https://github.com/frobtech commented:

I had originally expected to see this change before the caller side because it 
seemed to me like it would be the simpler one and would set the stage on the 
normalization questions.  But all the nuances about the normalization really 
span both changes and just need to be addressed (or noted as expected 
limitations, whatever) across the pair together one way or another.  It's hairy 
enough to merit some detailed commentary.  But it's niche enough that it's not 
clear there need to be any "good" answers to all those nuances.  To a first 
approximation, they can all just be part of the reason why everyone should just 
be encouraged to conform to C23 / C++ requirements and always use prototypes.  
The specific caveats for using these metadata / analysis features just need to 
be clearly noted if users will be allowed to mix troublesome code patterns 
(lack of prototype declarations for all function pointer types) with the new 
features that have caveats specific to those code patterns.

https://github.com/llvm/llvm-project/pull/212863
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to