Bug#618805: britney2: wrong uninstallable count in some cases? (libebml+libmatroska transition)

["Adam D. Barratt" <adam@adam-barratt.org.uk>]

On Fri, 2011-03-18 at 16:53 +0100, Julien Cristau wrote:
> > b2:
> > Trying hint from jcristau: libmatroska/1.1.0-2 libebml/1.2.0-2
> > leading: libmatroska,libebml
> > start: 113+0: i-7:a-2:a-2:i-2:k-45:k-45:m-3:m-2:p-2:s-2:s-1
> > orig: 113+0: i-7:a-2:a-2:i-2:k-45:k-45:m-3:m-2:p-2:s-2:s-1
> > recur: [] libmatroska,libebml 403/0
> 
> b2 believes grass, grass-dev, libgdal1-1.6.0-grass, libqgis-dev,
> qgis-plugin-grass are already uninstallable by that point.
[...]
> > trying: grass
> > accepted: grass
> >    ori: 113+0: i-7:a-2:a-2:i-2:k-45:k-45:m-3:m-2:p-2:s-2:s-1
> >    pre: 382+0: i-35:a-27:a-26:i-26:k-69:k-69:m-27:m-26:p-26:s-26:s-25
> >    now: 382+0: i-35:a-27:a-26:i-26:k-69:k-69:m-27:m-26:p-26:s-26:s-25
> >    all: libmatroska libebml grass
> > 
> > i.e. it seems to have a different (broken?) uninstallable count at the
> > start of the hint already...

This may (but equally may not) be related to an issue I fixed earlier
today and pushed as 347022bb01182704cc1976ffb8fd2de916933620.
Unfortunately I haven't managed to reproduce this particular issue in my
tests so far, which is making it difficult to confirm whether one of the
recent fixes may have resolved it.

Regards,

Adam