On Feb 16, 2009, at 11:15 AM, Ben Laurie wrote: > This fixes a currently failing test by using the > RangeConstraintManager.
These aren't broken tests; their just testing expected behavior. The expected behavior is that BasicConstraintManager doesn't handle these cases well. > > Index: test/Analysis/null-deref-ps.c > =================================================================== > --- test/Analysis/null-deref-ps.c (revision 64627) > +++ test/Analysis/null-deref-ps.c (working copy) > @@ -1,6 +1,6 @@ > -// RUN: clang -analyze -std=gnu99 -checker-simple -verify %s && > -// RUN: clang -analyze -std=gnu99 -checker-simple > -analyzer-store-region -analyzer-purge-dead=false -verify %s && > -// RUN: clang -analyze -std=gnu99 -checker-cfref > -analyzer-store-region -verify %s > +// RUN: clang -analyze -analyzer-range-constraints -std=gnu99 > -checker-simple -verify %s && > +// RUN: clang -analyze -analyzer-range-constraints -std=gnu99 > -checker-simple -analyzer-store-region -analyzer-purge-dead=false > -verify %s && > +// RUN: clang -analyze -analyzer-range-constraints -std=gnu99 > -checker-cfref -analyzer-store-region -verify %s Please don't remove the RUN lines that invoke BasicConstraintManager. We still wanted that tested. As we discussed via IM, please add the separate RUN lines and #ifdef logic for the alternate warnings (some comments in the test case are also good to explain the divergent behavior). _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
