libhunspell SONAME transition

[Rene Engelhard <rene@debian.org>]

Hi,

hunspell 1.2.2 is supposed to be released RSN.

It fixes many bugs - amongst them a crash together with libhyphen
(which OOo uses for hyphenation).

To work around this crash, for OpenOffice.org, since 1:2.4.0~rc1-3, we build
with internal hunspell, but this always was just a temporary measure until he
have a hunspell
 - to which the most important fixes (the crash) can be backported to/is
   integrated in
 AND
 - which doesn't fail the testsuite on anything else than i386 and amd64
  (This is #449369).

Both points are now done with 1.2.2, unfortunately it changed the ABI
and therefore the SONAME. It's completely ABI compatible, though.

I do *not* think we should release OOo linked with internal hunspell
when we can avoid it.

Affected Source Packages in alphabetical order:
 - enchant
 - iceape
 - icedove
 - iceweasel
 - openoffice.org
 - xulrunner

Because 1.2.2 is completely API-compatible they are all
(except openoffice.org, which needs the rules change to use system-hunspell
again) bin-NMUable.

As openoffice.org is not (well, it needs source changes to enable
--with-system-hunspell again) I'll do the switch to system-hunspell for it
for the next source upload.

But openoffice.org will still be coupled to the suitesparse transition as the
OOo package in testing depends on the old one and the new hunspell would
break it unless the new OOo also goes in (and OOo itself is besides
suitesparse also blocked by gcc-defaults/gcc-4.3/gnat-4.2/gnat-4.3/binutils)

Opinions?

Regards,

Rene