https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124266
--- Comment #5 from Wilco <wilco at gcc dot gnu.org> --- Looks like the issue is caused by aarch64_emit_cfi_for_reg_p() - not sure why it even exists, but it seems incorrect to block emitting frame notes on registers that we must save/restore... The crash is caused by notes sometimes getting partially emitted anyway - this mismatch is then detected by dwarf2cfi.
