This is the mail archive of the
mailing list for the GCC project.
Re: add dbgcnt and opt-info support for devirtualization
- From: Xinliang David Li <davidxl at google dot com>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Fri, 16 May 2014 10:45:43 -0700
- Subject: Re: add dbgcnt and opt-info support for devirtualization
- Authentication-results: sourceware.org; auth=none
- References: <CAAkRFZKGLC9vW3-niUu2QrQGKZzUeq6Qv989bOaHdLjkn=6JqA at mail dot gmail dot com> <20140516160323 dot GE20755 at kam dot mff dot cuni dot cz> <CAAkRFZL3Zfmz=YCS78gYNZdpm2KL3B70BsT9Rm2icd0D-Xw7nw at mail dot gmail dot com> <20140516165115 dot GA31943 at kam dot mff dot cuni dot cz>
On Fri, May 16, 2014 at 9:51 AM, Jan Hubicka <email@example.com> wrote:
>> I have been chasing a runtime failure of a very large test built with
>> gcc-4_9. The bad code either calls a pure function or turn a virtual
>> call into __builtin_unreachable (incomplete target set). The indirect
>> info shows the otr type to be !maybe_derived_type, and the outer-type
>> gets cleared during inline update. I isolated a small test case -- but
>> the good news is that gcc-4_9 @head already fixed the problem.
>> I will check in the test case to trunk later.
> Good, testcase would be welcome. I guess it was the fix for placement_new bug.
> It disables some valid devirtualizations (and I thus may revisit the fix for
> 4.10), so it would be good to know if your testcase differs from the original
> PR one.
I thought so too -- but when I backed out this single change from
trunk, the test still passes, so something else is also going on.
>> > Honza