[Lumiera] Gtk3 Is Underway

Michael Fisher mfisher31 at gmail.com
Wed Aug 15 18:21:45 CEST 2012


Good news about gcc 47.  I just did what the error told me to hehe.  I went
ahead merged those changes into gtk3 and gdlmm3, no need to specify gcc-4.4
for those two branches.

On Wed, Aug 15, 2012 at 10:38 AM, Ichthyostega <prg at ichthyostega.de> wrote:

> Am 15.08.2012 10:32, schrieb Michael Fisher:
> > I was able to make decent progress with migrating to a gtkmm-3.0 gui.
> > Under wheezy, the lumiera gui will build against the provided gtk3 (and
> > gdl3) libraries.  I have created three new branches for the endeavor.
> >
> > mfisher31/gtk3 - Bulk of gtk3 specific transformations here
> > mfisher31/gdlmm3 - Includes above + gdlmm-port (works)
> > mfisher31/gcc-4.7 - Changes specific to allow compiling with the wheezy
> > stock compiler (non-gui).
>
> Wow....
> thanks! I'll have a more close look next days.
>
>
> > I'm not exactly sure what I have going on in the gcc-4.7 branch is what
> > 'should' be happening so I made it separate.
>
> I've looked at some of the changes yesterday. It seems that gcc-47 is more
> picky when it comes to parsing template definitions. Basically the language
> requires that each function name used in a template needs to be known
> (declared
> somewhere) right away. BUT there is the exception regarding the "parameter
> dependent names": the compiler can't conclude anything regarding symbols,
> where the exact meaning *could* depend on the concrete parameters used
> to instantiate the template. In these cases, the lanugage requires the
> compiler to postpone the search for a declaration to the point when
> the template is actually instantiated with concrete type parameters.
>
> Now, in the cases I saw, the 4.7 compiler complained about a name not
> declared; you did the right thing to prefix the names with this->
> because in these cases the type of the "this" pointer indeed depends
> on the template parameters, and indeed the intention was to mix in
> these member functions via a base class provided on template instantiation.
>
> So, your solution to make that names parameter dependent and thus postpone
> the compiler check to actual instantiation time was the canonical solution.
>
> I rather don't understand why the older compiler didn't complain; probably
> it's just a sloppiness, and thus it's good to fix those situations.
>
> Cheers,
> Hermann
>
> _______________________________________________
> Lumiera mailing list
> Lumiera at lists.lumiera.org
> http://lists.lumiera.org/cgi-bin/mailman/listinfo/lumiera
> http://lumiera.org/donations.html
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lumiera.org/pipermail/lumiera/attachments/20120815/9f6fe890/attachment.html>


More information about the Lumiera mailing list