> I'm not really a fan of file-wide crc/md5s. I prefer per-frame crc/md5 > values, where seeking isn't an issue, but that only works for decoding > tests, not encoding tests... Also there's the never-resolved issue of > how to do regression testing on metadata reading... Not to mention > subtitles. > > What do you have in mind for encoding tests instead of md5s?
I saw that h264_mp4toannexb_bsf.c has no coverage. (Sure, a Code-In student might see it, but it seems unlikely they'll know how to test it.) The code file is small but I have a test that covers about 80% of it. In this case, the output from 'md5:' does match the MD5 of what gets laid down on disk as out.mpg (no seeking involved in the write, I guess). But I don't think we should be afraid of writing files via FATE tests. Alternatively, we create an internal output target that simulates a file entirely in memory. -- -Mike Melanson _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel