I'm having some issues trying to clean up the memory allocations from using
libavformat and libavcodec, but I'm getting seemingly arbitrary segfaults when
doing so. In particular, I cannot call avcodec_close, or
av_close_input_stream, without generating a segfault.
The locations where these segfaults appear to be happening don't make much
sense to me. For avcodec_close, it claims to be faulting after this call
chain:
(I'm using static libav* libraries, linked into a dynamic JNI library, running
on my Android phone, so the debugging information is limited to running the
stack trace addresses through addr2line:)
#00 pc 00526724 av_freep /home/kutulu/workspace/ffmpeg/libavutil/mem.c:153
#01 pc 003e6304 avcodec_default_free_buffers
/home/kutulu/workspace/ffmpeg/libavcodec/utils.c:989
#02 pc 003e6c4c avcodec_close
/home/kutulu/workspace/ffmpeg/libavcodec/utils.c:729
According to this, the segfault is happening here:
for(j=0; j<4; j++){
av_freep(&buf->base[j]); // <--- here
buf->data[j]= NULL;
}
If I skip avcodec_close and just call av_close_input_stream, the stack trace
gives me:
#00 pc 0001066a /system/lib/libc.so
#01 pc 0000afae /system/lib/libc.so
#02 pc 00328aa4 av_parser_close
/home/kutulu/workspace/ffmpeg/libavcodec/parser.c:238
#03 pc 000beafc av_close_input_stream
/home/kutulu/workspace/ffmpeg/libavformat/utils.c:2452
which is here (though I have also seen the segfault on the previous line, so
the addresses might be off a bit):
if (s->parser->parser_close)
s->parser->parser_close(s);
av_free(s->priv_data); // <--- here
av_free(s);
I can't see why any of these things wouldn't be safe to free; I'm not doing
anything unusual with the contexts once I have them besides pass them to
av_read_frame and avcodec_decode_audio3.
Can anyone point me in the right direction?
--Mike
_______________________________________________
libav-user mailing list
[email protected]
https://lists.mplayerhq.hu/mailman/listinfo/libav-user
