Re: [PATCH] Fix base-installer FTBFS

[Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de>]

Joey Hess wrote:
> Thiemo Seufer wrote:
> > Joey Hess wrote:
> > > Thiemo Seufer wrote:
> > > > di_log isn't defined (anymore?), and base-installer depends already on
> > > > libdebian-installer4-dev (>= 0.17) anyway.
> > > 
> > > di_log is not present in the unreleased libdebian-installer 0.17.cvs.
> > 
> > I doubt that, since I use current CVS sources and it built on mips with
> > this change. di_log is not #define'ed but available.
> 
> I'm only going by the changelog entry which says it was removed in
> 0.17.cvs.

Then this was fixed already. My patch is still needed for current CVS.

> > It was the only offending instance of di_log I encountered.
> > 
> > > This is distinct from version 0.17, which in in the archive and which
> > > base-installer depends on. As far as I know, that version has di_log.
> > > 
> > > I asked Bastian to revert the changes he made to CVS head for
> > > libdebian-installer 0.17, put them in a branch if necessary, and not
> > > upload it. We can't afford incompatable changes to core d-i components
> > > right now; that is the purpose of this freeze. Perhaps you were confused
> > > by CVS HEAD having an unreleased, and currently unreleaseable version of
> > > libdebian-installer.
> > 
> > Is there a freeze branch of the CVS?
> 
> That's what head is supposed to be..

This can never work. It would effectively mean to stop further development
of d-i over the lifetime of sarge, since there has to be a "critical fixes
only" source tree for the stable release.

I strongly recommend to create a "sarge" branch. This leaves HEAD unimpeded
and allows stricter control over the release branch. Otherwise you'll have
to scare off your co-developers all the time from doing what they like
most: Implementing shiny new features.


Thiemo