andrzej zaborowski wrote:
> On 25/04/2008, Jan Kiszka <jan.kiszka@web.de> wrote:
>>  > The checking of whether the guest filled enough data happens in
>>  > wm8750_dac_dat(), I don't see why do it second time here.  The only
>>  > place we need an additional check is before s->dat_req call, which you
>>  > remove.
>>
>> True, that's redundant, let's fix it this way (this even obsoletes the
>>  flush in out_cb):
> 
> Ok, although I insist that it doesn't obsolete the check in out_cb
> (hence why I added it).  Imagine s->req_out was 5000, and the guest
> already wrote 1000 bytes. Now suddenly audio API calls out_cb with
> free_b == 1500. In this case we have half of the buffer ready to
> write, *before* we call s->data_req.

Yes, buffer/2 was attacking the symptom, wrongly, see my other mail.

> BTW I just committed a change to the MusicPal LCD, please check if it
> still works.  Thanks.

Still fine!

Jan