On 24.10.2015 11:16, Tanu Kaskinen wrote: > On Fri, 2015-10-23 at 12:08 +0200, Georg Chini wrote: >> When the null sink changes state from suspended to idle (for example >> when attaching a loopback module to the monitor source), u->timestamp >> must also be updated. >> Thanks to Olivier Delbeke for reporting the bug and teting the patch. >> >> --- >> src/modules/module-null-sink.c | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/src/modules/module-null-sink.c b/src/modules/module-null-sink.c >> index 0270ba2..e6d261a 100644 >> --- a/src/modules/module-null-sink.c >> +++ b/src/modules/module-null-sink.c >> @@ -95,6 +95,8 @@ static int sink_process_msg( >> >> if (PA_PTR_TO_UINT(data) == PA_SINK_RUNNING) >> u->timestamp = pa_rtclock_now(); >> + if (pa_sink_get_state(u->sink) == PA_SINK_SUSPENDED && PA_PTR_TO_UINT(data) == PA_SINK_IDLE) >> + u->timestamp = pa_rtclock_now(); >> >> break; > I guess the goal is that the timestamp should be reset to "now" when > the sink transitions from INIT or SUSPENDED to IDLE or RUNNING, and not > at any other time. Now the timestamp is reset also in the IDLE -> > RUNNING transition, and INIT -> IDLE isn't handled. I guess this should > be changed? > > -- > Tanu INIT -> something should not be a problem because the timestamp is set to "now" when the thread starts up. The IDLE -> RUNNING case is handled like it was before the patch, but I was also thinking that this is not completely correct. I can change it if you prefer. Bug report is here: http://lists.freedesktop.org/archives/pulseaudio-discuss/2015-October/024553.html
