https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123853
--- Comment #23 from John Paul Adrian Glaubitz <glaubitz at debian dot org> --- (In reply to Mikael Pettersson from comment #22) > (In reply to John Paul Adrian Glaubitz from comment #21) > > (In reply to Mikael Pettersson from comment #18) > > > Created attachment 64643 [details] > > > patch I'm testing > > > > > > Patch to reject the problematic insns by adding a predicate to detect them > > > and check it in the conditions for the *movsi_m68k and *movsi_m68k2 > > > insns. > > > If someone knows of a better way please let me know. Only lightly tested > > > so > > > far, but seems to do the right thing on the test case in #c11. > > > > > > I'm going to bootstrap and regression test this natively on Aranym, but > > > that > > > will take weeks as I need to establish a current baseline first. > > > > Has this patch been posted to gcc-patches yet? > > No, because as I wrote native bootstrap and regression testing will take > weeks. I would recommend qemu-virt which emulates an 68040 CPU with about 2 GHz and 3 GB RAM. I will also do some more testing, I suspect the patch might fix PR/120165.
