set_scheduler() assumes that if sys/resource.h was found then we will
find RLIMIT_RTTIME there, but this is a non-POSIX extension on Linux.
Change the check to ensure that RLIMIT_RTTIME is actually defined.
Linux indeed defines this as a macro, and POSIX specifies that the other
RLIMIT_ constants must be macros, so having this as an #ifdef seems
correct.
bootstrap.sh uses some non-POSIX features of bash, so we can't use
/bin/sh. Unlike /bin/sh, bash can be installed anywhere in the path, so
we should use /usr/bin/env to find it.
This helps systems that have bash in /usr/local/bin, for example.
PCM Devices which have the BATCH flag set update the PCM pointer only with
period size granularity. Using timer based scheduling does not have any
advantage in this mode. For one devices which have that flag set usually update
the position pointer in software after getting the period interrupt. So
disabling the period interrupt is not possible for this kind of devices.
Furthermore writing to or reading from the buffer slice for the current period
is not possible since the position inside the buffer is not known. On the other
hand the tsched algorithm seems to get easily confused for this kind of
hardware, which results in garbled audio output. This typically means that timer
based scheduling needs to be manually disabled on systems with such devices.
Auto disabling tsched in this case allows these systems to run with the default
configuration.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
the main intent is to make testing different sample rate resampling
implementations easier; so far there is only global control via
resample-method (command line argument and /etc/pulse/daemon.conf)
module-remap-*'s only purpose is resampling (comprising format conversion,
channel remapping, sample rate adjustment), it can easily be introduced
into any audio pipeline
Signed-off-by: Peter Meerwald <p.meerwald@bct-electronic.com>
by using pa_modargs_get_sample_rate() we avoid inconsistant validity
checking of the sample rate in various places
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
buf in struct ffmpeg_data is reset() initially and freed, but never
actually used
when a new block is allocated ffmpeg_data->buf[c].length is used
(which is always 0) to compute the new block size
so, drop buf
Signed-off-by: Peter Meerwald <p.meerwald@bct-electronic.com>
When setting attribute foo, or in this case the card profile, in my
opinion the thing passed to the set_foo() function should be of the
type of foo, not a string identifier that can be used to search for
the actual foo in set_foo().
This is mostly a question of taste, but there's at least some small
benefit from passing the actual object: often the profile object is
already available when calling pa_card_set_profile(), so passing the
card name would cause unnecessary searching when pa_card_set_profile()
needs to look up the profile from the hashmap.
Don't call pthread_join() to join a thread from a different
process than the thread was created in. Doing so can lead to
undefined behaviour.
On OpenBSD, the symptom was a pulseaudio process with a single
thread waiting forever for other threads to join. Since that
process also held the autospawn lock, starting new pulseaudio
processes with --start kept failing. The problem was analyzed
with help from Philip Guenther.
This patch adds a pa_thread_free_nojoin() function which can
be used to free resources for a thread without a join, as
suggested by Tanu Kaskinen.
See https://bugs.freedesktop.org/show_bug.cgi?id=71738
f434087e42 introduced the potential to not select a card profile if
all the profiles were marked as unavailable.
While this is very unlikely, it's a theoretical posibility, so if the
initial choice of a profile fails, try harder.
When parsing device properties, missing adapter will result in
device_info_valid being set to -1. It is then logical that if the
adapter goes missing at a later point, device_info_valid gets set to
-1 also in that situation.
The function did two things: set device_info_valid to -1 and called
device_free() for each device in the hashmap. Setting
device_info_valid to -1 was unnecessary. The main purpose of that was
to fire DEVICE_CONNECTION_CHANGED as a side effect, but that hook is
fired anyway in device_free(), as a side effect of removing all
transports. Calling device_free() can be delegated to pa_hashmap, when
freeing or emptying it.
Normally DEVICE_CONNECTION_CHANGED is fired when the first transport
becomes connected, but it may happen that the first transport becomes
connected already before the device properties have been received. In
that case the hook should be fired at the time the device properties
are received. This patch makes the hook to be fired at the right time.
At this point this doesn't make any other practical difference than
making the code more logical, but in the next patch I'll fire the
DEVICE_CONNECTION_CHANGED hook in set_device_info_valid(), and at that
point it's important that the device isn't marked valid too early,
because otherwise external code would see "valid" devices that however
don't have the adapter set.
This reverts commit c327850d9e as
the workaround in that commit is no longer needed after the real
bug has been fixed.
Conflicts:
src/pulsecore/core-util.c
Commit 7e344b5 hade the side effect of forcing every socket to
be non-blocking on Windows. This is because of a (documented)
side effect of WSAEventSelect(). So we need to make sure to restore
blocking behaviour afterwards for relevant sockets.
When connecting to a remote server your local generated authentication
cookie is used. If remote server's cookie is different from your local
one you aren't allowed to connect. You can use the cookie argument
or define a wider acl in remote server configuration for
module-native-protocol.
It's bad form to assume in free() that any member of the struct has
been initialized. I ran into problems with this when I reordered
things in pa_sink_input_new() and pa_source_output_new().
leftover_buf points to the output buffer of a stage containing leftover
data; similar for leftover_buf_size and have_leftover
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
some resampler implementations (e.g. libsamplerate and ffmpeg) do not consume
the entire input buffer; the impl_resample() function now has a return value
returning the number of frames in the input buffer not processed
these frames must be saved in appropriate buffer and presented together with
new input data
also change the parameter names from in_samples, out_samples to in_n_frames,
out_n_frames, respectively (n_frames = samples / channels)
Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
When a card is being created and no profile has been assigned
pa_card_new will attempt to select one from the list but it does that
without checking the available flag which can lead to select profiles
not available.
The size of pa_card_profile_info cannot change even if it just a field
appended to end because each entry is appended to a contiguous memory
and accessed by offset this may lead clients to access invalid data.
To fix a new struct called pa_card_profile_info2 is introduced and shall
be used for now on while pa_card_profile_info shall be considered
deprecated but it is still mantained for backward compatibility.
A new field called profiles2 is introduced to pa_card_info, this new field
is an array of pointers to pa_card_profile_info2 so it should be possible
to append new fields to the end of the pa_card_profile_info2 without
breaking binary compatibility as the entries are not accessed by offset.
These kcontrol names have started to show up lately, in
combination with surround internal speakers.
BugLink: https://bugs.launchpad.net/bugs/1236965
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
