Age | Commit message (Collapse) | Author |
|
- remove duplicated frame parsing code
- parse each mpeg audio frame
=> generate accurate pts, the viz plugin hack to get things in sync is no more needed
- parse the full xing header
=> compute the length (in s) of the stream without "magic" stuff
=> use the Xing TOC to seek "accurately" in vbr streams.
- prepare id3v2 tags parsing
I think this big change was needed.
Don't hesitate to blame me for bugs.
CVS patchset: 5420
CVS date: 2003/09/28 14:16:00
|
|
recognized as a mp3 header and the bitrate was computed from this identifier !
- Detect all id3v2 version, not only 2.3
- 1 kbit is 1000 bits, not 1024
- remove all funny BE_32() clones
- some cleanups
CVS patchset: 5410
CVS date: 2003/09/22 23:16:14
|
|
CVS patchset: 5401
CVS date: 2003/09/17 23:57:15
|
|
CVS patchset: 5369
CVS date: 2003/09/13 19:46:53
|
|
It's still a ugly hack but it should work better now.
CVS patchset: 5359
CVS date: 2003/09/13 11:50:06
|
|
CVS patchset: 5358
CVS date: 2003/09/13 00:49:07
|
|
It should fix the audio lag reported on xine-devel.
I've tested with these files:
ftp://mplayerhq.hu/MPlayer/samples/avi/nandub-desync/*.avi
http://www.vis.uni-stuttgart.de/~hopf/tmp/tst3.avi
and everything is ok ;-)
CVS patchset: 5352
CVS date: 2003/09/10 23:11:53
|
|
that the RHS is a int32_t which then needs expanding
to int64_t. The expansion takes the high bit of the int32_t,
which is the sign bit, and sets all the higher bits to the same sign.
This in effect turns the result into a negative number, when it
should just be a large positive int64_t number.
This fix should probably be duplicated in all other demuxers, but I
don't have time for that.
CVS patchset: 5350
CVS date: 2003/09/10 00:01:59
|
|
- Handle packet "sequence" number.
Streaming servers used to discard packets at overrun, it was not detected, causing some bad artifacts and crashes.
Now, when a bad sequence is detected, the decoder is reseted and the demuxer waits for a keyframe before sending data to the decoder.
This should fix problems with live streams recorded by xine or mmsclient.
I've tested with mms TV links from gxine mediamarks.
CVS patchset: 5349
CVS date: 2003/09/07 20:10:59
|
|
CVS patchset: 5344
CVS date: 2003/09/05 00:12:04
|
|
CVS patchset: 5342
CVS date: 2003/09/04 17:46:00
|
|
(for Details see manpage of ogmtools)
CVS patchset: 5334
CVS date: 2003/09/03 13:41:10
|
|
CVS patchset: 5333
CVS date: 2003/09/02 18:59:06
|
|
.WAV files with badly contructed headers.
CVS patchset: 5325
CVS date: 2003/09/01 13:19:28
|
|
Info from www.id3.org
CVS patchset: 5320
CVS date: 2003/08/31 16:37:32
|
|
Make sure it waits for a "program_stream_pack_header"
before demuxing anything else.
CVS patchset: 5315
CVS date: 2003/08/28 16:42:07
|
|
Refactor identical code into a function.
CVS patchset: 5314
CVS date: 2003/08/28 16:10:16
|
|
fft*post are untouched (fft: for now).
CVS patchset: 5310
CVS date: 2003/08/25 21:51:37
|
|
CVS patchset: 5296
CVS date: 2003/08/19 12:43:46
|
|
The demuxer was zeroing 8 bytes of every frame output.
This seemed not to matter for software decode, so maybe the
liba52 is not doing CRC checks as it should.
CVS patchset: 5295
CVS date: 2003/08/16 20:15:55
|
|
CVS patchset: 5285
CVS date: 2003/08/12 20:00:34
|
|
CVS patchset: 5284
CVS date: 2003/08/12 19:44:35
|
|
CVS patchset: 5281
CVS date: 2003/08/12 18:41:08
|
|
CVS patchset: 5268
CVS date: 2003/08/10 16:11:05
|
|
CVS patchset: 5267
CVS date: 2003/08/10 16:03:21
|
|
Any PMT with stream id's >= 0x80 are now handled correctly.
It currently only handles AC-3 or A52, but more can be easily added when needed.
CVS patchset: 5264
CVS date: 2003/08/09 23:01:58
|
|
CVS patchset: 5261
CVS date: 2003/08/09 19:27:35
|
|
CVS patchset: 5260
CVS date: 2003/08/09 17:50:16
|
|
CVS patchset: 5257
CVS date: 2003/08/07 09:26:53
|
|
and hence the correct length of songs, fix calculation of the bitrates (with 1 kbps = 1024 bps)
CVS patchset: 5254
CVS date: 2003/08/05 18:23:56
|
|
CVS patchset: 5229
CVS date: 2003/08/01 17:06:32
|
|
CVS patchset: 5225
CVS date: 2003/07/30 22:00:42
|
|
CVS patchset: 5223
CVS date: 2003/07/27 23:35:44
|
|
CVS patchset: 5211
CVS date: 2003/07/25 21:02:04
|
|
CVS patchset: 5208
CVS date: 2003/07/22 20:36:51
|
|
2) improve accuracy of time based seeking in long files
CVS patchset: 5207
CVS date: 2003/07/22 20:14:09
|
|
CVS patchset: 5205
CVS date: 2003/07/21 19:28:05
|
|
Fix playback of this stream :
http://cubeonline23.sourceforge.net/kirbylan.avi
CVS patchset: 5194
CVS date: 2003/07/19 19:11:45
|
|
2) support files containing multiple audio/video streams
CVS patchset: 5190
CVS date: 2003/07/19 16:11:39
|
|
accepted the input (fixes high seek lateny with DVDs)
CVS patchset: 5185
CVS date: 2003/07/19 11:57:29
|
|
CVS patchset: 5181
CVS date: 2003/07/19 00:18:49
|
|
add checks to prevent segfaults when max number of streams is exceeded
CVS patchset: 5176
CVS date: 2003/07/18 23:34:36
|
|
CVS patchset: 5173
CVS date: 2003/07/17 16:07:52
|
|
CVS patchset: 5168
CVS date: 2003/07/16 14:14:17
|
|
CVS patchset: 5167
CVS date: 2003/07/16 00:52:45
|
|
CVS patchset: 5165
CVS date: 2003/07/16 00:13:55
|
|
avi format definitely sucks ;)
CVS patchset: 5164
CVS date: 2003/07/14 18:49:13
|
|
now xine can play everything in this directory except tm20.avi (duck true motion 2) :
ftp://ftp.mplayerhq.hu/MPlayer/samples/testsuite
CVS patchset: 5153
CVS date: 2003/07/12 18:42:44
|
|
I.E. PES packets larger than fifo buffer size.
CVS patchset: 5144
CVS date: 2003/07/12 00:45:04
|
|
stream id), the last two bytes are not reliable
CVS patchset: 5143
CVS date: 2003/07/11 20:32:34
|