nikic wrote: > I'm also experimenting with comparing the results between Clang and the ABI > library again, currently with this patch: > https://gist.github.com/nikic/7fd434e8b427a58ed532cc3a3b19c9b1 It only checks > that the Kind is the same. I'll create a PR for the validation when I have > time to finish it.
PR is up at https://github.com/llvm/llvm-project/pull/203281, which I think should validate all the cases that are relevant to the current implementation. There are ~60 test failures with the validation + ABI library force enabled. The most concerning issue I saw was mismatches in IndirectByVal. I think some of the issues were already flagged by review comments above, e.g. the issue with x86_fp80 size. Should try to fix at least most of these failures (I think it's okay to land if only some weird edge cases are left). Can also add `-fexperimental-abi-lowering` RUN lines to some of those currently failing tests once they are fixed. (We probably don't want `-fexperimental-abi-lowering` on *all* X86 tests because there's so many...) https://github.com/llvm/llvm-project/pull/194718 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
