I don't see anything attached to your email, but I don't think crashes aren't required for a test case for this. The behavior I was thinking of testing is that the old code would fail to set the operand location properly for the type attribute, and the new case does set it. So any case which relies on that location being set to something valid would suffice.
~Aaron On Sun, Mar 9, 2014 at 11:55 AM, Logan Chien <[email protected]> wrote: > > If you are interested in, here's reduced test case. It was reduced from > the Android jni.h. I feel this test is not good enough because clang will > not crash if you delete some of the lines in crash_pch.h. > > {F47242} > > http://llvm-reviews.chandlerc.com/D3020 _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
