On Tue, 14 Aug 2007, Andy Lester wrote:
> On Aug 14, 2007, at 6:27 PM, Andy Dougherty wrote:
>
> > see above. It would have been nice if gcc simply rejected the attribute
> > so that testing the exit status of gcc would be sufficient. gcc-3.4
> > doesn't have this problem. The test could either be made to only work
> > for gcc > 3.3, or it could try parsing the gcc-3.3 warning output (being
> > careful to handle localization issues appropriately, and avoiding false
> > positives for unrelated warnings)
>
> The attribute sniffer looks at the compiler's output and says if it sees
> "warning" or "error", that the attribute must not work.
Ah yes, I see that you've added that. I didn't see any discussion about
why such an approach was needed. It's worthwhile to record such things so
that future maintainers don't have to rediscover why you had to work so
hard or worry about breaking it!
--
Andy Dougherty [EMAIL PROTECTED]
