Quoting Mark Thompson (2016-06-23 01:35:27) > On 21/06/16 18:50, Anton Khirnov wrote: > > This makes sure the actual stream parameters are used, which is > > important mainly for hardware decoding+filtering cases, which would > > previously require various weird workarounds to handle the fact that a > > fake software graph has to be constructed, but never used. > > This should also improve behaviour in rare cases where > > avformat_find_stream_info() does not provide accurate information. > > --- > > avconv.c | 245 > > ++++++++++++++++++++++++++++---------------------------- > > avconv.h | 11 +-- > > avconv_filter.c | 33 ++------ > > avconv_opt.c | 44 ---------- > > avconv_qsv.c | 1 - > > 5 files changed, 134 insertions(+), 200 deletions(-) > > Testing it I no longer need funny hwupload instances in vaapi transcode, so > yay! > > Probably ok? I'm not sure of the trickier implications around the > queueing - previously it would have been happy to initialise and run > even if some graph inputs aren't there when we start, but now it will > buffer the other streams indefinitely waiting for them?
Yes, I discussed this with Martin on IRC and we agreed that it should only be an issue in really pathological cases which would not work very sanely anyway, and the advantages of this patch are more important. -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
