't help show me the failure point, and often fills my terminal scrollback buffer.
For AMDGPU test cases for example, every function has a long chunk of metadata that is irrelevant to the majority of tests. This floods my terminal even if a single function failed between check labels. I think this should revert to the old behavior. I think dump full on input would only be a reasonable default on build bots (where even then it might need some restrictions to avoid gigantic build logs if a really broken patch were committed). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D81422/new/ https://reviews.llvm.org/D81422 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
