No subject

> A Dilluns, 9 de mar=E7 de 2009, Vincent Torri va escriure:
> > On Mon, 9 Mar 2009, Albert Astals Cid wrote:
> > > A Dilluns, 9 de mar=E7 de 2009, carlo.bramix va escriure:
> > >> Hello!
> > >>
> > >>> Apparently if you generate a configure file on unix
> > >>> from configure.in with a recent autoconf, it
> > >>> does not produce a configure script that works on
> > >>> MINGW+MSYS.
> > >>
> > >> I just rebuild the configure scripts under Debian 5.0 (stable, not=

> > >> testing or unstable) with colinux and everything works fine when I=

> > >> compile under mingw+msys.
> > >>
> > >> Although all my efforts, I was never able to make autotools workin=
g
> > >> under msys. Instead, autotools under cygwin normally worked fine b=
ut,
> > >> for some reasons that I could not understand very well, I remember=
 that
> > >> I had some troubles with CR-LF somewhere. So I decided to (re)buil=
d the
> > >> scripts under linux: in that manner I'm sure that everything is co=
rrect
> > >> in every part.
> > >>
> > >> I downloaded Poppler 0.10.4.
> > >> I deleted generated files.
> > >> I rebuild the scripts and the makefiles.
> > >> I compiled and installed poppler. Success.
> > >>
> > >> autogen.sh under linux said:
> > >>
> > >> autoconf 2.61
> > >> automake 1.10.1
> > >> libtool 1.5.26
> > >>
> > >> These packages are not that old, perhaps you have really "too new"=

> > >> packages. I believe it would be better to rebuild the scripts with=
 a
> > >> stable distribution, at least we are sure that it will work on
> > >> everything. Anyways, at the moment I believe it is a good idea to =
leave
> > >> the AC_LIBTOOL_WIN32_DLL macro included into configure script, thi=
s will
> > >> ensure the compatibility with the old, with the stable and the
> > >> testing/unstable. Personally, I believe that forcing it is a regre=
ssion
> > >> because lot of software have not it declared in their configure sc=
ripts
> > >> but they could be already compiled as shared without problems, but=
 this
> > >> is just my opinion.
> > >
> > > It's an autotools regression/feature, if you look a bit around the
> > > internet you'll see the mingw people complaining to the autotools p=
eople
> > > (i think the "problem" is really in libtool) for the change.
> > >
> > > BTW i'm now using
> > > autoconf 2.61
> > > automake 1.10.1
> > > libtool 2.2.4
> >
> > if you want to build dll on windows with msys/mingw you either *have =
to*:
> >
> > 1) use AC_PROG_LIBTOOL and call AC_LIBTOOL_WIN32_DLL before and add
> > -no-undefined to libtool flag in the Makefile.am that build the libra=
ry.
> > This way is considered deprecated since 2004-02-20 (in automake 1.9b)=
.
> >
> > 2) or use LT_INIT([win32-dll]) and add -no-undefined to libtool flag =
in
> > the Makefile.am that build the library. LT_INIT was introduced after
> > libtool > 1.5.
> >
> > As poppler development has certainly begun after 2004 (poppler 0.2 wa=
s
> > released in 2005, there is no date for 0.1), I think that you can saf=
ely
> > use the new libtool calls.
> 
> I'd better leave it as it is now if it's working.
> 
> Albert
> 
> >
> > Vincent Torri
> 
> 
> _______________________________________________
> poppler mailing list
> poppler at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/poppler
>