dblaikie added a comment. In D106615#2950964 <https://reviews.llvm.org/D106615#2950964>, @yonghong-song wrote:
> The attributes should be revolved during semantic analysis stage. So looks > like replace-style attribute setting is not really needed. I will change to > add attribute parameter during creation time. Thanks for suggestion. I think semantic analysis runs alongside code generation - eg, if you have one attribute applied to a declaration, then a function definition using that type, then another declaration of the same type, but with a different attribute, then another function definition using the type - if you build the type only once, on the first use, you might not capture all the attributes? Anyway, yeah, if all this works out without the replacement style API, I /think/ that's probably the better call. Might be worth including a test like that to show how it's expected to be handled. (do these attributes need to work on declared-but-not-defined types? (eg: a type declared, and then a function that takes pointer-to that type, but the type is never defined in this translation unit) If so, could you include testing. for that situation?) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D106615/new/ https://reviews.llvm.org/D106615 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
