Re: Depending on C99 (Re: GtkBindingSignal changes)

[Matthias Clasen <mclasen redhat com>]

On Thu, 2006-01-05 at 12:28 +0100, Michael Natterer wrote:
> On Thu, 2006-01-05 at 12:01 +0100, Tim Janik wrote:
> > On Wed, 4 Jan 2006, Xavier Bestel wrote:
> > 
> > > On Wed, 2006-01-04 at 15:26, Matthias Clasen wrote:
> > >
> > >> Thats a gcc extension/C99 addition though. We use [1] in other places
> > >> where flexible arrays are used.
> > >
> > > Oh, I imagine there must be good reasons (read: already debated to
> > > death) for not taking advantage of C99 improvements. Sorry.
> > 
> > no, it actually hasn't as far as i know. and i think it makes sense to
> > at least start discussing this possibility.
> > 
> > 
> > could people please speak up if they think depending on C99 would
> > be a bad idea for glib & gtk+ (e.g. with the next major release) and
> > why this would be a problem for them?
> > 
> > in the end, the w32 compilers are C99 compliant i'd assume and on
> > older unixes when there're no modern compilers available, there's
> > always a C99 compliant gcc that can be used.
> 
> I'm all for this step. GIMP struggles over stone age compilers
> on old unixes all the time, and asking people to use a complier
> that's a bit younger than 20 years isn't too much to ask for.
> 
> So please, let's get rid of some legacy and enable everybody
> to use a bit more modern C.
> 

We still get regular bug reports from people using gcc 2.95 whenever
we acidentally put a variable declaration in the middle of a block. 

That doesn't mean that we can't declare gcc 2.95 unsupported at some
point, but we should probably discuss which C99 features we actually
want to use in the GTK+ stack. I'm personally not looking forward to
have variable declaration sprinkled all over the code, with the
notable exception of for (int i = 0; ...)

Matthias