[Bug 770935] RTP/RTSP/TCP transport RTSP TIMEOUT should start at once when tcpi_ca_state is TCP_CA_Loss

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=770935

Dag Gullberg <opaque.se at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |opaque.se at gmail.com

--- Comment #1 from Dag Gullberg <opaque.se at gmail.com> ---
Created attachment 334899
  --> https://bugzilla.gnome.org/attachment.cgi?id=334899&action=edit
RTP/RTSP/TCP transport RTSP TIMEOUT should start at once when tcpi_ca_state is
TCP_CA_Loss

This patch depends on patch in
https://bugzilla.gnome.org/show_bug.cgi?id=770934

Understanding that this likely is a Linux specific fix, I still put it up for
review. Should perhaps be set in a platform independent API. Suggestions most
welcome.

The idea is to make sure that we capture the first possible moment when there
is loss to trigger the RTSP TIMEOUT.

Having the TCP stack indicating TCP_CA_Loss would be that moment.

With this fix together with the previous one in 770934, the time gap is reduced
in one case from 14s to 0s, making the TCP tunneling TIMEOUT-wise behave like
the UDP case.

A desirable feature.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.