[Bug c/102268] Wrong code with aliasing union pointers

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102268

--- Comment #6 from rguenther at suse dot de <rguenther at suse dot de> ---
On Fri, 10 Sep 2021, davmac at davmac dot org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102268
> 
> --- Comment #5 from Davin McCall <davmac at davmac dot org> ---
> Hi Richard,
> I think you marked as a duplicate of the wrong bug. It is indeed a duplicate of
> 82224 - sorry, I didn't realise that there was already a bug filed, also this
> test case is "fixed" by version 8.1 and then regresses again in 10.1 (but I
> assume that is coincidental rather than the underlying issue was fixed).

The bug now happens / resurfaced because of 101641, I have not checked
the only slightly different testcases in 82224 (but that bug as a quite
big audit-trail with multiple testcases).

There have been multiple fixes in this area in the last years but the
situation in PR101641 now persists where we lose track of a change
of the active union member because we elide a store into it as
redundant (writing the same bit pattern into the same location).