Bug#403639: Fix committed in svn repo

[Loïc Minier]

On Thu, Jan 11, 2007, Frans Pop wrote:
> I have tested the udeb from that location by rebuilding the installer with 
> as only change that udeb, and I can confirm that the memory leaks are 
> (almost?) completely gone. I have done a full installation and not 
> noticed any regressions or problems.

 Ah, I'm relieved.  I was fearing the pathes we might have to follow
 would the patch not improve the situation. :)

> Loïc: How invasive is this patch? How would you judge the risk of 
> regressions? Do you feel comfortable defending it for the RMs?

 I was never enchanted by the nature of the directfb backport to gtk
 2.8 and this is based on an upstream fix, so I don't feel too dirty
 in appliying it if it plugs an important memory leak.
   The cleanup code that it fixes seems to have been incomplete in our
 directfb backport, we didn't keep the backport up-to-date (which would
 have been too much effort anyway), so it was to be expected that such
 issues pop up.  This code is also very tricky, which means I can't
 really criticize it technically as I don't understand it fully.
   Finally, since the patch involves directfb specific code and since
 d-i is the main consumer of this code, I think this is ok to go.  In
 fact, I don't have anything better to offer. :)

 I've uploaded these debs to unstable.

-- 
Loïc Minier <lool at dooz.org>