Currently the rewind logic for the source output is broken if the output
does not implement a process_rewind() callback. In that case, the read
index of the delay memblockq is rewound. This is wrong, because the data
that is going to be re-written was not yet read. Instead the write index
should be rewound and the read index left untouched. This is the reason
for the rewind glitches of monitor sources.
The original atomic implementation in pulseaudio based on
libatomic stated that the intent was to use full memory barriers.
According to [1], the load and store implementation based on
gcc builtins matches sequential consistent (i.e. full memory barrier)
load and store ordering only for x86.
I observed random crashes in client applications using memfd srbchannel
transport on an armv8-aarch64 platform (cortex-a57).
In all those crashes the first read on the pstream descriptor
(the size field) was wrong and looked like it contained old data.
I boiled the relevant parts of the srbchannel implementation down to
a simple test case and could observe random test failures.
So I figured that the atomic implementation was broken for armv8
with respect to cross-cpu memory access ordering consistency.
In order to come up with a minimal fix, I used the newer
__atomic_load_n/__atomic_store_n builtins from gcc.
With
aarch64-linux-gnu-gcc (Linaro GCC 7.3-2018.05) 7.3.1 20180425
they compile to
ldar and stlxr on arm64, which is correct according to [1] and [2].
The other atomic operations based on __sync builtins don't need
to be touched since they already are of the full memory barrier
variety.
[1] https://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html
[2] <https://community.arm.com/developer/ip-products/processors
/b/processors-ip-blog/posts/armv8-a-architecture-2016-additions>
It was omitted. This patch fixes unexpected behavior that avoid-
resampling does not work in some cases.
Signed-off-by: Sangchul Lee <sangchul1011@gmail.com>
Brings things in line with the autotools build, and adds ALSA mixer
paths and profile-sets into the meson build system as well.
The module installation path is also now customisable.
This patch adds a new feature to the core which allows to send messages
to objects. An object can register/unregister a message handler with
pa_message_handler_{register, unregister}() while a message can be sent
to the handler using the pa_message_handler_send_message() function.
A message has 4 arguments (apart from passing the core):
object_path: The path identifying the object that will receive the message
message: message command
message_parameters: A string containing additional parameters
response: Pointer to a response string that will be filled by the
message handler. The caller is responsible to free the string.
The patch is a precondition for the following patches that allow clients
to send messages to pulseaudio objects.
There is no restriction on object names, except that an object path
always starts with a "/". The intention is to use a path-like syntax,
for example /core/sink_1 for a sink or /name/instances/index for modules.
The exact naming convention still needs to be agreed.
The current code uses a pa_strbuf to construct the escaped string. This
will generate a linked list member for each character which may be very
inefficient.
This patch avoids the use of pa_strbuf by allocating a sufficiently large
string which can be filled with the output data.
These events were missing, because the
pa_core_update_default_sink/source() calls were assumed to send the
subscription events when necessary. Often that indeed is the case, but
if the current configured default sink doesn't exist, and then the
current default sink is set as the configured default sink, the
configured default sink changes but the default sink doesn't, and in
this case pa_core_update_default_sink() doesn't send the change event.
module-default-device-restore relies on getting a notification whenever
the configured default sink changes, and the missing event meant that
the files containing the configured sink and source weren't updated in
some cases.
Fixes: https://gitlab.freedesktop.org/pulseaudio/pulseaudio/issues/648
So far PulseAudio only supported two different work formats: S16NE if
it's sufficient to represent the input and output formats without loss
of precision and FLOAT32NE in all other cases. For systems that use
S32NE exclusively, this results in unnecessary conversions from S32NE to
FLOAT32NE and back again.
Add S32NE remap operations and make use of them (for the COPY and
TRIVIAL resamplers) if both input and output format are S32NE. This
avoids the back and forth conversions between S32NE and FLOAT32NE,
significantly improving performance for those cases.
Currently, virtual sinks and sources are not suspended when the master sink
or source is suspended. To implement this, the slave must be able to track
the suspend cause of the master.
With this patch, the sink input suspend callback will not only be called
when the sink or source is changing state, but also when the suspend cause
changes. Similar to the set_state_in_*_thread_cb() functions, the suspend
callback receives a state and a suspend cause as additional arguments.
Because the new state and suspend cause of the sink or source have already
been set, the old values are passed to the callback.
pa_split_in_place() and pa_split_spaces_in_place() are modifed
to use size_t type instead of integer type.
alsa-ucm.c is revised according to this change.
Signed-off-by: Sangchul Lee <sc11.lee@samsung.com>
pa_card_profile_set_available needs to check if the card is linked
before firing PA_CORE_HOOK_CARD_PROFILE_AVAILABLE_CHANGED, so callbacks
connected to it receive a fully initialized card object.
This fixes a crash introduced by commit 30a551bbc
"switch-on-port-available: Check if we need to change the active
profile".
It is helpful to improve reproducibility build [1] since
PA_SRCDIR/PA_BUILDDIR contains build path,
--disable-running-from-build-tree could drop these macros at
precompilation.
[1] https://reproducible-builds.org/
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Sample format(e.g. 16 bit, 24 bit) was not considered even if the
avoid-resampling option is set or the passthrough mode is used.
This patch checks both sample format and rate of a stream to
determine whether to avoid resampling in case of the option is set.
In other word, it is possble to use the stream's original sample
format and rate without resampling as long as these are supported
by the device.
pa_sink_input_update_rate() and pa_source_output_update_rate() are
renamed to pa_sink_input_update_resampler() and pa_source_output
_update_resampler() respectively.
functions are added as below.
pa_sink_set_sample_format(), pa_sink_set_sample_rate(),
pa_source_set_sample_format(), pa_source_set_sample_rate()
Signed-off-by: Sangchul Lee <sc11.lee@samsung.com>
mpv and vlc play "normal" 5.1 AC3 and DTS files as if they had a
"5.1 (Side)" layout. Which is fine and consistent with ITU
recommendations if the user has a proper 7.1 system. But if the user
actually has a 5.1 system, PulseAudio will try to remap, poorly, between
the "5.1 (Side)" and "5.1" layouts, sending either an average between
front and rear channels, or an attenuated version of that average,
depending on the remixing-use-all-sink-channels setting.
This is not desired, the "Side" channels should be sent to "Rear", it is
only an unfortunate nomenclature confusion.
This patch does not fix 5.1 <-> 7.1 remixing.
Signed-off-by: Alexander E. Patrakov <patrakov@gmail.com>
This is based on the initial protocol_native library that is already
defined, and then by looking at the Makefile.am to work out the
dependencies.
It's not clear whether we really need database_c_args, maybe there's
things that can be simplified.
Signed-off-by: Arnaud Rebillout <arnaud.rebillout@collabora.com>
We move over helper functions to get rate, channels, channel map and
sample format (if PCM) in the public API, so users of the extended API
are more easily able to pull out these values from pa_format_info.
Having a single level macro for stringizing LADSPA_PATH doesn't work,
because the '#' preprocessor operator doesn't expand any macros in its
parameter. As a result, we used the string "LADSPA_PATH" as the search
path, and obviously no plugins were ever found.
This adds a two-level macro in macro.h and uses that to expand and
stringize LADSPA_PATH.
Bug link: https://bugs.freedesktop.org/show_bug.cgi?id=107078
pa_sink_get_state() and pa_source_get_state() just return the state
variable. We can as well access the state variable directly.
There are no behaviour changes, except that module-virtual-source
accessed the main thread's sink state variable from its push() callback.
I fixed the module so that it uses the thread_info.state variable
instead. Also, the compiler started to complain about comparing a sink
state variable to a source state enum value in protocol-esound.c. The
underlying bug was that a source pointer was assigned to a variable
whose type was a sink pointer (somehow using the pa_source_get_state()
macro confused the compiler enough so that it didn't complain before).
I fixed the variable type.
pa_sink_input_get_state() and pa_source_output_get_state() just return
the state variable. We can as well access the state variable directly.
There are no behaviour changes, except that some filter sources accessed
the main thread's state variable from their push() callbacks. I fixed
them so that they use the thread_info.state variable instead.
The only thing that the drained state was being used for was "pacmd
list-sink-inputs". In all other cases the drained and running states
were treated as equivalent. IMHO, this usage doesn't justify the
complexity that the additional state brings.
This patch was inspired by a bug report[1] that pointed out an error in
an if condition in pa_sink_input_set_state_within_thread(). The buggy
code is now removed altogether.
[1] https://bugs.freedesktop.org/show_bug.cgi?id=106982
Current code does not check whether pa_play_file call failed. Hence no error is
reported in the cli interface if playback failed because e.g. file isn't
readable by the daemon.
