[cfe-commits] [PATCH] Emit error for unknown warning options

[Chandler Carruth]

On Tue, May 10, 2011 at 9:29 PM, Eli Friedman <eli.friedman at gmail.com>wrote:

> See PR9701.  Essentially, the issue is that if we don't emit an error
> for unknown warning options, there isn't any supported way for
> configure scripts to check whether clang supports a given warning
> option, which leads to a bunch of useless warnings about unknown
> warning options.  Given that gcc emits an error for unknown warning
> options,  and we've at least stubbed out all the common warning
> options, I think changing the default behavior like this makes sense.
>

The fact that we have warning options stubbed out but not implemented means
that merely passing flags doesn't guarantee a warning will fire, and it
would be annoying to force further proliferation of stubbed out warning
options as GCC and potentially other compilers we want general commandline
compatibility with change and evolve their warnings.

It's not clear to me why configure scripts if they really care can't pass
-Werror?

>From the other perspective, based on the stubbed out flags we already have,
maybe we should make this default ignore.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110510/2540162c/attachment.html>