Changes in short:
* Re-work of session handling to use sound and ISDN threads, which
greatly reduces latency and CPU usage
* Complete re-work of ISDN subsystem to use CAPI 2.0
* Complete re-work of sound subsystem to use ALSA
* Re-work of recording subsystem to produce better results
* Re-work of debug and error message handling (centralized)
* Added --sleep and --wakeup remote commands to facilitate release
and re-acquire of CAPI connection
* Code documentation (partial)
Reduced latency:
----------------
We are now using separate threads to handle ISDN input (and send it to audio
output) and audio input (which again sends it to ISDN). This means, we are
not dependent on GTK GUI reaction time. This enables very good sound quality
with low latency and low CPU consumption (normally not measurable).
Tested with 3x parallel "while true; do true; done" and "find /". No sound
problems, even though the threads don't run with real time priority (which
can be implemented now relatively easily, though).
ISDN system:
------------
We now support CAPI 2.0. This means, any card supported by CAPI 2.0 should
work and also ant-phone should be able to coexist with other ISDN
applications.
New CAPI code was tested with Fritz!Card, both with fcpci driver and with
mISDN driver.
There are also remote-CAPI implementations which forward local CAPI requests
to a remote host (e.g., a Fritz!Box router). They should be also usable with
new CAPI subsystem (not tested).
Sound system:
-------------
Since we are now using ALSA, it's much easier to coexist with other sound
applications. However, sound device specification has been changed. You
should now type in ALSA names (normally 'default') instead of raw devices.
Recording:
----------
Recording subsystem now uses ring buffers and the buffers are flushed in GTK
thread on timeout. Although it would be possible to have special thread for
this purpose, it doesn't really make sense, since we buffer about 2 seconds.
In 2 seconds, the application better responds... Anyway, in the worst case,
there will be some samples skipped. Also, recording tries to stitch together
incoming samples on local as well as remote channels as it best fits based on
time code. Certain jitter is compensated for automatically, bigger jitter may
produce some cracks. However, in tests there was no such problem.
Debug and error message handling:
---------------------------------
All debug and error messages should now go through macros dbgprintf() and
errprintf(). This allows for implementing of debug log viewer later.
For new/rewritten code, all debug and error messages start with "COMPONENT: "
prefix. This allows for easier identification in logs.
Additionally, there are further debug levels, up to 4. Debug level 3 and
higher prints very noisy information about each audio packet processed.
Sleep and wakeup commands:
--------------------------
Some ISDN card drivers need to be unloaded before suspending the machine to
disk and loaded again after resume. This is not possible, as long as the
application is holding the CAPI connection. Therefore, two new command-line
commands have been implemented, which allow turning off and on the CAPI
connection (--sleep and --wakeup, respectively). They can be integrated into
suspend and resume scripts to take care of stopping and resuming CAPI.
Code documentation:
-------------------
Documentation for new code and for big rewritten parts has been changed to
Doxygen format. Instead of plain comments, you'll see comments in form
/*! ... */ with tags in form @tag, which allow Doxygen to generate
pretty-printed source code reference documentation. Doxygen script file
has not been actually produced. The rest of the files should be changed
as well to use Doxygen format tags in documentation of functions and
structures.
Ivan Schreter