[llvm-commits] [llvm-gcc-4.2] r81788 - /llvm-gcc-4.2/trunk/gcc/testsuite/g++.apple/eh1.C

[Chris Lattner]

On Sep 14, 2009, at 2:58 PM, Mike Stump wrote:

> On Sep 14, 2009, at 2:39 PM, Eric Christopher wrote:
>> Mmm.. either way. I'm happy to revert as it's a bug by that
>> definition :)
>
> Reversion is best I think as I suspect there are many failures in the
> testsuite.

Yes, I agree with mike, reverting is best IMO.  If we were going from  
1 failure to 0, I'd be ok with this, if we're going from 1000 failures  
to 999, it is best to leave it failing.

Adding a comment to the testcase with a radar number pointing out the  
llvm issue would be a good idea though so that future people analyzing  
the issue won't redo the work you've done.

-Chris

>  The thing being tested is actually a feature of gcc.
> Testing for a non-feature, would be odd.  If you wanted to test
> results to be cleaner, you can mark it as expected.  I can dig up the
> older radar number for when we switched gcc to text, if you want
> (please don't ask though).
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits