Andre [Tue, 9 May 2006 17:25:12 +0000 (19:25 +0200)]
check barriers also in afs_send_chunk()
As this function is called even if the select timeout didn't expire,
we'll have to check the eof_barrier and the data_send barrier also
early in afs_send_chunk() and return if there's time left.
Andre [Sat, 6 May 2006 14:09:16 +0000 (16:09 +0200)]
grab: init gc->audio_format_num to -1
Previously it was 0, which happens to be the mp3 audio format number.
Therefore, grab without args only grabbed mp3 files. This patch makes
it grab any audio format.
Andre [Thu, 20 Apr 2006 18:56:04 +0000 (20:56 +0200)]
use git-tar-tree to make the tarball
This simplifies Makefile.in quite a bit, speeds up the tarball
creation, and reduces the size of the compressed tarball from 560KB
to 408KB, mostly because it no longer contains the screenshot images.
Andre [Wed, 19 Apr 2006 21:59:08 +0000 (23:59 +0200)]
Makefile.in: Remove special treatment of ortp_send/ortp_recv
they no longer depend on glib, so use the generic rule for building
object files. This made gcc spit out some trivial warnings about
unused variables which are also fixed in this patch.
Andre [Wed, 19 Apr 2006 20:35:04 +0000 (22:35 +0200)]
net.c: make it compile without ucred
This patch is ugly as hell and only compile-tested. It obviously
needs more work (or a totally different approch like using the usual
send/recv functions in case ucred is not available).
Andre [Wed, 19 Apr 2006 02:54:20 +0000 (04:54 +0200)]
NEWS: gcc-2.95 is no longer supported
The admin of the single remaining system I have access to which was
still running gcc-2.95 recently upgraded to gcc-3 (a version which
is merely 3 years old rather than the 7 years old gcc-2.95).
This means that upcoming paraslash releases will no longer be
compile-tested with gcc-2.95.
gcc-2.95
finally didn't manage to survive
it served all well for all the time
but no one really cares to whine
as we all meet in some other life
Andre [Tue, 18 Apr 2006 19:53:48 +0000 (21:53 +0200)]
split play.c and rename para_play to para_write
This separates the code to write an audio stream into several
independent pieces:
write.h: definition of writer-related structures
write.c: the para_write main program
write_common.c: functions to be shared between para_write
and para_audiod
write_common.h: exported functions of write_common.c
alsa_writer.c: writes the stream to an alsa sound device
file_writer.c: writes the stream to a file
Other writers (aka output plugins) can be added easily.
Andre [Tue, 18 Apr 2006 18:42:15 +0000 (20:42 +0200)]
add new writer to para_play: file
It simply writes the input to a random filename under ~/.paraslash.
Also, add some writer-related macros to config.h and two new options
for para_play: --writer and --list_writers.
Andre [Tue, 18 Apr 2006 03:28:55 +0000 (05:28 +0200)]
Further para_play abstraction: struct writer_node_group
This structure contains the list of writer nodes and information common to all
of these nodes. The idea is that the wng group code will be shared between
para_audiod and para_play.
Andre [Mon, 17 Apr 2006 18:44:22 +0000 (20:44 +0200)]
play.c: reorder and rename some functions.
Move everything which depends on ALSA to the three functions
alsa_init(), former set_alsa_params(), alsa_write(), former
pcm_write(), and alsa_shutdown().
Andre [Sun, 16 Apr 2006 17:56:09 +0000 (19:56 +0200)]
fix more struct timeval related warnings
Use tv2ms() which returns long unsigned where appropriate.
For the remaining cases, cast the members of struct timeval
to long unsigned if used via a format string.
Andre [Fri, 14 Apr 2006 13:31:48 +0000 (15:31 +0200)]
com_sa: pedantic mem leak fix.
In fact, this memory leak doesn't matter much because it happens in the child
which calls exit shortly afterwards. But it's considered good programming
practice to free all memory, so..
Andre [Mon, 10 Apr 2006 17:40:11 +0000 (19:40 +0200)]
improved version of split_args()
Thanks to Lorenzo Bettini for pointing this out.
This patch also fixes a bug with the default filter configuration which
allocated too little memory for the array of filter configurations. Now
we always allocate space for at least three entries.