emaxx-google wrote: An update: The miminization of the reproducer is sadly still ongoing. A longer version of this TLDR:
--- As the reduction was making slow progress, I've tried the proposal from @vgvassilev to add the "successfully compiles in the non-modular mode" criterion to the interestingness test. It did let the reduction finish in just a couple of days, but the result turned out to suffer from the same problems as before: lots of broken code and commands, scattered through the files not "reachable" in non-modular builds (for headers only mentioned in cppmaps but not included directly or transitively). Hence I returned back to the idea with minimizing without `-fallow-pcm-with-compiler-errors`. I improved a few reduction heuristics, but also the progress stalled a few times due to various issues (disk space, OOMs and compiler hangs, bugs in the new heuristics). It looks like it's a very tough input for the C-Reduce-like minimization, but unfortunately it's the only viable approach we have at the moment. I'll provide an update here once I have something to share (likely in a few days). Thanks for the patience! https://github.com/llvm/llvm-project/pull/133057 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
