On 21 March 2017 at 09:17, Jakub Jelinek <ja...@redhat.com> wrote: > I've tested in the mean time the following patch with both gcc from yesterday > where > pr35878_3.C fails as expected, and with the latest cc1plus where > it succeeds both with -m32 and -m64. Scanning the tree dump has the > advantage that you test it everywhere, it works even with -masm=intel etc.
Works for me. I can't write a tree scan to save my life, hence the scan-assembler approach taken in my patch. Not that I can really grok assembler either much.
