[Bug ipa/113359] [13 Regression] LTO miscompilation of ceph on aarch64

["rguenther at suse dot de" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113359

--- Comment #13 from rguenther at suse dot de <rguenther at suse dot de> ---
On Tue, 6 Feb 2024, jakub at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113359
> 
> --- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
> Just going from the demangled name of
> std::pair<std::chrono::time_point<ceph::mono_clock,
> std::chrono::duration<unsigned long, std::ratio<1l, 1000000000l> > > const,
> Context*>
> it would surprise me if it was ODR violation in the testcase, because class
> Context
> is only defined in Timer.ii, not in the other preprocessed source,
> ceph::mono_clock is defined in both but looks the same (and it is empty class
> anyway), and the pair uses Context* as second type anyway, so it is unclear how
> it could become something smaller than pointer.
> But, admittedly I haven't looked up at this under the debugger and not even
> sure where to look at that.

Might be also an interaction with IPA ICF in case there's a pointer to
the pair involved?