Re: Gtk2::GLExt build problems

[Mark Glines <mark glines org>]

muppet wrote:
> On Jun 2, 2006, at 5:07 PM, Mark Glines wrote:
>
> > I'm having some difficulty using Gtk2::GLExt-0.90.
>
> Be warned that this module has not been touched since Oct '04.  I don't
> recall with what version of gtkglext it was developed.

Thanks for the tip.  Is it deprecated in favor of something else, or
does it just need a little love? :)


> > not a valid file descriptor at
> > /usr/lib/perl5/vendor_perl/5.8.8/i686-linux-thread-multi/Glib/ParseXSDoc.pm
> >
> > line 105
>
> That line is
>
>         store_fd $parser->{xspods}, select;
>
> which attempts to write the data collected for pod generation with
> Storable.  The immediate question is "why would that give you an invalid
> file descriptor?"
>
> What version of Storable do you have?

2.15.  I get the same symptoms on two Gentoo Linux machines, probably
with the same versions of everything installed.


> As a workaround, if you set FORCE_DATA_DUMPER=1 in the environment when
> running Makefile.PL, Glib::MakeHelper::do_pod_files() should fall back
> to using Data::Dumper instead of Storable.  We switched to Storable
> earlier this year to reduce the outrageous memory consumption of
> documentation generation on Gtk2; however, Gtk2::GLExt is pretty small,
> so there should be no problems.

That worked.  My home machine fails "make test" due to a mismatch
between nvidia kernel module version and header file version... not
Gtk2::GLExt's fault.  Guess it got upgraded since I rebooted last.


> > When installed, your examples/example.pl.sdl script runs, but it emits a
> > couple of warnings.
> >
> > Prototype mismatch: sub main::GL_NICEST: none vs () at
> > /usr/lib/perl5/5.8.8/constant.pm line 103.
> > Prototype mismatch: sub main::GL_COMPILE: none vs () at
> > /usr/lib/perl5/5.8.8/constant.pm line 103.
> >
> > Turns out SDL::OpenGL *does* export those two constants, at least the
> > version on this box (sdl-perl-2.1.3) does.
>
> That should be pretty harmless, although it's annoying.
>
> Does anybody know offhand a decent way to define a constant if it's not
> already defined?


It doesn't seem to be complaining about redefined constants; its
actually talking about a prototype mismatch.  So the warnings might go
away if we explicitly set the right prototype.

I don't know how to specify a prototype with "use constant", or indeed
if its possible at all.  But replacing the constants with the following
lines doesn't emit any warnings for me:

sub GL_NICEST { return 0x1102};
sub GL_COMPILE { return 0x1300};

And it makes no difference whether I declare them with empty parens or not.

Mark