A dynamic array is a nice simple container, but the old interface
wasn't quite what I wanted it to be. I like GLib's way of providing
the free callback at the container creation time, because that way
the free callback doesn't have to be given every time something is
removed from the array.
The allocation pattern was changed too: instead of increasing the
array size always by 25 when the array gets full, the size gets
doubled now (the lowest non-zero size is still 25).
The array can't store NULL pointers anymore, and pa_dynarray_get() was
changed so that it's forbidden to try to access elements outside the
valid range.
The set of supported operations may seem a bit arbitrary. The
operation set is by no means complete at this point. I have included
only those operations that are required by the current code and some
unpublished code of mine.
It's easier to work with the port description if it can be assumed
that it's always non-NULL. I have checked that the current code base
always ensures a non-NULL description.
With the new behaviour, you will not always get a callback after a
successful write. Make sure the callers can properly handle this.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
This patch removes all occurrences of double and triple
newlines.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-a -not -name 'adrian-aec.*' -a -not \
-name reserve.c -a -not -name 'rtkit.*' \
-exec sed -i -e '/^$/{N;s/^\n$//}' {} \;
Two passes were needed to remove triple newlines.
The excluded files are mirrored files from external sources.
This patch replaces every occurrence of ')\n{' with ') {'.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-a -not -name core-util.c -a -not \
-name adrian-aec.c -a -not -name g711.c \
-exec sed -i -e '/)$/{N;s/)\n{$/) {/}' {} \;
The excluded files are mirrored files from external sources.
This patch replaces every occurrence of '){' with ') {'.
The ffmpeg source tree was excluded since it will disappear anyways.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-exec sed -i -e 's/){/) {/' {} \;
This patch replaces every occurrence of 'if(' with 'if ('.
The ffmpeg source tree was excluded since it will disappear anyways.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-exec sed -i -e 's/ if(/ if (/' {} \;
This patch removes all tabs hidden inside the source tree and replaces
them with 4 spaces.
Command used for this:
find . -type d \( -name bluetooth \) -prune -o
-regex '\(.*\.[hc]\|.*\.cc\)' -a -not -name 'reserve*.[ch]'
-a -not -name 'gnt*.h' -a -not -name 'adrian*'
-exec sed -i -e 's/\t/ /g' {} \;
The excluded files are mirrored files from external sources containing
tabs.
This code is from heftig, but the mistake that I'm fixing here is my
own. Before applying heftig's patch, I downgraded the level of one of
the log messages. I managed to downgrade a different message than what
I intended, so now I'm undoing that mistake.
file_path contains the last tried file name, including the suffix, so
the error message was wrong:
Tried to open target file '/tmp/test.log.99', '/tmp/test.log.99.1',
'/tmp/test.log.99.2' ... '/tmp/test.log.99.99', but all failed.
The outputs are removed from the idxset before output_free() is
called. Trying to remove them again in output_free(), and asserting
that it should succeed caused crashing whenever outputs were freed.
This bug was introduced in commit
061878b5a4.
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=65901
We need the mainloop lock to be taken around pa_mainloop_api_once() to
prevent an assert due to the defer event creation and setting of the
destroy callback not being performed atomically.
This needs us to expose a bit of implementation detail, but this seems
to be the cleanest way without an API change.
The specific problem is that pa_mainloop_api_once() needs to first
create a defer event and then set its destroy callback. If the defer
event is completed before the callback is set, an assert will be
trigerred.
Now that we don't *always* get a callback after having written
something, make sure we can continue writing as long as it fully
succeeds.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
To save some CPU (in low latency scenarios), don't re-enable the
"writable" event after it has succeeded. It is very likely the next
write will succeed right away too.
This means that we always need to handle EAGAIN/EWOULDBLOCK as a
successful write of 0 bytes, so I also verified that all callers to
pa_iochannel_write handled this correctly.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
The tsched_watermark is in bytes, not in usecs. Fix this by introducing
a new variable, and also use that variable in some places for optimisation.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
If there is a "Line Out" jack present, then add this path. The fallback
analog-output will be a subset of this path and removed.
I only use the "Line Out Jack" or "Line Out Front Jack" for actual jack
detection - without anything connected to the front jack, it makes little
sense to enable the port.
(Another option could perhaps be to use different paths for stereo line out
and surround line outs, but that could be a possible future improvement.)
Assume that the headphone port volume is lower than the speaker volume.
When plugging in headphones, if the path is active, while the jack is
being inserted and before it is actually detected as being plugged in,
it will still receive the signal being played (which is at a higher
volume than it will be when plugged in completely). The volume
difference manifests as a volume spike when the headphones are plugged
in, before the final volume is set.
This patch is required to prevent such a volume spike when plugging in
headphones. The problem is not fixed completely, but the spike is
shortened. To be fixed completely, we need to apply the port volume
before unmuting the new path.
In the default configuration, PulseAudio's rlimit-rttime is set to
1000000 (100%), which is higher than what RealtimeKit requires from
its clients (200000, 20%).
Make an attempt to still get realtime scheduling by clamping the
current RLIMIT_RTTIME to what RealtimeKit accepts. Warn about doing
this.
This makes the test more robust by:
1. Decreasing the '1' threshold during calibration - the RMS value for
the sine wave will be 0.5, so the previous code was making us take
the ALSA mixer past 0dB.
2. Using the difference rather than absolute value for 0->1 transitions,
so that we're somewhat independent noise in our calculations.
This test is intended to measure real latency by playing a sample to a
sink and capturing that over a loopback interface. The loopback can
either be physical (cable running from headphone out to line in) or
virtual (monitor source or module loopback).
Also included in this is calibration code to make sure that volumes are
sufficiently adjusted to be able to detect the played back signal (and
that there aren't false positives due to line noise).
One of the objectives of all this is to later factor out the setup code
to allow us to easily write more loopback tests for various
functionality (volumes, resampling, mixing, etc.).
This pushes all avahi-client code to a threaded mainloop from the PA
mainloop context. We need to do this because avahi-client makes blocking
D-Bus calls, and we don't want to block the mainloop for that long.
The only exception to this now that I don't see a workaround for is
during module unload time. However, this shouldn't be a huge problem
since in most cases, this will only happen at server shutdown time.
The bulk of the change is partitioning the data so that PA core objects
only (well, mostly) get accessed in the PA mainloop and Avahi calls
happen only in the Avahi threaded mainloop.
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=58758
