https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93738
--- Comment #28 from Segher Boessenkool <segher at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #24) > (In reply to Segher Boessenkool from comment #23) > > Hi! > > > > On Wed, Jan 14, 2026 at 12:23:37PM +0000, cvs-commit at gcc dot gnu.org > > wrote: > > > Thus, I think we need to limit to non-paradoxical SUBREGs. > > > > Why does subreg_lowpart_p ever allow paradoxical subregs? Those by > > definition are never a low part! > > > > Should we not just fix that, instead? > > I'm afraid such a change is too risky for stage4, for stage1, why not. Yeah, okay. > GCC has behaved that way since forever, certainly e.g. 3.2 has that in, and > I'm pretty sure at least some of the > 80 callers will rely on the current > behavior. And *not* rely on it on purpose. but it has only ever been tested with that weirdosity in place. If we really want these semantics, we should change the name of this so that it somewhat describes what it tests! Yeah, a short name is hard to come up with, it makes no sense after all :-(
