segfaults on all .cin videos from Quake II.
patch by (Matti Hamalainen <mhamalai ratol fi>)
Originally committed as revision 3956 to svn://svn.ffmpeg.org/ffmpeg/trunk
This patch adds the support for INTRA PCM macroblocks in CAVLC and CABAC
mode, the deblocking needed a small modification and so did the
intra4x4_pred_mode prediction.
With this patch, the 5 streams of the conformance suite containing INTRA
PCM macroblocks now decode entirely, 4 are completely corrects, 1 is
incorrect since the first B slice because of deblocking in B slice not
yet implemented.
The code is not optimized for speed, it is not necessary IPCM
macroblocks are rare, but it could be optimized for code size, if
someone want to do this, feel free.
Originally committed as revision 3941 to svn://svn.ffmpeg.org/ffmpeg/trunk
3 more streams of the conformance suite decode to the
end instead of failing on an assert (CABACI3_Sony_B.jsv,
CABAST3_Sony_E.jsv and CABASTBR3_Sony_B.jsv), they are still false since
the first B slice probably because of deblocking in B slices not yet
implemented.
Originally committed as revision 3932 to svn://svn.ffmpeg.org/ffmpeg/trunk
which isn't the advanced one; and indeed, using adv. profile parser fails.
Using normal parser works, and that's what is done
- attempt at taking care of stride for NORM2 bitplane decoding
- duplication of much code from msmpeg4.c; this code isn't yet used, but
goes down as far as the block layer (mainly Transform Type stuff, the
remains are wild editing without checking). Unusable yet, and lacks the AC
decoding (but a step further in bitstream parsing)
patch by anonymous
Originally committed as revision 3931 to svn://svn.ffmpeg.org/ffmpeg/trunk
With this patch, the streams CAWP5_TOSHIBA_E.264 and CVWP5_TOSHIBA_E.264
are completely decoded
Originally committed as revision 3917 to svn://svn.ffmpeg.org/ffmpeg/trunk
- no more error with norm6 codes
- almost skipped frames (~110 bits) are not overflowed
However:
- there are very few bits left in those frames, so I wonder if enough MB
data is actually left
- NORM6/DIFF6 (4/5) bitplanes are a lot more present, which I don't find
obvious.
patch by anonymous
Originally committed as revision 3915 to svn://svn.ffmpeg.org/ffmpeg/trunk
vc9data.h to select the behaviour
- modified decode_b_picture_primary_header so that
decode_bi_picture_header is no more required
Fixes:
- modifications for some more using of MpegEncContext
- fixed a bit frame type decoding
- switch from pointers to index for AC tables
- corrected P CBPCY decoding (p_cbpcy += 4)
patch by anonymous
Originally committed as revision 3906 to svn://svn.ffmpeg.org/ffmpeg/trunk
- avctx and gb elements were removed from VC9Context, hence a larger diff
- some code was added to h263dec.c regarding CODEC_ID_WMV3 (should apply to CODEC_ID_VC9 too)
- VLC tables and other related tables were made global whenever this seemed necessary; appropriate changes were therefore made to other parts of the code using those tables
- the change for the bitplane management to a struct (some of them should eventually be mapped to MpegEncContext arrays) wasn't associated with the proper frees; should be fixed now
patch by anonymous
better names for globalized tables by me
Originally committed as revision 3905 to svn://svn.ffmpeg.org/ffmpeg/trunk
- i/p/b_decode_mbs functions now checks the proper values (from
bitplane). I didn't find it very clear, but it seems "raw" = "to read at
the MB layer". I used a structure, but we could spare that indirection and
add a flag in the VC9Context for each array.
- Properly decode (or so I think) the CBPCY (as it is predicted)
- many more things
patch by anonymous
Originally committed as revision 3882 to svn://svn.ffmpeg.org/ffmpeg/trunk
