Greg Bedwell <[email protected]> writes: > Thanks for the review. > > > Looks good with one little nit, below. Do you need me to commit this > for you? > > I should be fine to commit myself, thanks! > > > -// %clang -E -dependency-file bla -MT %t -MP -o %t -x c /dev/null > > +// RUN: not %clang_cc1 -E -dependency-file bla -MT %t -MP -o %t -x c / > dev/null 2>&1 | FileCheck %s > > This could use -verify instead of FileCheck. > > I previously tried this approach, but ran into an issue that -verify doesn't > seem to work in this case with both -E and -o on the command line. I could > get rid of one or both of them, but then I get no errors at all so I'd need // > expected-no-diagnostics. However, without any visibility of the original bug > report I'm assuming that these are both prerequisites for the original crash > this was supposed to test. Unfortunately, my attempts to go and build a > revision prior to the bugfix (April 2011) are met with failure so I can't try > and reconstruct a test from that. > > Bearing this in mind, it seemed prudent to just use FileCheck instead and try > to keep the test as unchanged as possible. Sorry, I should have explained > this in my original review summary! > > Can you think of anything else I can try, or should I commit as-is?
Thanks for the explanation. This is fine as is. Please commit. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
