alexfh wrote: > Interesting, changing `private header "wrap_foo.h"` to `header "wrap_foo.h"` > in `b.wrap_foo` stops reproducing the error. I'm not sure it guarantees > > > whenever resolving module for the header, pick one that has the header as > > modular over textual. > > But seems we are closer to this behavior than I expected.
@vsapsai I guess, it's a good sign? :) Do you see how our use case can be supported by a trivial and low-risk forward fix? If not, I'd insist on a revert before we can figure out the way forward. We can run this sort of a change through our testing before it relands, and ensure it doesn't break our code. We did this on multiple occasions before (especially for changes that touch Clang header modules). https://github.com/llvm/llvm-project/pull/138227 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
