Add some comments and a new debug message about underrun

This commit is contained in:
Mike Brady
2021-06-04 09:41:15 +01:00
parent d33de05ae9
commit e4c82b7700
+22 -3
View File
@@ -1659,14 +1659,15 @@ int do_play(void *buf, int samples) {
if ((samples != 0) && (buf != NULL)) {
// jut check the state of the DAC
// just check the state of the DAC
if ((state != SND_PCM_STATE_PREPARED) && (state != SND_PCM_STATE_RUNNING) &&
(state != SND_PCM_STATE_XRUN)) {
debug(1, "alsa: DAC in odd SND_PCM_STATE_* %d prior to writing.", state);
}
// debug(3, "write %d frames.", samples);
snd_pcm_state_t prior_state = state; // keep this for afterwards....
debug(3, "alsa: write %d frames.", samples);
ret = alsa_pcm_write(alsa_handle, buf, samples);
if (ret == samples) {
stall_monitor_frame_count += samples;
@@ -1699,7 +1700,25 @@ int do_play(void *buf, int samples) {
frame_index = 0;
measurement_data_is_valid = 0;
if (ret == -EPIPE) { /* underrun */
debug(1, "alsa: underrun while writing %d samples to alsa device.", samples);
// It could be that the DAC was in the SND_PCM_STATE_XRUN state before
// sending the samples to be output. If so, it will still be in
// the SND_PCM_STATE_XRUN state after the call and it needs to be recovered.
// The underrun occurred in the past, so flagging an
// error at this point is misleading.
// In fact, having put samples in the buffer, we are about to fix it by now
// issuing a snd_pcm_recover().
// So, if state is SND_PCM_STATE_XRUN now, only report it if the state was
// not SND_PCM_STATE_XRUN prior to the call, i.e. report it only
// if we are not trying to recover from a previous underrun.
if (prior_state == SND_PCM_STATE_XRUN)
debug(2, "alsa: recovering from a previous underrun.");
else
debug(1, "alsa: underrun while writing %d samples to alsa device.", samples);
int tret = snd_pcm_recover(alsa_handle, ret, 1);
if (tret < 0) {
warn("alsa: can't recover from SND_PCM_STATE_XRUN: %s.", snd_strerror(tret));