[MPlayer-dev-eng] [PATCH] Use "-nocache" option in TOOLS/midentify.sh

[Reimar Döffinger]

On Sat, Feb 11, 2012 at 02:00:01PM -0300, Rodrigo Campos wrote:
> On Sat, Feb 11, 2012 at 05:44:17PM +0100, Reimar Döffinger wrote:
> > On Sat, Feb 11, 2012 at 01:16:59PM -0300, Rodrigo Campos wrote:
> > > On Sat, Feb 11, 2012 at 01:08:56PM -0300, Rodrigo Campos wrote:
> > > > On Sat, Feb 11, 2012 at 11:46:16AM +0100, Reimar Döffinger wrote:
> > > > > On Wed, Feb 08, 2012 at 11:07:42AM -0300, Rodrigo Campos wrote:
> > > > > > Perhaps more gain for this case and cases where the demuxer seeks back and forth
> > > > > > could be achived with the other option you suggest (playing with "cache-min"). I
> > > > > > will probably play with them, but I don't know which demuxers do that. May I ask
> > > > > > you how can I know so I can try different "caches" options ? :)
> > > > > 
> > > > > I don't know it out of my head, but "-noconfig all -cache-min 0" seems
> > > > > to me to give reasonable results (actually the same as -nocache, but
> > > > > should potentially avoid some issues).
> > > > 
> > > > As far as I tested, it gives the same results for me too :)
> > > > 
> > > > > What do you think?
> > > > 
> > > > I think it seems like the right approach, thanks! :)
> > > > 
> > > > I mean (if I got it right), if there is no need to seek back and forth, and we
> > > > are running with "-cache-min 0" and "-frames 0", it will get the info and exit
> > > > ASAP.  Just like if "-nocache" was used. But if it needs to seek back and forth,
> > > > it will use the cache and avoid downloading two times the same thing. And it
> > > > will not wait to fill the cache, since cache-min is set to 0 and as soon as it
> > > > gets all the info it will exit. Or did I misunderstand something ?
> > > 
> > > Also, shouldn't we use "-cache-seek-min 0" too, to avoid filling the cache when
> > > a seek is done ? Or this option only affect seeks made by the user ?
> > 
> > It affects all seeks, but this is about whether MPlayer will do a real
> > seek (in case of http: open a new connection) vs. just reading on until
> > it has the right data.
> > That one specifically should not be changed IMO, to avoid exactly the
> > problems I suspect with -nocache.
> 
> Yeah, right. So, I guess a patch with "-noconfig all -cache-min 0" should be
> accepted ? Should you send it as it was your idea ? :)

I committed the change.