iains added a comment. I have no comment the patch itself, that seems fine, I even agree that this could be convenient :-)
However, I am concerned that this now means that '-fprebuilt-module-path' has meaning for both clang modules and c++20 modules - which is something we agree is generally undesirable (AFAIU, we agree it would be better to have orthogonal command line switch sets). At some stage, we need to abstract the general mapping between module names and BMI names; I think the ownership of that mapping needs to be in the module loader (or, at least, available when the compiler instance has no pre-processor). This is because when building a pre-processed source that imports a module we could still need to carry out the module-name to BMI name translation (the alternative seems bad, which would be to bake the BMI path in the pre-processed output). So, I'm not 100% convinced about lookup schemes for BMI names that depend on header search being available. maybe other modules folks have a view? CHANGES SINCE LAST ACTION https://reviews.llvm.org/D120874/new/ https://reviews.llvm.org/D120874 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
