FFmpeg 3.0.4

[eumagga0x2a]

We've missed again a chance to ship Avidemux with a current FFmpeg release. From the changelog:

Code Select Expand

+version 3.0.4:
+- avformat/avidec: Check nb_streams in read_gab2_sub()
+- avformat/avidec: Remove ancient assert
+- avformat/avidec: Fix memleak with dv in avi
+- lavc/movtextdec.c: Avoid infinite loop on invalid data.
+- avcodec/ansi: Check dimensions
+- avcodec/cavsdsp: use av_clip_uint8() for idct
+- avformat/movenc: Check packet in mov_write_single_packet() too
+- avformat/movenc: Factor check_pkt() out
+- avformat/utils: fix timebase error in avformat_seek_file()
+- avcodec/g726: Add missing ADDB output mask
+- avcodec/avpacket: clear side_data_elems
+- avformat/movenc: Check first DTS similar to dts difference
+- avcodec/ccaption_dec: Use simple array instead of AVBuffer
+- avformat/mov: Fix potential integer overflow in mov_read_keys
+- swscale/swscale_unscaled: Try to fix Rgb16ToPlanarRgb16Wrapper() with slices
+- swscale/swscale_unscaled: Fix packed_16bpc_bswap() with slices
+- lavf/utils: Avoid an overflow for huge negative durations.

https://github.com/FFmpeg/FFmpeg/commit/b9a1d389b2a05269042d013cf6da4a09ce6ccc24

[mean]

Only relevant one is this set

avformat/movenc: Check packet in mov_write_single_packet() too
avformat/movenc: Factor check_pkt() out
avformat/movenc: Check first DTS similar to dts difference

The other ones are not used at all

[eumagga0x2a]

That's true, there is not really much behind an incremented patch level digit. I've created a local git branch for ffmpeg-3.0.4 anyway, testing the first build with 3.0.4, looks good so far.

Off-topic: when encoding using the Mpeg2 (ff) encoder, the terminal gets flooded (one line for each frame?) with messages from libavformat/mpegenc.c:948 "buffer underflow st=%d bufi=%d size=%d\n", like

Code Select Expand

[adm_lavLogCallback]  [lavc] buffer underflow st=0 bufi=26180 size=41616

This is not specific to ffmpeg-3.0.4, of course.