01.06.2013 14:50, Michael Tokarev wrote:
> [Replying to a very old email which I missed initially]
>
> Todd, do you still have the issue you describe?
>
> I mean, not that we changed lots in that area (I guess we
> changed nothing there), but i'm not sure I understand what
> is going on there, and why the behavour changed the way you
> describe. I guess there's some other bug somewhere which
> you see now.
Hm. So I finally see the issue here, after I started messing
with displays in qemu.
> 12.03.2013 20:42, Todd T. Fries wrote:
>> So I'm having problems that I believe I've traced back to the commit
>> ab51b1d568e02c80b1abf9016bda3a86dc1db389 on qemu-system-sparc.
>>
>> Specifically, the commit states that:
>>
>> -nographic
>>
>> is equivalent to (replacing the 'null' with 'none'):
>>
>> -serial none -parallel none -monitor none -display none -vga none
>>
>> However on qemu-system-sparc this is not the case.
>>
>> Something about '-nographic -daemonize' works on qemu 1.3.1 to permit
>> further specifying the serial port elsewhere (unix: or tcp: for example).
This is because of the following code:
hw/sparc/sun4m.c:sun4m_hw_init():
slavio_serial_ms_kbd_init(hwdef->ms_kb_base, slavio_irq[14],
display_type == DT_NOGRAPHIC, ESCC_CLOCK, 1);
where 3rd arg for slavio_serial_ms_kbd_init() is `disabled'.
So without -nographic, the sun serial port becomes disabled
and obviously does not work.
So apparently we have other uses of -nographic in the code.
Another are 2 referenses in qemu-char.c:
qemu-char.c: stdio_allow_signal = display_type != DT_NOGRAPHIC;
qemu-char.c- backend->stdio->signal =
qemu-char.c: qemu_opt_get_bool(opts, "signal", display_type !=
DT_NOGRAPHIC);
So we need to fix these too.
Should it be switched to display=none? Or alternatively,
we have this in qemu-char.c already:
qemu-char.c: if (is_daemonized()) {
qemu-char.c: error_report("cannot use stdio with -daemonize");
so maybe tie it to is_daemonized() ?
If -nographic is legacy, what are the alternatives to the above usages
of it?
Thanks,
/mjt
