brettw added inline comments.
================ Comment at: clang-tools-extra/clang-doc/BitcodeReader.cpp:336 Reference *I, FieldId &F) { switch (ID) { case REFERENCE_USR: ---------------- haowei wrote: > This block ID came from L582, which read from a bitcode file when > `BI_REFERENCE_BLOCK_ID` is encountered. I am not familiar with the content of > an arbitrary bitcode file. But after your patch, if this block ID is > encountered, it will result in an error (a behavior change). > > It looks like `REFERENCE_IS_IN_GLOBAL_NAMESPACE` is only defined in in > clang-doc and not part of LLVM's bitcode header. So it looks like it is not a > standardized ID to me. Could you confirm it? > > If `REFERENCE_IS_IN_GLOBAL_NAMESPACE` only presents in bitcode file generated > from BitcodeWriter from clang-doc and we are not expecting to provide > compatibility to an arbitrary bitcode generated from elsewhere, I am OK with > deleting this field (I still don't think this is the right approach to do but > I don't have a strong opinion on this either). Correct, this whole bitcode system is used only to communicate within clang-doc between threads (!) (I think it should be entirely removed but I don't have the time for that). It is never serialized to a file and is not shared with any other component. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D134235/new/ https://reviews.llvm.org/D134235 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits