On 2025-11-03 06:47, Andreas Metzler wrote: > On 2025-11-02 Andreas Metzler <[email protected]> wrote: > [...] > > Also setting either no-schedule-insns2 or no-strict-aliasing for this > > function (instead of O1) for a -O2 build lets the arm64 build succeed. > > or alternatively > -O1 -fschedule-insns2 -fstrict-aliasing > breaks the build.
The segfault also does not occur with -O3, and the important difference seems to be that -O3 enables -fschedule-insns (without trailing '2'). Building with -O2 -fschedule-insns fixes pamtilt.
