The HSP and HFP profiles expect that a device function only as an audio
gateway or as an headset, which is the normal behavior for a headset,
a hands-free car unit or a phone.
In case of a desktop, it can perform both functionalities, but there's
no interest to get them at the same time as the bidirectional audio
is already supported.
I'm not 100% sure if this was breaking SSE41 builds on the official build system (I'm building Pipewire
with a different process), but I suspect it was, because you can't combine these into a single translation
unit to sidestep it without including multiple copies of resample-native-impl.h which isn't desirable.
We sync the filter graph in two places, make a function so that both
places do the same thing.
Make node_reset clear the setup flag so that we don't have to do that
twice.
If the the audioconvert.filter-chain.N property is set early, they will
be added to the active_graphs list but with setup = false. When the node
starts, setup_convert is called, but the graphs aren't added to
filter_graphs. Run the do_sync_filter_graph at the end of setup_convert
to add them.
The current implementation only send the +CIEV:<call>,<active> event
if there's an active modem in ModemManager. This may lead to headset
disconnection as in (1) if the profile is by another application than
telephony one, e.g. a conference application/website.
This commit improves dummy call status update by adding a new
"bluez5.disable-dummy-call" props param in bluez5 device, allowing
external application like WirePlumber to set it dynamically.
(1) https://gitlab.freedesktop.org/pipewire/pipewire/-/issues/1744
Fixes: https://gitlab.freedesktop.org/pipewire/pipewire/-/merge_requests/2606
Parse TMAP / GMAP features from MediaEndpoint:SupportedFeatures and pass
them onto the codec in SelectProperties, so it can determine which
mandatory features the device supports.
Add configuration option for specifying which TMAP / GMAP feature bits
we advertise to remote side.
Although some of these could be determined automatically, for production
systems it's better to have explicit option to specify which ones should
be advertised as this may depend on HW capabilities.
When the dynamic data flag is set on the buffer data, it means the
consumer can deal with any data pointer set on the buffer and we can
simply pass the one from upstream to downstream. If the flag is not set,
we need to copy the buffer data.
See #5009
When we recalculate the headroom we also update the latency in frames.
We should express this latency in the graph rate. This is usually the
rate that is suggested in the target_rate but when we are forcing our
own rate (mostly when using IRQ or when in DSD/IEC mode) we should
ignore that value and use our own rate that we will force instead.
Fixes#4977
Add feedback and feedforward controls to the delay. This makes it
possible to make comb and allpass filters with the delay to build
custom reverb effects.
Otherwise we might end up with partial channels when code doesn't
check the unpositioned flag. It's better to set everything to unknown
when there is a mismatch between channel count and layout.
The "+BIND: <a>,<state>" reply to AT+BIND? should be sent for every
supported indicator.
"AT+BIEV= <assigned number>,<value>" should only be provided for
enabled indicators. The AG shall respond with an ERROR response code
if it receives updates for disabled or unknown HF indicators or values
that are out of bounds.
This allows to pass PTs tests:
- HFP/AG/HFI/BV-02-C
AG receives an updated HF Indicator value
- HFP/AG/HFI/BI-03-C
AG receives invalid updated HF Indicator values
When releasing multiple transports, call Release() simultaneously
instead of serializing the calls.
This operations still needs to be blocking currently, as all releases
have to finish before we do other state-modifying ops.
This works around broken firmware on Creative Zen Hybrid Pro with BAP,
whose Disable command misbehaves when shutting down sink + source CIS
otherwise. It's also anyway better to shut down everything at once.
Increase the cutoff frequency for the lower quality modes. This should
give significantly better high frequency preservation at the expense of
more (but likely inaudible) aliasing.
Reduce the cutoff for the higher qualities in blackman and exp to
compensate the the wider transition band in those windows.
Increase cutoff for kaiser because of the sharper rolloff.
Add a new SPA_TYPE_OBJECT_ParamDict object that contains a struct with
key/value pairs. We're using something similar for Tags but this is a
more generic version.
Make a new Capability param that uses the ParamDict object. This is
meant to be used to describe capabilities with the generic key/value
struct.
Make a new PeerParam object where the keys are generic ids of the peer
objects and the values any Pod. The idea is to use this object to store
a peer objects. Make some helpers to iterate the peers and their
objects.
Add a new PeerCapability param that uses the PeerParam object with
Capability objects. This can be used to send the collection of
Capabilities from all peers to a port. This is a bit like the tags but
in a more generic way. The tags could also be implemented in this new
generic way later.
Make the PeerFormats use the same PeerParam of Format objects.
The Capability param is set on ports. impl-link will collect all
Capability objects into a PeerCapability object and send this to the
peer. The difference with the Tag param is that these Capability params
are not in any way forwared on the node automatically (like what is done
in the loopback module) because they represent the capabilities of the
ports betweem the link.
Some devices refuse to enable microphone if Streaming Context metadata
is just Unspecified.
Set some reasonable values for the stream context we create along TMAP,
and try follow CAP rules for selecting the PAC.
With BAP codec configuration selection goes via multiple functions,
which will need to maintain some private state.
Adjust media_codec to allow for that.
Use it for get_qos().
Based on HFP specs, the audio connection is independent of the active
call status, which should be managed by the ModemManager part of the
plugin.
But when using HFP AG without modem attached, e.g. during zoom meeting,
the connection will be closed after a while unless call status has been
forced to active,
cf. https://gitlab.freedesktop.org/pipewire/pipewire/-/issues/1744.
Currently and for HFP AG PTS tests requesting to get an audio connection
in 3 seconds after a call activates, this prevent to start audio
connection before starting a call.
This commit prevents to force the call status during audio (dis)connection
if a modem is available.
Get the ModemManager interfaces when the ModemManager starts after native
HFP has started.
Also add log topic to be able to select log level for modemmanager plugin
part.
Do not parse, and do not report the negotiated frame rate. It is
completely ignored. Until support for the `FrameDurationLimits`
libcamera control is implemented, ignore any and all frame rates.
This field is only used during the setup of buffers, so it can be
transformed into a local variable, but even that is not needed
because the data type can be chosen on a per-plane basis.
Make the params introspectable, use the names to configure them in
audioconvert.
Only use precomputed filters when using the defaults.
Tweak the kaiser filter alpha and n_taps based on values calculated from
stopband attenuation and transition bandwidth.
