I'm not an arm guru, my knowledge of arm is near non-existent.
The commit in qemu which I bisected this to yesterday, apparently, is a bugfix, - previously qemu allowed an unaligned access here, while real hardware does not. Now qemu behavior matches that of the hardware. So it might as well be a bug in picolibc. I tried to run picolibc testsuite on a bare metal machine, not in qemu (on an arm debian porterbox). But I can't figure out how to run picolibc test binaries in this case. When running `meson test`, it just skips ALL tests. Maybe it has to be built in some other way, I don't know. It's best if someone who knows arm and/or picolibc to actually step in at this point and check how this stuff works on a real hw, - maybe picking up any of the failing tests as I did with qemu. As long as I'm not proven wrong, I assume the bug is on the picolibc side here. We did have another issue in this area in qemu after fixing this too-lose alignment defect, but it's fixed now too, while it does not help picolibc (https://gitlab.com/qemu-project/qemu/-/issues/2326). Thanks, /mjt
