Bug#870346: glib2.0 FTCBFS: .../.libs/gobject-scan: cannot execute binary file

[Simon McVittie]

On Tue, 01 Aug 2017 at 11:27:15 +0200, Helmut Grohne wrote:
> Reordering makes cross building succeed. Thus I am attaching the
> reordering as a patch.

I'm not sure this is really the right thing: we want the EXTRA_FLAGS to
be overridable individually by the deb and udeb flavours, so the deb
and udeb flags should come last.

> I do question whether having --disable-gtk-doc conditional to cross
> building is a good idea. If I understand the packaging correctly, the
> documentation is only needed for the libglib2.0-doc package, which
> happens to be Arch:all. Is there a good reason for not passing
> --disable-gtk-doc in arch-only native builds?

Yeah, I think this is probably just a matter of someone making it work.
This would also be nice for those of us who routinely do Arch:all
and Arch:any builds separately in order to ensure that the package
won't fail on the buildds - gtk-doc is pretty slow and only running
it once would be a big win (doubly so if you're building for both
amd64 and i386, like I am).

Trying out a potential patch now.

> It would be very nice to have this fixed, since glib2.0 is a key piece
> of the system, with basic packages including dbus and systemd depending
> on it.

dbus doesn't *have* to depend on it, although I was told to stop using
stage1 for the minimal build that doesn't involve GLib. (It isn't
<nocheck> because the installed-tests use GLib, so we lose nearly all
the autopkgtest coverage without it.)

    smcv