On 05/20/2015 05:01 PM, Martin Sebor wrote:
Now that I know a bit more from Jakub & Yuri's comments, I don't think
I opened compiler_rt/23600 and attached the sanitizer patch
to it:
http://llvm.org/bugs/show_bug.cgi?id=23600
If there are no objections to the bug I'll post the patch for
review next.
Thanks. Seems to be moving along. Looks like they were a bit surprised
that you'd included a patch ;-)
So I'll ask the question(s) in a slightly different way and see if that
guides us one direction or another.
Do the libbacktrace changes make sense independently? ie, are they the
right thing to do, even if they don't fix a visible bug? ISTM the
answer to both questions is yes. In which case, that part ought to go
forward now rather than waiting.
Yes. They fix the sort stability bug. The bug can be reproduced
without any ASan patches by compiling the ASan tests on powerpc64
with the -fasynchronous-unwind-tables option. It's also possible
that the bug can be reproduced with other programs on other
targets. It's just not triggered by the same tests on the targets
we commonly test.
OK. You can go ahead and commit the libbacktrace fix.
Let's hold off on the testsuite fixes until we've got the sanitizer &
libbacktrace fixes installed.
jeff