andykaylor wrote:

>  This test expects the warning message `warning: overriding 
> '-fcomplex-arithmetic=basic' option with '-fcomplex-arithmetic=promoted'` for 
> `clang -ffast-math -ffp-model=fast`. I'm not sure if this test is correct, 
> but I believe a new variable is necessary to avoid affecting it. If we stored 
> the history of non-Gcc options in the existing `GccComplexRangeOption`, the 
> above warning message would become `overriding '-ffast-math' option with 
> '-ffp-model=fast'`. Should I also fix all warnings related to complex range 
> options other than `-fno-fast-math`?

I've never been happy with the way we handle the warnings where aggregate flags 
change the complex range setting. For instance, in the case you cite, `warning: 
overriding '-fcomplex-arithmetic=basic' option with 
'-fcomplex-arithmetic=promoted'`, that has to be confusing to a user who has 
used neither '-fcomplex-arithmetic=basic' nor '-fcomplex-arithmetic=promoted'. 
In a perfect world, the message would say `warning: '-fp-model=fast' sets 
complex range to "promoted" overriding the setting of "basic" that was implied 
by '-ffast-math'`. I'm just not sure how much logic we'd need to construct such 
a message with reasonable grammar in all cases.

I think we just need to commit to each change making progress relative to the 
previous state rather than trying to fix everything.

I'll give your changes a closer review tomorrow, but I think you're on the 
right track.

https://github.com/llvm/llvm-project/pull/132680
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to