On Wed, Mar 15, 2023 at 02:39:43PM +0000, Richard Biener via Gcc-patches wrote:
> I've tested and most of the -Wuse-after-free testcases work with -Og
> before and after the change (unfortunately adding -Og via RUNTESTFLAGS
> doesn't work since explicit -O2 in the testcases overrides that).
>
> We also do not have much test coverage here and I'm hesitant to change
> this now.
>
> That said, with the adjusted first patch and the split out fix for
> g++.dg/warn/Wuse-after-free3.C the following is what I propose now.
> There's extra adjustments in gcc.dg/Wuse-after-free-2.c as we now
> mention the pointer that is used.
>
> I do think not running these kind of diagnostics very late, at least
> a diagnostic like this that's prone to false positives due to
> code motion. But I'm also happy to leave the state of affairs as-is
> for GCC 13, but I don't promise to be able to pick up things during
> next stage1 again.
>
> Re-re-re-bootstrap & regtest running on x86_64-unknown-linux-gnu.
>
> Unless I get clear opinions I'm going to throw some dice whether
> to apply 1/2 and will "defer" this one.
Let's go with both patches then, if somebody files false negative issues
for this warning later on, we can amend it further later.
Jakub
