Unless we actually have a compatibility reason to expose the implicit `_GUID_*` declarations that MSVC appears to inject, I'd prefer that we model these as we do for other type information objects: do not expose them to semantic analysis, and treat the `_GUID_` prefix and `@@3U__s_GUID` suffix as just being the way these implicit objects happen to be mangled (that is, what the patch does now). The implicit type and variables smell to me like an accident of the MSVC implementation rather than something we need to emulate.
http://llvm-reviews.chandlerc.com/D1357 _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
