Re: libkpathsea transition

[Hilmar Preusse <hille42@web.de>]

On 30.08.05 Frank Küster (frank@debian.org) wrote:
> Hilmar Preusse <hille42@web.de> wrote:

Hi all,

> > Hmm, wasn't it the case, that the old libkpathsea3 implements the
> > new TDS, while libkpathsea4 understands the new one? I.e. a
> > program linked with libkpathsea3 won't work with tetex 3.0? Or am
> > I totally broken?
> 
> Actually the TDS changes are mainly in texmf.cnf - there might be
> some new variables that only the new kpathsea understands.
> 
Hmm, OK.

> I think that a program linked against libkpathsea3 (and using this
> version, of course) would still find most files on a putative unstable
> system with teTeX-3.0. If it doesn't, that would be inconvenient for
> the unstable user
> 
I'd like to have it tested before simply assuming that the search lib
does not affect anything. E.g. the japanese version of dvips would
search map files etc. at the wrong place, which would completely
break the package.
But this is all IMHO. I've no clue how all this internally works.

> However, if we just upload teTeX-3.0 into unstable, libkpathsea3
> would vanish, and all packages that get recompiled after that are
> linked against libkpathsea4.  Then they could not enter testing
> before tetex-bin and libkpathsea4 enters testing.  And this would
> imply a library transition, which was forbidden by the release
> managers (for valid reasons, I think).
> 
Correct. Maybe I'm just to conservative. IMHO we should wait until
the next transition is allowed by the RM. Olaf said the soname was
changed just to make sure nothing breaks, i.e. the API wasn't changed
by intention. So I expect this transition should not take too long.
I prefer this to (temporary) broken packages.

H.
-- 
Traffic signals in New York are just rough guidelines.
		-- David Letterman
  http://hilmarpreusse.forum-rheinland.de/