On 5/23/22 06:30, Peter Maydell wrote:
So this changes the errno values in the not-gdb case from being
host errno values to the gdb protocol ones. Errnos in Arm semihosting
are an unspecified mess, so maybe we can get away with changing
the existing QEMU behaviour; but I'd rather we didn't do it
one syscall at a time in a big refactoring series if we can avoid it.
Ok.
Also, I think I mentioned this in the v2 cover but not here, that having done the errno
conversion here for arm semihosting, it worked less well for mips and xtensa, which have a
rather better defined set of errnos.
My question from v2 was: should we in fact convert back from gdb's errno to host errno in
gdbstub.c handle_file_io(), and then let each semihosting backend convert from host to guest?
r~
