AaronBallman wrote:

> LGTM with minor nits; however, we still have the open question of whether 
> this should be a CC1-only option.
> 
> @MaskRay, do you have an opinion on whether this should be exposed via a 
> Clang driver option (with or without a negative form)?

CC @rnk @zmodem as other folks interested in Windows behaviors.

> I think this makes sense to leave as a `-cc1`-only option for now. It's 
> always easy to promote this to a driver option, but pretty much impossible to 
> remove once it's in the driver.

FWIW, I'm starting to feel pretty strongly about this not being a cc1-only 
flag. We do not want to condition users to do `-Xclang -some-option` for 
controlling language dialects.  So to me, if this is important enough for users 
to be able to control explicitly, it should be a driver flag. (The situation 
where I think a cc1-only flag would be a good idea is if your downstream driver 
was going to enable this automatically for users so the user doesn't have to do 
the `-Xclang` dance, but I don't have the impression that's the case here.)

I looked through Microsoft's documentation and could not find a `/Zc` or other 
flag to control this behavior in cl, so I don't think a driver option would 
require any changes for clang-cl.

https://github.com/llvm/llvm-project/pull/176551
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to