In pa_create_stream_callback, a stream is inserted into
s->context->record_streams only if it's a record stream. Otherwise it's
inserted into s->context->playback_streams. However, in stream_unlink the
stream is removed from s->context->playback_streams only if it's a playback
stream and otherwise it's removed from s->context->record_streams.
Thus, if the stream is an upload stream, we first insert it into
s->context->playback_streams in pa_create_stream_callback and then try to
remove it unsuccessfully from s->context->record_streams in stream_unlink. This
means that we are leaking hashmap entries until the context is freed,
constantly consuming more memory with applications that upload and unload a
large number of samples through one context.
Of course, this begs the question whether upload streams even belong in either
of those hashmaps. I don't want to mess around with the code too much at this
point though, so this patch should be a sufficient improvement.
These are not used for anything at this point, but this
makes it easy to add ad-hoc debug prints that show the
memblockq name and to convert between bytes and usecs.
This patch introduces some extra protocol information, so protocol
version is bumped. This functionality is primarily needed to solve
a long standing issue in alsa-plugins, which should ignore underruns
if and only if it is obsolete, i e, if more data has been written to
the pipe in the meantime (which will automatically end the underrun).
BugLink: http://bugs.launchpad.net/bugs/805940
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
This piggy backs onto the previous changes for protocol 22 and
thus does not bump the version. This and the previous commits should be
seen as mostly atomic. Apologies for any bisecting issues this causes
(although I would expect these to be minimal)
Passing a NULL-terminated array of pa_format_info pointers is a bit
unwieldy for clients. Instead of this, let's pass in an array of
pointers and the number of elements in the array.
This quite is an old patch. It was added to N900 to avoid unnecessary
wake-ups when the phone is in power save mode (= blank screen and
no user interaction). In this situation if the user had a browser
window with flash animation open pulseaudio kept waking up every
10 seconds, causing a severe hit to use times.
Anyway I do not see any reason to send timing updates if the sink or
source where the stream is connected to is suspended.
When the sink format changes and we kill the stream, clients need a way
to know (a) what device they should reconnect to, and (b) what the
stream running time was when the stream got killed (pa_stream_get_time()
won't work after the stream has been killed). This adds these two bits
of information in the event callback's proplist parameter.
This is the beginning of work to support compressed formats natively in
PulseAudio. This adds a pa_stream_new_extended() that takes a format
structure, sends it to the server (=> protocol extension) and has the
server negotiate with the appropropriate sink to figure out what format
it should use.
This is work in progress, and works only with PCM streams. Actual
compressed format support in some sink needs to be implemented, and
extensive testing is required.
More details on how this is supposed to work is available at:
http://pulseaudio.org/wiki/PassthroughSupport
This prevents the smoother attached to the stream clock from being
updated while the stream is corked, which in turn ensures that once
corking is completed, pa_stream_get_time() always returns the same value
until the stream is uncorked - i.e., the clock does not advance when the
client believes that it will not.
The actual call to pa_smoother_put() happens on things like stream
suspend/unsuspend, which trigger timing updates. This changes the
smoother coefficients, which means that a call to pa_smoother_get() for
the same value of 'x' can return different values before and after a
timing update.
Second version after Tanu's feedback
TODO:
- notify client that volume control is disabled
- change sink rate in passthrough mode if needed
- automatic detection of passthrough mode instead of hard
coded profile names
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
All seeks/flushes that depend on the playback buffer read pointer cannot
be accounted for properly in the client since it does not know the
actual read pointer. Due to that the clients do not account for it at
all. We need do the same on the server side. And we did, but a little
bit too extreme. While we properly have not applied the changes to the
"request" counter we still do have to apply it to the "missing" counter.
This patch fixes that.
Do not subtract bytes the client sends us beyond what we requested from
our missing bytes counter.
This was mostly a thinko that caused servers asking for too little data
when the client initially sent more data than requested, because that
data sent too much was accounted for twice.
This commit fixes this miscalculation.
http://bugzilla.redhat.com/show_bug.cgi?id=534130
Since the stream identifiers (channels) are monotonically growing integer, it
isn't a good idea to use them as index to a dynamic array, because the array
will grow all the time. This is not a problem with client connections that
don't create many streams, but, for example, long-running clients that use
libcanberra for playing event sounds, this means that the client connection
effectively leaks memory.
Move the mainloop to monotonic based time events.
Introduces 4 helper functions:
pa_{context,core}_rttime_{new,restart}(), that fill correctly a
timeval with the rtclock flag set if the mainloop supports it.
Both mainloop-test and mainloop-test-glib works with rt and timeval
based time events. PulseAudio and clients should be fully functional.
This patch has received several iterations, and this one as been
largely untested.
Signed-off-by: Marc-André Lureau <marca-andre.lureau@nokia.com>
The primary reason for this change is to allow time graphs that do not
go through the origin and hence smoothing starting from the origin is
not desired. This change will allow passing time data into the smoother
while paused and then abruptly use that data without smoothing using the
'quick fixup' flag when resuming.
Primary use case is allowing recording time graphs where the data
recorded originates from a time before the stream was created. The
resulting graft will be shifted and should not be smoothened to go
through the origin.
