[MPlayer-users] 2 Bugs [?]

Hans Meine meine at kogs1.informatik.uni-hamburg.de
Wed Mar 12 15:08:12 CET 2003

On Wednesday 12 March 2003 10:38, Vladimir Mosgalin wrote:
> Btw, do developers know that seeking is somewhat broken? It works
> different for different type of files, sometimes incorrect, and seeking
> intervals smaller that 10 behave strangely... "seek +2" seems to be
> smallest seek in most files, and "seek -3" seeks the same amount
> backwards, for example... I can report more complete examples, if
> someone is interested.

Most of those effects can be explained with the need for a keyframe AFAIK. 
That means that as long as the player does not search for the last keyframe 
_before_ the point you want to seek to and calculates all the frames in 
between (and mplayer does not do so), it can not give you (a non-distorted 
version of) the exact frame you requested.

OTOH, I heard that mplayer at least had a bug at some point which evaluated 
the keypress as well as the keyrelease (when pressing [->] for example). 
However, there must be different keyboard drivers (depending on the 
environment - X / linux console / ...) and I don't know if the bug was/is in 
all those. I did not discover that problem yet.

To solve the problem that seeking forward/backward results in different steps, 
maybe mplayer should not seek to the first keyframe after the seekpoint 
(which i _guess_ it does) but to the _nearest_ keyframe (while making sure 
that it still seeks into the right direction*).


* Imagine the current frame is 2secs after a keyframe, the next keyframe is 28 
seconds in the future and you want to seek +10: The nearest keyframe to the 
seekpoint would be that 2seconds in the past, naively seen.

More information about the MPlayer-users mailing list