On 08/25/2011 07:53 AM, Raymond Yau wrote:
2011/8/23 Takashi Iwaitiwai@suse.de: be enough just to have a
boolean for underrun_detect yes/no.
So, the patch would be like below.
What do you think?
Takashi
diff --git a/pulse/pcm_pulse.c b/pulse/pcm_pulse.c index d6c6792..b0e52ab 100644 --- a/pulse/pcm_pulse.c +++ b/pulse/pcm_pulse.c @@ -42,6 +42,7 @@ typedef struct snd_pcm_pulse { int handle_underrun;
size_t offset;
int64_t written; pa_stream *stream;@@ -460,6 +461,7 @@ static snd_pcm_sframes_t pulse_write(snd_pcm_ioplug_t * io,
/* Make sure the buffer pointer is in sync */ pcm->last_size -= writebytes;
pcm->written += writebytes; ret = update_ptr(pcm); if (ret< 0) goto finish;@@ -585,6 +587,15 @@ static void stream_request_cb(pa_stream * p, size_t length, void *userdata) update_active(pcm); }
+#if defined(PA_CHECK_VERSION)&& PA_CHECK_VERSION(0,99,0)
compile fail at these line since PA_CHECK_VERSION is not defined in pulseaudio 0.9.14
I don't have PA 0.9.14 available, but that's why the line starts with "if defined(PA_CHECK_VERSION)" so that the second part should never be evaluated.
seem that pcm->wriiten is application pointer of pulse device in bytes
does it mean that pa_stream_get_underflow_index(0 is the hardware pointer ?
does it mean that pulse device no longer can be run 7 x 24 non-stop since pcm->written may overflow ?
Hmm. It's an int64, so your computer is likely to break down before there is a wraparound, but perhaps there is a slight chance that the comparision will be wrong if there is an underrun at the exact wraparound moment.