As the libosmogapk actually relies on external libraries for
audio coding, we should enable / disable particular codec
tests depending on the build configuration.
Closes: OS#2926
Change-Id: Ie4711294c43ff88b17431615883abf96d1ae02a6
This change fixes parallel building problem, when osmo-gapk was
being compiled before its libosmogapk dependency:
make[2]: *** No rule to make target '../src/libosmogapk.la',
needed by 'osmo-gapk'. Stop.
make[2]: *** Waiting for unfinished jobs....
For some reason, automake ignores a dependency if the full
path is provided:
$(top_builddir)/src/libosmogapk.la
while the relative path solves the problem:
libosmogapk.la
Closes: OS#2907
Change-Id: I3fdd1731bd372bbb42fe57981e757386e8ede0f0
The previous GAPK implementation was represented by a single
executable. So, all audio transcoding operations were available
only via calling the 'gapk' binary. This approach didn't allow
external applications to benefit from using GAPK API directly.
The following set of changes separates the common code into a
shared library called 'libosmogapk', linking the 'gapk' binary
against it:
- 95e6664 Introduce a shared 'libosmogapk' library
- 30209ce Install GAPK headers to '${includedir}/osmocom/gapk/'
- a8d4657 Add an 'osmo_gapk' prefix to the exposed symbols
- 40d59f1 Add a pkg-config manifest for libosmogapk
- 4f0a47d Add the symbol export map for libosmogapk
All memory management operations are now based on talloc library:
- 3c20dac libosmogapk: use talloc for memory management
- 5cabe1e osmo-gapk: use talloc for memory management
Integrated Osmocom logging framework:
- c35ba8a libosmogapk: use Osmocom logging framework
- 4b7cd2c osmo-gapk: drop useless printf calls
- 0fe18af osmo-gapk: use Osmocom logging framework
- 11943bf osmo-gapk: adjust application verbosity
Integrated GNU Autotest environment and basic test coverage:
- f069eb3 Init automake test environment
- 1fe6a9b tests: add procqueue test
- 3e9e57f tests: add pq_file test
- 9d2b15d tests: add pq_rtp test
- f59f3f1 tests: add format / codec transcoding tests
For more details, see commits history.
Change-Id: I3c6d4a9d326ee49153e4ad83823d094831c112da
Let's use the common string representation for item category
names, defined in the shared header, instead of defining
them in every file.
Change-Id: Ie0c449d77fa383cad27f67b8ce902bd071342dbb
Abusing the talloc hierarchical nature may cause some problems,
e.g. on embedded systems with emulated talloc API. Let's release
the memory allocated for a state explicitly.
Change-Id: Ie675a92b1e52a4886dc447af19f65ff5e12a4c40
The BENCHMARK_STOP should be called with a correct codec type and
a correct operation type (encode or decode). Otherwise the results
could be incorrect.
Change-Id: Ie90e85ca8d9ec3175a58dde60525e0b7d6daf608
During the HR codec initialization, a part of the state is
allocated, but not freed at exit. Let's fix this.
Found during debugging with Valgrind:
4,932 bytes in 1 blocks are definitely lost in loss record 177 of 179
at 0x4C2AB80: malloc (in vgpreload_memcheck-amd64-linux.so)
by 0x6381C8F: gsmhr_init (libgsmhr.c:63)
by 0x526DF62: osmo_gapk_pq_queue_codec (pq_codec.c:48)
by 0x401ACE: make_processing_chain (app_osmo_gapk.c:573)
by 0x401ACE: main (app_osmo_gapk.c:765)
Change-Id: Ie9ead89c2272782de8d928f29753e6a523cf8834
This test group is intended to check the format / codec transcoding
capabilities of the library. The reference files are used to ensure
that encoding or decoding was successful.
The following formats are currently being tested:
- amr_efr
- gsm
- racal_hr
- racal_fr
- racal_efr
- ti_hr
- ti_fr
- ti_efr
- rtp_efr
- rtp_hr_etsi
- rtp_hr_ietf
This test is intended to check the RTP source / sink operability.
To do this, two processing queues are being allocated:
"generator": source/random -> sink/rtp
"checker": source/rtp -> sink/checker
The first one generates some amount of random bytes (payload),
and stores them inside a buffer that is shared between both
queues.
After generation, a payload is being sent from the first
queue via an RTP sink, and then being received by the second
via an RTP source.
As both queues do use a shared buffer, the last item of the
second queue (named 'sink/checker') is able to compare a
received payload with expected.
This test is intended to check the file source / sink
operability. To do that, the following processing chain
is being composed:
source/file -> proc/dummy -> sink/file (stdout)
The source item opens the sample file named 'io_sample.txt'
for reading. The next processing item simply converts all
uppercase latters to the lowercase. The last one writes
the result to stdout.
This processing cycle is being repeated several times
with different block length values.
This test is intended to validate the processing queue
management API. Moreover, the talloc debugging API is
used to ensure that there are no memory leaks.
First, four processing queues are being allocated. One
of them is empty, while others have different count of
items. Then the human-readable description is being
generated for all of them. And finally, the processing
and exit cllback are being tested.
During the test execution, the talloc NULL-context
tracking feature is enabled, allowing to observe every
memory allocation within the libosmogapk, and to detect
memory leaks.
The talloc_size() call sets the current file name and the current
line number as name for chunk being allocated. This combination
is not so informative during debugging, so let's use the static
'.buffer' string as context name for item's output buffer.
In order to give advanced control over a processing queue,
it would be better to have the checking function separated from
the osmo_gapk_pq_prepare(). Moreover, this change introduces an
additional 'strict' checking mode that requires a queue to have
a source item first and a sink item in the last position.
This change adds two meta-information fields to the processing
queue item structure. Both of them will be used for more
detailed logging and for the human-readable processing
queue description.
There are currently three types of prcessing queue items:
- source (file, alsa, rtp)
- proc (format, codec)
- sink (file, alsa, rtp)
Let's assign corresponding type for each item.
This would facilitate logging and the queue checking.
Since this change, every processing queue may optionally have
an associated human-readable name. If name is not required,
NULL should be passed to the osmo_gapk_pq_create().
In order to simplify memory leak debugging, this change introduces
the library's internal talloc context that may be changed by
external application by calling the osmo_gapk_set_talloc_ctx().
There are not so much code, related to internal logging subsystem.
So, there is no reason to keep a few lines in a dedicated file.
In the future one may also be used for other routines.
Previously the osmo-gapk application used to exit as soon as all
the frames are processed, no matter has the sink finished its
internal processing (e.g. ALSA playback).
In some cases it's required to wait for some queue items
to finish processing. For example, the ALSA sink writes the
audio samples to the buffer in non-blocking mode, so as soon
as all of them will be written, a program may finish execution,
causing the playback abort.
To prevent that, this change extends the library's API, allowing
each queue item to have a processing state callback that returns
a positive integer if processing is not finished yet,
and 0 otherwise.
Instead of immediately shutting down the application, it is
better to try to break the processing queue first, and stop
the execution immediately if second SIGINT is received.
The printf() writes the text into stdout, which may be undesirable
in some use cases. Moreover, the printed information was redundant.
So, let's drop such calls.
Since this change, the libosmogapk uses the Osmocom logging
framework. By default, logging is disabled and could be enabled
by the external applications calling the osmo_gapk_log_init()
with a desired log target as an argument.
Since GAPK package contains a library and the representative
osmo-gapk application, the 'main.c' looks a bit confusing. Let's
use the common naming scheme.
The usage of linuxlist is more flexible than having a limited
array of pointers. This approach allows to have as much items
in a processing queue as required.
To simplify the benchrarking process via the library API, this
change introduces two new functions, which are intended to
provide total cycle and frame count.
Having statically allocated memory for benchmark data of every
codec causes high memory usage, especially if actual benchmarking
is not required for a particular use case. Instead of that, let's
provide an optional opportunity to enable benchmarking for a
particular codec by calling the osmo_gapk_bench_enable(). The
required amount of memory would be allocated, and then can be
freed by calling the osmo_gapk_bench_free() or manually.
The usage of a 'static inline' function definition in the 'bench.h'
is resulting in separate independent function definitions in each
translation unit from which the header is included. This is
increasing the size of compiled code unnecessarily.
There is no need to expose the implementation details of both
BENCHMARK_START and BENCHMARK_STOP macros via public header.
This change moves them to a separate private header 'bench.h'.
The benchmark_dump() is only used by the osmo-gapk binary, and
is intended to prepare and print benchmarking results to stderr,
what is most likely unusable for the library users.