ayokunle321 wrote: @erichkeane thanks a lot for the clarification. I think I have a hang of it now- I was kinda confused before. Now, my approach is to look through DiagnosticXKinds.td files -> check for a `select` with a significant number of items -> check its diagnostic ID -> grep the lib directory with this ID and find the uses of the diagnostic -> see if it’s up for replacement (caller has a lot of magic numbers/an enum made just for it).
Now I wanted to ask - what do you consider a lot - 3 or more uses with magic numbers? For a `select` with a significant number of items, I’m thinking of going with 4 or more items as it seems reasonable. I was also looking through the codebase using this heuristic exposed a lot more for replacement (rather than 5 which I had in mind before). Also could you just brief me through the process of a change. I know the .td files are used to generate header files. So do I have to change the select to an enum_select first in the .td files and then build and then reference the generated enum option in the caller (e.g. diag::MemClassWork::AliasDecl) and then rebuild. The thing about this is that the build would probably fail the first time as the callers will still make use of the magic numbers. I also found an enum declared just for a select that was already in a header file. For this case, I just remove this declaration right? And then replace the select. I guess I’m just confused about the whole header generation process - where the new enums would live since the relevant header files already exist before a build (e.g Sema.h, Decl.h) and since we’re making changes to .td files which are the header generators. Kinda new to clang stuff so forgive me for these questions 😅 https://github.com/llvm/llvm-project/pull/130868 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
