Re: [Gtk-osx-users] Failure while building gtk+
- From: John Ralls <jralls ceridwen fremont ca us>
- To: GTK+-2 OSX Users <gtk-osx-users lists sourceforge net>
- Subject: Re: [Gtk-osx-users] Failure while building gtk+
- Date: Wed, 15 Sep 2010 20:03:27 -0700
On Sep 15, 2010, at 6:44 PM, Richard Procter wrote:
>
> Looks like this has bitten others, too:
>
> http://groups.google.com/group/funtoo-dev/browse_thread/thread/
> efc00bdaa04f908d
>
> The gentoo guys rolled back gobject-introspection to 0.9.0 and stuck
> with pygobject-2.20. As pyobject-2.21 is an unstable development
> release, I've tried to do the same. However, pygobject-2.20 still
> fails to build, with a different error:
>
> ld: duplicate symbol __PyGObject_API in .libs/repo_la-bank-info.o
> and .libs/repo_la-bank-argument.o
>
> This appears due to the code in 'girepository' wrongly including one
> of pygobject's own header files (!), which defines "_PyGObject_API",
> presumably the offending symbol that has had an extra '_' prepended
> at some point: '__PyGObject_API' does not occur in the pygobject
> source tree.
>
> This is where I am tempted to get a little cranky with people
> responsible for releasing PyGObject but it's hard to believe they'd
> release something which did not compile fine for them.
>
> (After down-grading gobject-introspection, I built via
> $ jhbuild build gobject-introspection
> $ jhbuild buildone pygbject
> )
>
> Note that pygtk-2.16 depends on pygobject >= 2.12.1 according to its
> README file.
>
Don't get cranky with them unless you expect them to test on OSX. The duplicate symbols restriction is in Mach-O and not in ELF. Also, the linker adds an underscore to the front of symbols, so foo in source turns into _foo in the linker, and _foo turns into __foo.
Are the pygobject folks following the even = stable, odd = development convention? There's only one release of pygobject-2.20 in ftp.gnome.org, then they immediately started releasing 2.21.x. Makes me think they're not....
A big part of the problem seems to be that gobject-introspection isn't really stable. Perhaps it's too soon for modulesets-stable to use it. From a Gtk-OSX standpoint, it's useful only for PyGtk... so what do you guys think? Should I pull it and go back to --enable-introspection=no for a while in stable?
Regards,
John Ralls
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]