On Sunday 01 November 2009 09:02:22 Tom wrote:
> Mike Frysinger wrote:
> > On Saturday 31 October 2009 22:00:09 Tom wrote:
> >> Mike Frysinger wrote:
> >>> On Saturday 31 October 2009 13:37:39 Tom Rix wrote:
> >>>> +#ifdef DEBUG
> >>>> +static inline void print_device_descriptor(struct
> >>>> usb_device_descriptor *d) +{
> >>>> + serial_printf("usb device descriptor \n");
> >>>
> >>> do you really need serial_printf() ? what's wrong with debug() ? then
> >>> you dont even really need "#ifdef DEBUG" around all the functions ...
> >>
> >> The explicit serial_printf is done because this patch set changes the
> >> stdin and stdout for serial to usbtty.
> >>
> >> When you use printf to debug printf, you regress into a bad state.
> >
> > so in your specific use case it makes sense *some* of the time (usbtty is
> > enabled and the console has been changed to it), but in the general use
> > case (usb debugging), it does not. why not make it intelligent instead
> > of penalizing everyone to use their serial console:
> > - default to debug()
> > - if usbtty support is enabled, check the current stdout console to see
> > if it's set to a usbtty, and only then fall back to forcing
> > serial_printf()
>
> A better solution would be to combine this logic into debug().
yes and no. debug() should not change, but a new debug function would
probably make sense. debug_stdio() where the first argument would be the
device name you're debugging ("usbtty" here) and would handle sending directly
to the serial functions if current stdio is set to that.
i look forward to your proposed code ;)
-mike
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
