GitLab

Allow passing arbitrary data to async callback

Closes #7

See merge request mobian1/callaudiod!5

Arnaud Ferraris authored Nov 05, 2020 and Arnaud Ferraris committed Jan 12, 2021

Fixes #7

Add documentation

Closes #2

See merge request mobian1/callaudiod!2

meson: Fix libcallaudio out of tree builds

Closes #9

See merge request mobian1/callaudiod!4

Removes hardcoded include flags in favour of using include_directories

This will make gtk-doc slightly more happy

Some operations wrongfully used g_error(), causing the daemon to exit in 
error cases where it could have continued working.

This patch fixes this by using g_critical() instead, and also ensures we 
don't call pa_operation_unref() on a null pointer.

When the card doesn't have a "Voice Call" profile and no usable sink, 
switching mode would result in the operation never returning.

Making sure the operation can be executed solves this issue.

pulse: assert valid CadOperation type

See merge request mobian1/callaudiod!1

This commit simplifies the way the output port is selected so we make 
sure we always use the highest priority available port, except in a few 
cases.

Fixes #3

Fixes #4

Fixes #1

Using a uint doesn't make sense here, while a boolean named `success` 
will be much more intuitive

Rest of the project uses 4 spaces, make it coherent.

As we now automatically check which port should be used as output, we no 
longer need to keep track of the earpiece port.

Forcing the earpiece port can be a wrong idea when e.g. the person is
wearing headphones: we want to keep outputting sound through the
headphones instead of suddenly switching to the earpiece.

This is done by retrieving the highest priority available port which
isn't the speaker. It requires priorities to be set right and having no
other output than speaker, earpiece or headphones, but for now it's a
resonable constraint.

For cards with a single profile, we still want to offer a voice call 
mode. This can be achieved by simply switching the output port to the 
earpiece.

In order to offer a voice call mode for devices with a single UCM 
profile (which can be a legitimate use case), we want to track whether 
the card has a voice call profile, as well as the earpiece/handset 
output port if it has one.

Instead of manually setting result code, calling the user callback and 
so on, we can just call operation_complete_cb().

Tracking the default port (which is actually the non-speaker active 
output port) can be tedious and prone to errors.

Instead of trying to track which port was in use when we switched to 
speaker, we can simply check, when it's time to switch back, which 
available port has higher priority and use this one.

This will also help solve corner cases: switch to call mode (earpiece), 
enable speaker, plug in headphones, then disable speaker: the current 
code will automatically switch back to headphones instead of the 
earpiece, which sounds like the good thing to do in that case.