>I'm saying we might want to verify that these do in fact fail. UNSUPPORTED
>causes the test to not be run, XFAIL tests to ensure failure, which is
>possibly what we want here. 
>OTOH, our test configuration isn't set up to do
> conditional _compile time_ failures, so maybe we should just settle for 
> UNSUPPORTED.

Your right. That's probably a good idea. <atomic> is unique in that it actually 
explicitly errors when it is included but won't work.
`.fail.cpp` tests are expected to fail at compile time. I think we should do is 
leave these tests as unsupported and add a `cxx03.fail.cpp` test 
that ensures including <atomic> results in a compiler failures. 

Here's my gripe. I like the separation between
1. XFAIL - tests that should be fixed.
2. UNSUPPORTED - tests I don't care about at all.

It makes it a lot easier to watch just the tests I care about. Adding one XFAIL 
test seems a touch cleaner, and will guarantee the same thing.

http://reviews.llvm.org/D5955



_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to