Hi Zhen,
On Fri, Dec 16, 2022 at 10:43 AM Leizhen (ThunderTown)
<thunder.leizhen@xxxxxxxxxx> wrote:
> On 2022/12/16 15:42, Leizhen (ThunderTown) wrote:
> > On 2022/12/15 22:51, Geert Uytterhoeven wrote:
> >> On 30f3bb09778de64 with your debug patch v2:
> > I've set up the qemu environment, and I'll try to solve it by tomorrow at the latest.
>
> It seems that the problem is still strcmp(). After I commented strcmp() in
> arch/m68k/include/asm/string.h, and force it to use the one in lib/string.c,
> it works well.
I can confirm that.
One difference is that the one in lib/string.c always return -1/0/1,
while the m68k version can return other negative or positive numbers.
However, adding:
if (res < 0) return -1;
if (res > 0) return 1;
to the m68k version doesn't make a difference.
Renaming the m68k version (changed to -1/0/1) to m68k_strcmp(), and
the generic version to lib_strcmp(), and adding a wrapper that calls
and compares both, shows that both functions do return the same value,
and the test succeeds.
Moving the m68k version inside lib/string.c makes the test pass, too.
So it must be related to the function being inline, and gcc making
(incorrect) assumptions...
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
