[MPlayer-cvslog] r33352 - in trunk/gui: cfg.c cfg.h interface.c

[Reimar Döffinger]

On Sat, May 07, 2011 at 01:47:31PM +0200, Ingo Brückl wrote:
> Reimar Döffinger wrote on Sat, 7 May 2011 09:23:47 +0200:
> 
> > On Fri, May 06, 2011 at 09:31:26AM +0200, Ingo Brückl wrote:
> >> I wrote on Wed, 04 May 2011 02:13:29 +0200:
> >>
> >> > I can offer an alternative solution that avoids all this clutter with a few
> >> > core code changes inside *existing* CONFIG_GUIs.
> >>
> >> > The advantage: Clear and simple. The GUI can react now on *all* MPlayer's
> >> > exits at each point in time. No issues in any corner case. guiDone() is
> >> > sufficient, the obscure guiExit() stuff unneeded.
> >>
> >> Reimar, I need your decision whether I shall go on with the cluttering r33352
> >> approach or with the suggested alternative which I regard as the far cleaner
> >> and preferrable solution.
> 
> > I can't say anything since the first patch isn't complete.
> 
> Well, I wanted to avoid a possible waste of time and just give the idea of
> the patch without elaborating the gui part to see whether you would accept
> those changes in principle or be strictly against them.

Well, it seemed suboptimal to me and I couldn't suggest improvements
without all of it.

> > E.g. why does guiDone even need to know how far initialization
> > got, it should be able to figure it out itself.
> 
> Of course can the GUI figure out itself, but that is not the point (and not
> the problem). The point is that MPlayer won't call guiDone() until it has
> called guiInit(), which is wrong.
> 
> So maybe, we should go even easier, although the name of the symbolic
> constant may no longer be that suitable now. Attached is a complete patch
> (minus indenting) which makes MPlayer call guiDone() on every of its possible
> exits if it has been calling a GUI function before.

Yes, that was about what I had in mind, looks really good to me.