Re: [gnome-db] libgda glib 2.8 issues.

[Bas Driessen <bas driessen xobas com>]

On Wed, 2006-01-25 at 09:11 +0100, Vivien Malerba wrote:
> 2006/1/25, Bas Driessen <bas driessen xobas com>:
> >  On Tue, 2006-01-24 at 09:18 +0100, Vivien Malerba wrote:
> >  2006/1/24, Bas Driessen <bas driessen xobas com>:
> > > Hello All,
> > >
> > > Just downloaded the latest CVS libgda sources and while attempting to
> > > compile I noticed 2 issues:
> > >
> > > -1 undefined reference to `g_access'
> > > -2 undefined reference to `g_mkdir_with_parents'
> > >
> > > Apparently 'g_access' and 'g_mkdir_with_parents' are only available in
> > glib
> > > 2.8, while the system I am working on (FC4) has glib 2.6.6 installed. If
> > we
> > > really want to use these 2 new calls, configure.in has to be modified to
> > > check for glib version >= 2.8 rather than 2.0.0 as it is doing now.
> > >
> > > On the other hand, these new calls are only made in 3 files
> > > (gda-diagnose.c, gda-test-models.c and gda-test-sql.c) in the /testing
> > > directory. do we really want to up the glib requirement just for these 3
> > > testing files? An easy workaround for glib 2.6 is to replace g_access with
> > > access and g_mkdir_with_parents with g_mkdir. Not sure if that is
> > compatible
> > > with 2.8, but we could use #ifdef settings based on the glib version in
> > case
> > > it is not.
> > >
> > > Personally I would like to keep libgda compatible with glib 2.6.
> > >
> >
> > I agree with you. Can you provide a patch?
> >
> >
> >  Changes applied to cvs HEAD as discussed. There are many glib 2.6 specific
> > function calls in the functions in the /testing directory. Therefore I have
> > set the glib requirement in configure.in to a minimum of 2.6.0. This should
> > be a reasonable requirement for a modern library as libgda. The glib 2.6
> > library has been around for more than a year anyway.
> >
> >  If anyone has a problem with this, please respond.
>
> I would have prefered that you use #ifdef to keep the
> g_mkdir_with_parent() when possible as it does more than the
> g_mkdir(): it creates the missing directories along the way.
>
> Anyway it's not a big deal, so don't bother correcting it.
>
> Thanks for the patch,
Correct, that is why I suggested the #ifdef situation. After reading up on the web on both g_mkdir and g_mkdir_parent they both appear to translate to the "mkdir -p" situation. I could not see any difference other than a better (?) name. I will double check in the morning and download and look in the actual code and find the differences if any.

Probably not a big deal, like you said, but if we make a change, let's do it the proper way :)

Bas.