Bug#401969: please build using hunspell
rene at debian.org
Fri Dec 8 02:49:12 CET 2006
Mike Hommey wrote:
> On Thu, Dec 07, 2006 at 11:26:34AM +0100, Rene Engelhard <rene at debian.org> wrote:
> > [ also Cc'ing iceweasels bug ]
> > Am Donnerstag, 7. Dezember 2006 08:12 schrieben Sie:
> > > block 401969 by 324637
> > > thanks
> > [...]
> > > > MySpell is obsolete.
> > > > Please build against Hunspell, which is an improved version of MySpell
> > > > retaining full backwards compatibility. That also would make the usage
> > > > of hunspell-de-* in iceweasel possible.
> > > (...)
> > >
> > > I won't do that until hunspell comes with a shared library.
> > And I won't add a shared library myself and invent a SONAME. (Like Fedora does).
> Why is that ? Not for binary compatibility, obviously (which is
> generally the reason to not *change* a soname)
Because of this. Binaries which link to a SONAME which only exist in
distributions inventing a SONAME don't work elsewhere. Binaries built
on Fedora (and Debian if i'd apply it if shared libs' future in hunspell
is uncertain) are not able to use it somewhere else where only the
"normal" upstream contents are packaged. Of course, binary-only things
suck and you most times have the source to rebuild it, but...
Anyway, the Fedora patch was now upstreamed to sourceforges tracker. If
upstream adopts that one I will build a libhunspell-1.1-0 package.
(tried that already, works)
> > (BTW, I'll probably remove MySpell after etch so then this is RC;
> > I consider building against system-hunspell (and telling that) still better
> > than building with an internal copy of MySpell)
> > Especially since Mozilla now doesn't support hunspell dicts but hunspell-* and
> > myspell-* have the same file. I guess I need to make hunspell-* conflict with any
> > Mozilla out there *sigh*)
> > There's no big difference in using hunspell and myspell, except that hunspell dictionaries
> > then will also work. And you show that hunspell is used so the security team knows
> > that mozilla needs to be rebuilt (which probably won't happen, no one ever found a security bug
> > in either hunspell or myspell). And in any case, there's already enchant and openoffice.org building
> > with static hunspell (openoffice.org does build far longer than ice*)
> It's not because it's done elsewhere that it's not a bad thing.
What is your argument exactly for not linking to link against that static
lib? Only because it's static? That argument IMHO doesn't really
Many things link against static libs - simply becaus ethere's no shared
lib (yet) upstream.
My points are:
- you are blocking usage of superior dicts (hunspell-*) elsewhere
(OOo, enchant/abiword) by not supporting hunspell in Ice* for those
that use OOo or enchant/abiword and some Ice animal.
(And I want to get myspell removed for etch+1...)
- even with hunspell being static you point the security
team that you link with it (not that a security update for hunspell
should ever be necessary. so this isn't a problem either)
- it won't change behaviour of the spellchecker (except improve it's
ability to check words)
.''`. Ren? Engelhard -- Debian GNU/Linux Developer
: :' : http://www.debian.org | http://people.debian.org/~rene/
`. `' rene at debian.org | GnuPG-Key ID: 248AEB73
`- Fingerprint: 41FA F208 28D4 7CA5 19BB 7AD9 F859 90B0 248A EB73
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
Url : http://lists.alioth.debian.org/pipermail/pkg-mozilla-maintainers/attachments/20061208/08781040/attachment-0001.pgp
More information about the pkg-mozilla-maintainers