As reported in http://kpaste.net/04f1f3f
it is possible to call enumeration_is_subset with null pointers.
Handle that case instead of crashing. (It is also possible that
Tanuk's pending element_is_subset patch solves the issue, but this
nevertheless gives some extra security.)
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
We want to set the volume callbacks only if volume sharing
is not used. When volume sharing is used, we don't want to
mess with the stream volumes.
This was broken in 6c6b50
Now that subset mixer paths are removed, this workaround is no longer needed.
This effectively reverts 1c38b5d478 but due
to me forgetting to add files and adding a couple extra workarounds after,
it's easier to just do this manually rather than run git-revert.
In order to try and avoid 'spamming' the user with port choices,
attempt to detect and remove any pointless paths in a path set. That is
any paths which are subsets of other paths.
This should solve a problem case with some USB Headsets which result in
two paths both involving the 'Speaker' element. When no 'Master' element
exists (which is quite common on head/handsets), then the first path
(analog-output) will contain the 'Speaker' in a way that completely fits
with in the use of the 'Speaker' element in the other path
(analog-output-speaker).
Unification is really just a 'lowest common denominator' system. If any
paths do not support volume, mute or decibels, then mark them all as not
having them.
This was originally done this way because the flags set on sinks that
dictate if it supports h/w volume, mute etc. could not be changed after
the sink was created.
The fact that these flags could not change has now been change in the
previous commits, and thus there is now no need to use this 'lowest
common denominator' approach as we can fully support the various
different combinations, even if they change after initial creation
of the sinks/source.
Some sink flags are really just a product of what callbacks
are set on the device. We still enforce a degree of sanity
that the flags match the callbacks set, but we also set the
flags automatically in our callback setter functions to
help ensure that a) people use them and b) flags & callbacks
are kept in sync.
This is not currently useful but future commits will make further
changes concerning automatic setting of flags and event delivery
that makes this structure necessary.
This picks sane defaults for the sample spec used (32 kHz, mono) and
preprocessing (on by default). This should make it unncessary to provide
additional parameters in the default desktop case.
The main exception would be decreasing the sample rate for hardware with
limited processing power (can bring it down to 16 or 8 kHz).
This is a workaround - these usb headsets have one output volume
control only, labeled "Speaker". This causes the default profile
set to not control the volume at all, which is a bug. Workaround
that by creating a separate profile set.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
When a TCP socket is created the size of the send buffer (SO_SNDBUF) used is
determined by the OS, using the net.ipv4.tcp_wmem sysctl parameter. Previously
a call to setsockopt set the buffer size to a value that was too small, and
that in some cases could result in underruns and choppy playback. This
setsockopt call has now been removed so that the value determined by the OS
is used unchanged.
Note that the value used for the send buffer size is the 2nd value in
net.ipv4.tcp_wmem, e.g. if this is set to "4096 65536 8388608" the send buffer
size is set to 65536.
This adds code to specifically support legacy entries.
I kept this code in a separate commit so that it can be (relatively)
easily removed at some point in the future.
This simply exposes the formats that a device supports
via a simple protocol extension that will allow clients
to setup what a connected receiver supports format wise.
This has the advantage of allowing versioned updates in the future,
thus allowing us to be more user friendly going forward (as opposed
to just ignoring entries from old versions).
The primary motivation for this, however, is to allow variable length
storage in each entry which will be needed for upcoming work.
At present this commit will ignore any legacy entries but support
for reading and subsequently converting legacy entries will be added
shortly.
The previous logic in ade0a6f884
does not work with for input volumes.
This was discussed on the mailing list:
https://tango.0pointer.de/pipermail/pulseaudio-discuss/2011-May/010091.html
This approach can introduce a problem when setting the volumes
for sources. What follows is Tanu Kaskinen's analysis:
[quote]
I'll quote the log:
D: protocol-native.c: Client pavucontrol changes volume of source alsa_input.pci-0000_00_1b.0.analog-stereo.
D: alsa-source.c: Requested volume: 0: 45% 1: 45%
D: alsa-source.c: in dB: 0: -20.71 dB 1: -20.71 dB
D: alsa-source.c: Got hardware volume: 0: 45% 1: 45%
D: alsa-source.c: in dB: 0: -21.00 dB 1: -21.00 dB
D: alsa-source.c: Calculated software volume: 0: 101% 1: 101% (accurate-enough=no)
D: alsa-source.c: in dB: 0: 0.29 dB 1: 0.29 dB
D: source.c: Volume going up to 29273 at 270475970821
D: source.c: Volume change to 29273 at 270475970821 was written 34 usec late
D: alsa-source.c: Written HW volume did not match with the request: 0: 45% 1: 45% (request) != 0: 42% 1: 42%
D: alsa-source.c: in dB: 0: -21.00 dB 1: -21.00 dB (request) != 0: -22.50 dB 1: -22.50 dB
Looking at the last line, the requested volume seems to hit exactly the
right step (-21.00dB), but for some reason Alsa decides to choose
something else. I'm pretty sure that this happens because of rounding
errors. In the first phase we ask Alsa what dB value we should set, and
it returns -21.00 dB. The value is given as a long int, but we convert
that to pa_cvolume. Then when we set the volume, we convert the
pa_cvolume value back to a long integer. At this point I believe it gets
converted to -2101. This is not visible in the debug message for some
reason - the rounding algorithm must be different from what was used
with the pa_cvolume -> long conversion.
[/quote]
The commit after this contains a patch that addresses this issue.
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)
We were using the block size in bytes instead of samples, which meant
preprocessing was broken. This fix makes a large-ish difference in the
quality of echo-cancellation with speex.
The smoother is paused on initialization and resumed when the sink
state is set to running. Otherwise, early latency estimates are
too low since there is some delay between module initialization and
entering the running state.
After the smoother is initially resumed, it is paused when the sink
state is not running. The previous behavior was to pause only when
the sink enters suspended state, however, this would lead to large
errors in latency estimates after the sink has been idle for some
time.
Add a variable to track whether the actual volume is set or not.
Suppose this:
min volume: -126 max volume: 0
then when user wants to set some constant volume to -10, it would fail.
While the alsa values are typically positive, some values are "funky"
and have negative values. It is desirable to fix this at the alsa
level so that the numbers are positive, but it's not technically
invalid, and thus we have to support it.
Discussed here:
http://thread.gmane.org/gmane.comp.audio.pulseaudio.general/9832
and
http://thread.gmane.org/gmane.linux.alsa.devel/85459
If module initialisation fails, the speex done() function might try to
free a value that's not been allocated yet. Adding protection for this
condition.
The echo suppress attenuation value was being incorrectly modified.
Fixed and added 2 arguments to change the attenuation of the residual
echo filter. Default values of the speex preprocessor will be used when
omitted.
This allows the selective enabling of speex' preprocessing algorithms
before running the echo-canceller -- for now this includes automatic
gain control, noise suppression and echo suppression. It's all off by
default for now, though at some point in the near future we might want
to enable at least denoising by default.
The denoising works pretty well, though we might want to add a way to
tweak the noise-suppression knob that libspeex provides.
The AGC option is just a stop-gap -- we need a real AGC mechanism that
tweaks the source volume rather than doing this in software.
The speex documentation mentions VAD and dereverb, but it appears that
these are not complete yet.
We don't do all this in a separate module from module-echo-cancel to
avoid the overhead of adding another virtual source. It makes more sense
to make a separate virtual source module that can be used for cases
where preprocessing is useful but AEC is not (for e.g. noise suppression
for fan noise in a recording application).
Another reason to keep this integrated with the AEC module is that the
echo suppression bits use the speex echo canceller state. This does leak
some information about the AEC implementation into module-echo-cancel,
but this is unavoidable.
