[gccsdk] Autobuilder libglib2.0-0

Ron beeb at woosh.co.nz
Thu Feb 6 02:16:19 PST 2014


In message <37d474d553.Jo at hobbes.bass-software.com>
          John Tytgat <John.Tytgat at aaug.net> wrote:

> In message <44a0fbd453.beeb at ron1954.woosh.co.nz>
>           Ron <beeb at woosh.co.nz> wrote:
> 
> > In message <4c62ebd453.Jo at hobbes.bass-software.com>
> >           John Tytgat <John.Tytgat at aaug.net> wrote:
> > 
> > > In message <18f0e7d453.beeb at ron1954.woosh.co.nz>
> > >           Ron <beeb at woosh.co.nz> wrote:
> > > 
> > > > I had a problem where libffi was not linking when trying to build
> > > > libglib2.0-0
> > > > After examining the make file I found plain native gcc was being
> > > > used rather than arm-unknown-riscos-gcc.
> > > > In the libglib2.0-0 setvars file I remarked out these lines.
> > > > 
> > > > # build for local system to build local tools
> > > > #CC=$HOST_CC CXX=$HOST_CXX ./configure --prefix=$GCCSDK_INSTALL_CROSS
> > > > #make install
> > > > #make clean
> > > > 
> > > > And then libglib2.0-0 built successfully.
> > > > Can someone explain the intention of the above or perhaps
> > > > correct my misunderstanding.
> > > 
> > > In order to build libglib for RISC OS you need to have the glib-genmarshal
> > > executable for your build machine.  That's why this gets first built
> > > using your native gcc compiler and this gets then installed in
> > > $GCCSDK_INSTALL_CROSS (where the RISC OS cross-compiler is installed).
> > 
> > By 'this' I assume you mean a whole native libglib2.0-0, that 
> > appears to be what is happening anyway.
> 
> Yes.
> 
> I had a go myself to build libglib2.0-0 from scratch.  It turns out that
> the native glib-genmarshal is not (no longer ?) needed for building libglib
> itself.  So I removed the part of building native libglib, basically
> your suggestion.
> 
> I'm not sure whether native glib-genmarshal isn't needed by other libraries
> or libglib using programs, but as it is already suggested by AutoBuilder's
> 'build' file, you better have it installed natively so I don't see the
> point of building it ourselves.  I hope this doesn't have any fallout.
> 
> Hope this helps,
> John.

Well, I compiled gtk-2.0 and gtk+-2.0 with just the cross-compiled 
libglib2.0-0 in place, no complaints. But no testing to see if all
is working properly yet.
Sorry I dont know enough about glib-genmarshal yet.
The cross-compiler would complain if it tried to run the RISC OS
elf variant to do a job though I think.
I /did/ have libglib2.0-0 installed already on my linux, is it possible
the autobuilder used that glib-genmarshal? 
 
I think now that my linux libffi was not installed properly,
(sudo) apt-get purge and install of libffi6 was the likely cure
and not my addition to setvars of
(shell /home/ron/.../pkgconfig..
which as noted by last-failure did nothing.
(pkgconfig is a directory and for that sort of thing the
ro-pkg-config would have to be used.
I was barking up the wrong tree though, as reinstalling libffi6
and libffi-dev (native linux) seems to be what fixed the native
build error with libffi.

BTW gtk+-2.0 built happily with the same patches as gtk-2.0 but 
I haven't any proof that either are useable.
Does anyone know of a small known working port that should
compile and work with them?

Thanks, Ron M 




More information about the gcc mailing list