Age | Commit message (Collapse) | Author |
|
CVS patchset: 1396
CVS date: 2002/01/13 15:30:39
|
|
CVS patchset: 1387
CVS date: 2002/01/10 21:30:10
|
|
dxr3 decoder only functions if the dxr3 video_out driver loaded.
xine_t changed.
CVS patchset: 1381
CVS date: 2002/01/09 22:33:03
|
|
Replaced SETPTS calls now that I understand them.
Always call SETPTS with the vpts just before a WRITE.
If PTS=0, call SETPTS with a vpts = 0 ;
CVS patchset: 1372
CVS date: 2002/01/07 23:36:37
|
|
CVS patchset: 1369
CVS date: 2002/01/07 21:26:04
|
|
CVS patchset: 1347
CVS date: 2002/01/05 22:06:44
|
|
seem to need it, and it'll fix the problem with disappearing stills
(menu's, copyright notices) introduced by the still frame change.
Needs more testing, obviously.
CVS patchset: 1303
CVS date: 2001/12/24 19:31:13
|
|
types.
(the file is .overlay/resXxXxX, generated by em8300's autocal)
this was probably introduced sometime between xine 0.9.4 and 0.9.7
in xine-ui. Fix is to temporarily set the locale to "en_US".
CVS patchset: 1302
CVS date: 2001/12/24 18:40:13
|
|
the flushing by video_out.c causes the dxr3 to drop the still image.
consider dxr3 still menu support to be broken for now.
CVS patchset: 1301
CVS date: 2001/12/24 16:31:57
|
|
CVS patchset: 1300
CVS date: 2001/12/24 16:01:15
|
|
more like the old behaviour and seems good enough for av sync
- tried improving flush to get out of stuck /dev/em8300_mv writes; not
entirely succesful :-) commented out.
CVS patchset: 1293
CVS date: 2001/12/23 18:13:35
|
|
CVS patchset: 1289
CVS date: 2001/12/23 04:27:33
|
|
the kernel driver ignores SETSCR commands when the difference is
smaller than 7200 (a very arbitrary value of 2 frames). Since my
favourite oss audio driver has a GAP_TOLERANCE of 5000, this resulted
in about a gazillion 'audio out: adjusting master clock' messages.
Now added an offset to simulate adjustments < 7200; as soon as the
offset goes over 7200 SETSCR is called and offset set to 0.
CVS patchset: 1288
CVS date: 2001/12/23 04:08:25
|
|
- mpeg playback optionally syncs (SETPTS) every frame; works only for
constant frame duration (PAL movie) not most NTSC movies.
metronom doesn't seem to dig varying durations, gives pts values that
upset the dxr3 hardware.
- detect repeat first field in mpeg, disable sync every frame if found.
- small dxr3 encoder fixes
- moved fame encoding from copy to display method; should help with
frame skip.
- added callback functions for some config/dxr3* variables
(those callbacks are cool BTW; only wish the xine-ui setup screen would
show which vars have callbacks)
- added Dan Hollis field swapping trick; not sure if it really does what
he wants; I don't have an NTSC tv, so his 720x480 test video will look
interlaced no matter what.
- merry xmas
CVS patchset: 1287
CVS date: 2001/12/23 02:36:54
|
|
subtitles.
CVS patchset: 1275
CVS date: 2001/12/19 01:53:43
|
|
probably need some tweaking...
CVS patchset: 1268
CVS date: 2001/12/17 22:28:33
|
|
- minor change in dxr3_decoder.c
CVS patchset: 1267
CVS date: 2001/12/17 16:20:42
|
|
(the old cannot-play-a-2nd-avi-in-the-same-xine-session bug)
CVS patchset: 1260
CVS date: 2001/12/16 22:34:21
|
|
libfame and librte encoders.
CVS patchset: 1257
CVS date: 2001/12/16 19:06:49
|
|
decoding if using the dxr3 decoder plugin, uses on-the-fly mpeg
encoding otherwise (provided encoding support is compiled in).
some remarks:
- dxr3enc is no more. I've added some transition code in loadplugins.c
(look for the IGNORE_DXR3ENC) to prevent loading a stale dxr3enc plugin
from a previous install and to print a message if someone tries to
run xine -V dxr3enc.
- small updates to configure.in and _xine.m4. Mostly about the messages,
no new checks or anything.
- both dxr3_vo_standard.c and dxr3_vo_encoder.c are no more. The one and
only dxr3 video out driver is aptly named dxr3_video_out.c
CVS patchset: 1256
CVS date: 2001/12/16 19:05:44
|
|
If it isn't, it will report 0 capabilities, thus giving libmpeg2
a chance.
For users this means you can finally say xine -V Xv dvdnav://:main
to use libmpeg2 + Xvideo
and xine -V dxr3 dvdnav://:main to use dxr3 hardware decoding.
Previously, it used the dxr3 decoder plugin no matter which output plugin
was active.
CVS patchset: 1249
CVS date: 2001/12/15 20:56:21
|
|
Good news is that we're now using the new rte API almost 100%
(exception is an omission in the current new rte API, see FIXME)
CVS patchset: 1234
CVS date: 2001/12/13 03:41:31
|
|
at all mpeg framerates, not just 25 Hz :-)
CVS patchset: 1232
CVS date: 2001/12/13 00:36:32
|
|
CVS patchset: 1221
CVS date: 2001/12/11 15:30:05
|
|
easy to split into core + fame + rte
- now official support for librte-0.4 (zapping.sf.net); updated configure
scripts, checks for librte added. Possible to compile both librte and libfame
support in at the same time, configuration via .xinerc.
- fixed YUY2 decoding, both for libfame and librte.
Note: unifying dxr3enc (non-mpeg) and dxr3 (mpeg) seems rather easy
right now. I think I'll wait with it for a bit, because perhaps the
dxr3-dvd fixes need major replumbing in the dxr3 vo driver (but I don't
expect that; all mpeg stuff is done in the dxr3 decoder plugin)
CVS patchset: 1220
CVS date: 2001/12/11 02:26:58
|
|
(zapping.sf.net). librte is a wrapper for the mp1e backend.
This works very well. I'm considering throwing out support
for libfame (current default) and libffmpeg to be able to clean
up the code. It's now a tangled mess of defines...
Read the comments at the top of dxr3_vo_encoder.c to find out
how to enable librte.
CVS patchset: 1161
CVS date: 2001/12/02 21:14:51
|
|
CVS patchset: 1159
CVS date: 2001/12/02 07:10:02
|
|
seemingly not doing its job.
CVS patchset: 1158
CVS date: 2001/12/02 07:08:59
|
|
libfame). There's a massive A/V sync problem because the mpeg is
encoded and sent in the frame_copy function. It should be done in the
display_frame one, but that currently gives buffering problems that
upset the dxr3 card. define MP1E_DISPLAY_FRAME to 1 the source and see
for yourself... Perhaps Mike can conjure up another magic dxr3 register
variable to make things right like he did last time :-)
CVS patchset: 1157
CVS date: 2001/12/02 03:40:27
|
|
to libfame...
CVS patchset: 1152
CVS date: 2001/12/01 19:37:21
|
|
CVS patchset: 1151
CVS date: 2001/12/01 19:36:30
|
|
for mp1e sources though. libfame encoder still the default.
CVS patchset: 1150
CVS date: 2001/12/01 19:32:44
|
|
CVS patchset: 1133
CVS date: 2001/11/29 07:17:07
|
|
it still throws off metronom so audio is missing for a while...
CVS patchset: 1113
CVS date: 2001/11/26 01:12:00
|
|
now removed.
CVS patchset: 1111
CVS date: 2001/11/25 21:13:15
|
|
metronom/demux/whatever behaviour. It works pretty well for me now,
but definately needs more testing.
I've tagged the old style (as of xine 0.9.5) as DXR3_095 prior
to commit these changes.
CVS patchset: 1110
CVS date: 2001/11/25 20:21:25
|
|
CVS patchset: 1102
CVS date: 2001/11/24 11:52:19
|
|
CVS patchset: 1101
CVS date: 2001/11/24 11:09:30
|
|
- outline top black border at 16 lines, so old and new
macroblocks overlap (might reduce re-encoding artefacts)
- print libfame's quantizer setting (based on quality parameter)
divx4:
- removed version check out of init stage to decoding stage, so as to
cause minimal annoyance (you need to increase divx4 priority to get
there)
- re-enabled decore release command in close; this is vital now that
xine closes and reopens the decoder when seeking (when did that happen?)
otherwise we get 5M memory per seek
- upped the version requirement to oktober 2001 (was august). This due
to bug reports that decore release caused sigsev. (that's why release
was disabled till now)
read updated README.divx4 for more info.
CVS patchset: 1095
CVS date: 2001/11/21 20:40:47
|
|
is now under its own banner (dxr3enc).
CVS patchset: 1078
CVS date: 2001/11/19 17:07:15
|
|
CVS patchset: 1077
CVS date: 2001/11/19 15:06:12
|
|
CVS patchset: 1066
CVS date: 2001/11/18 09:41:32
|
|
CVS patchset: 1065
CVS date: 2001/11/18 08:25:46
|
|
acceleration?). Merge xine-utils header files to a new one "xineutils.h".
Update xine-lib C/headers to reflect those changes.
dxr3 headers are no more installed ine $includdir, but $includdir/xine.
CVS patchset: 1054
CVS date: 2001/11/17 14:26:36
|
|
dxr3_flush, as it doesn't help...
CVS patchset: 1034
CVS date: 2001/11/14 17:37:02
|
|
CVS patchset: 1033
CVS date: 2001/11/14 15:17:01
|
|
a bug that prevented it, but Vivien Chappelier has fixed that now
- removed unused fame mpeg1 object
CVS patchset: 1026
CVS date: 2001/11/12 23:56:31
|
|
CVS patchset: 1014
CVS date: 2001/11/10 19:15:50
|
|
CVS patchset: 1008
CVS date: 2001/11/10 06:20:39
|
|
CVS patchset: 1001
CVS date: 2001/11/09 07:20:44
|