[FFmpeg-devel] [PATCH 0/5] Interruptibility enhancements

[Andrey Utkin]

This patchset is inspired by issue of slow reaction of libavformat to user
interrupt request, when operating on a file situated on a slow storage (e.g.
network share) - for example, when user wants to quit current file
playback, or seek to another position.
Order of actions in retry_transfer_wrapper() changed, so it checks for
interrupt before going for (potentially blocking) I/O operation.
retry_transfer_wrapper() is used when regular files are read.
ff_network_wait_fd_timeout() has been changed the same way, just because they
do merely the same thing and are similar internally. But
ff_network_wait_fd_timeout() is used for network data transmission.
Feature was added to 'file' protocol to limit maximum block size transmitted at
once. I.e. if you set blocksize=4096, and demuxer requests 32768 bytes from
input file, demuxer gets maximum of 4096 bytes. It will get more data on next
data request calls (from retry_transfer_wrapper()). This is done to be able to
check for interrupt more frequently. By default, block size limitation is
disabled and behaviour is the same as before.
Commit 1/5 "Special case for ff_check_interrupt" is not essential, and is
questionable. I formed a patch just to not loose the idea proposed by Michael.

Andrey Utkin (5):
  Special case for ff_check_interrupt
  retry_transfer_wrapper(): check for interrupt before operation
  ff_network_wait_fd_timeout(): check for interrupt before operation
  file: Add 'blocksize' option
  Document file protocol options

 doc/protocols.texi    | 10 ++++++++++
 libavformat/avio.c    | 12 +++++++-----
 libavformat/avio.h    |  5 +++++
 libavformat/file.c    | 12 ++++++++++--
 libavformat/network.c |  4 ++--
 5 files changed, 34 insertions(+), 9 deletions(-)

-- 
1.8.1.5