Hi, august wrote > checking into this more, I made a little test app that only does > bgav_open and bgav_close on media (that exist and that do not). > > please see attached.
Ok, I checked this a bit, I played with several inputs and with valgrind --leak-check=full --track-fds=yes : * Existing local file: No leaks (fixed one in the mov demuxer though) * Non-existing local file: No leaks * http://wnycam.streamguys.com: Funny stream actually, they send an ID3V2 tag at the start of the stream. gmerlin-avdecoder reads and handles it :) I fixed 2 memory leaks (one in the URL parser, another in the ID3V2 parser), but didn't get file descriptor leaks * http url gotten redirected: Fixed one file descriptor leak and 2 or so memory leaks BTW test_openclose.c seems to be ok, I just changed the calls to load_file() with argv[1] to make it easier to use :) Now one can make a scripted leak test over a directory of random files :) BTW I consider gmerlin-avdecoder ready for testing again. One new feature is a native Matroska demuxer and webm support. For VP8 decoding one should use the latest ffmpeg (libvpx sucks). There might be some inaccuracies with respect to sample accurate seeking (especially in Vorbis tracks). I'm thinking to write a tool for testing sample accurate seeking. Burkhard ------------------------------------------------------------------------------ This SF.net email is sponsored by Sprint What will you do first with EVO, the first 4G phone? Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first _______________________________________________ Gmerlin-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/gmerlin-general
