ffprobe and ffmpeg functions return different last frame for same file

[hxuanyu]

I met an issue that the output of ffprobe -show_frames seems different from what av_read_frame+avcodec_decode_video2 ouputs

The ffprobe and ffmpeg libs were built from same latest code.

I used ffprobe -show_frames -select_streams v input to print out all frame info. and here's the info of last 2 frames

[FRAME]
media_type=video
key_frame=0
pkt_pts=N/A
pkt_pts_time=N/A
pkt_dts=167501
pkt_dts_time=167.501000
pkt_duration=N/A
pkt_duration_time=N/A
pkt_pos=103794518
pkt_size=74
width=1920
height=1080
pix_fmt=yuv420p
sample_aspect_ratio=1:1
pict_type=P
coded_picture_number=4016
display_picture_number=0
interlaced_frame=0
top_field_first=0
repeat_pict=0
FRAME
[FRAME]
media_type=video
key_frame=0
pkt_pts=N/A
pkt_pts_time=N/A
pkt_dts=167542
pkt_dts_time=167.542000
pkt_duration=N/A
pkt_duration_time=N/A
pkt_pos=103794920
pkt_size=74
width=1920
height=1080
pix_fmt=yuv420p
sample_aspect_ratio=1:1
pict_type=P
coded_picture_number=4017
display_picture_number=0
interlaced_frame=0
top_field_first=0
repeat_pict=0
FRAME

on the other hand, our application uses av_read_frame and avcodec_decode_video2 to read a avpacket and decode into a frame. Below are some traces

Line 100: 00:00:33.384 MAIN rhbc73 process packet, Packet.dts = 0, Packet.pts = -9223372036854775808
Line 101: 00:00:33.384 MAIN rhbc73 process packet, Packet.dts = 42, Packet.pts = -9223372036854775808
Line 102: 00:00:33.400 MAIN rhbc73 process packet, Packet.dts = 83, Packet.pts = -9223372036854775808
Line 103: 00:00:33.400 MAIN rhbc73 process packet, Packet.dts = 125, Packet.pts = -9223372036854775808
Line 104: 00:00:33.400 MAIN rhbc73 process packet, Packet.dts = 167, Packet.pts = -9223372036854775808
Line 105: 00:00:33.400 MAIN rhbc73 -- get a full frame, pFrame->pkt_dts = 0, pFrame->pkt_pts = -9223372036854775808, BestEffortTimeStamp = 0, pFrame->coded_picture_number = 0
Line 107: 00:00:33.400 MAIN rhbc73 process packet, Packet.dts = 209, Packet.pts = -9223372036854775808
Line 108: 00:00:33.400 MAIN rhbc73 -- get a full frame, pFrame->pkt_dts = 42, pFrame->pkt_pts = -9223372036854775808, BestEffortTimeStamp = 42, pFrame->coded_picture_number = 1
Line 110: 00:00:33.400 MAIN rhbc73 process packet, Packet.dts = 250, Packet.pts = -9223372036854775808
Line 111: 00:00:33.416 MAIN rhbc73 -- get a full frame, pFrame->pkt_dts = 83, pFrame->pkt_pts = -9223372036854775808, BestEffortTimeStamp = 83, pFrame->coded_picture_number = 2

Line 12140: 00:00:51.168 MAIN rhbc73 process packet, Packet.dts = 167501, Packet.pts = -9223372036854775808
Line 12141: 00:00:51.168 MAIN rhbc73 -- get a full frame, pFrame->pkt_dts = 167334, pFrame->pkt_pts = -9223372036854775808, BestEffortTimeStamp = 167334, pFrame->coded_picture_number = 4012
Line 12143: 00:00:51.184 MAIN rhbc73 process packet, Packet.dts = 167542, Packet.pts = -9223372036854775808
Line 12144: 00:00:51.184 MAIN rhbc73 -- get a full frame, pFrame->pkt_dts = 167376, pFrame->pkt_pts = -9223372036854775808, BestEffortTimeStamp = 167376, pFrame->coded_picture_number = 4013

so from this we can see last frame has different pkt_pts and coded_picture_number from the output of ffprobe.

The file I tested is here ​https://dl.dropboxusercontent.com/u/89678527/Threw%20It%20On%20The%20Ground.avi

[Stefano Sabatini]

If you show no application code there is no way to reproduce the issue.
For example, can you reproduce the issue using some of the programs in doc/examples?

Also, are you correctly flushing the decoder at the end of the demuxing loop?

Please reopen if you can provide the missing information.