[cfe-commits] r143792 - in /cfe/trunk: lib/Headers/avxintrin.h lib/Headers/emmintrin.h lib/Headers/tmmintrin.h lib/Headers/xmmintrin.h test/Sema/x86-builtin-palignr.c

[Bob Wilson]

On Nov 5, 2011, at 11:05 AM, David Blaikie wrote:

>> +   // FIXME: The "incompatible result type" error is due to pr10112 and should
>> +   // be removed when that is fixed.
>> +   return _mm_alignr_pi8(a, b, c); // expected-error {{argument to '__builtin_ia32_palignr' must be a constant integer}} expected-error {{incompatible result type}}
> 
> Strangely, I'm seeing this fail (on my Ubuntu box, if that's relevant)
> - the second error (the one you have the fixme about) isn't firing:
> 
> $ /home/dblaikie/Development/llvm/clean/build/x86-64/Debug+Asserts/llvm/Debug+Asserts/bin/clang
> -cc1 -internal-nosysroot-isystem
> /home/dblaikie/Development/llvm/clean/build/x86-64/Debug+Asserts/llvm/Debug+Asserts/lib/clang/3.0/include
> -ffreestanding -fsyntax-only -target-feature +ssse3 -target-feature
> +mmx -verify -triple i686-apple-darwin10
> /home/dblaikie/Development/llvm/clean/llvm/tools/clang/test/Sema/x86-builtin-palignr.c
> error: 'error' diagnostics expected but not seen:
>  Line 9: incompatible result type
> 1 error generated.
> 
> But the bots don't seem to be complaining, so I assume it's not a
> widespread issue. (perhaps I've done something wrong)

That's very strange.  You don't happen to have a (partial) fix for pr10112 in the sources, do you?  ;-)
Unless you can help figure out what's going on, I'm really not sure what to do.  It's a bit scary to think that there may be some undefined behavior involved here.  The best thing would be to get pr10112 fixed.