On Sat, Nov 28, 2020 at 02:46:54PM -0500, Andriy Gelman wrote: > From: Andriy Gelman <andriy.gel...@gmail.com> > > We currently use the same interrupt_callback function for both muxers > and demuxers to break out of potential infinite loops. > > The function decode_interrupt_cb() checks for how many SIGINT/SIGTERM > interrupts have been received, and (usually) two interrupts are needed to > break out of an infinite loop. > > If this condition is seen on the muxer, however, we will fail to flush the > avio buffers (see retry_transfer_wrapper()). > An example of this issue is seen in Ticket #9009 (which would be fixed > by this patch). > > A more robust alternative maybe to break out of demuxers with one > interrupt. The error should propagate through, close the muxers and > allow them to flush properly. > (If the infinite loop is present on the muxer then a second interrupt would > cause it break out and have the same behavior as before.) > > This patch adds this behavior. I've labelled it RFC because it > potentially touches many demuxers. > --- > fftools/ffmpeg.c | 6 ++++++ > fftools/ffmpeg.h | 1 + > fftools/ffmpeg_opt.c | 2 +- > 3 files changed, 8 insertions(+), 1 deletion(-)
The idea is that a single SIG* would try to exit normally if you on a single one do interrupt the IO this could result in some corruption on the input side. If the goal is to improve the user experience and only requirte a single SIGINT even in light of infinite loops then some sort of automated staged interruption could be done. that is keep trak of progress being made. If the input is stuck and makes no progress for a period of time then a signal could enable interruption immedeatly similar for the output side. otherwise a signal would only break the main loop cleanly Thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Dictatorship naturally arises out of democracy, and the most aggravated form of tyranny and slavery out of the most extreme liberty. -- Plato
signature.asc
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
