alexfh added a comment. In http://reviews.llvm.org/D15797#317793, @mattsta wrote:
> It's difficult to track down *why* the invalid locations are happening > because by the time we get to an invalid location, all source location > information is lost. The best I've been able to come up with is fixing the > early return conditions (which were previously impossible to reach due to > asserts catching the invalid conditions first). > > It's easy to have these happen on any large codebase (e.g. try to recursively > clang-tidy the erlang source tree with readability-braces-around-statements). I'd still like to add a test for this (and I have a different idea on how to fix this, namely return an invalid location from `findRParenLoc` instead of asserting). If you can't come up with a test case, I'll try to find one myself by running on some files, but we absolutely have to add a test to avoid regressions when this code is changed again. Repository: rL LLVM http://reviews.llvm.org/D15797 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits