jdoerfert marked 2 inline comments as done. jdoerfert added inline comments.
BRANCHES master /clang/test/OpenMP/PR44893.c:3 > Can you add some sort of positive test, related to where the crash was? I did, in deduplication.ll. > If this was a debug info related crash, maybe you can test that the debug > info IR is reasonable (@aprantl may be able to help). It was, kind of. With debug info enabled (-g) OpenMP code generation does some stuff that it does not without. It has actually nothing to do with debug info per se but with the location of the `struct ident_t`. > If not, please remove the -g from the command-line, and just test that the > normal IR is reasonable. I have the IR as a test (deduplication.ll). Dropping -g here does not make sense, if it is a problem I would drop the entire test instead. /clang/test/OpenMP/PR44893.c:1 I don't think I can, or at least I did not figure out how without loosing the desired effect (see below). Is it bad to invoke clang here? I actually did a `grep` of `%clang` and found other tests doing it so I figured it was fine. What I "need" is -g and -fopenmp, if I can tell that to cc1 I'm happy to change it. Users: jdoerfert (Author) https://reviews.llvm.org/rG396b7253944e _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
