[Freeciv-Dev] Re: TODO

[Daniel Sjölie <deepone@xxxxxxxxxx>]

On 2000-09-22 12:45:45, Raimar Falke wrote:
> On Fri, Sep 22, 2000 at 10:56:58AM +0100, Jules Bean wrote:
> > On Fri, Sep 22, 2000 at 11:55:43AM +0200, Erik Sigra wrote:
> > > Falk Hueffner wrote:
> > > > 
> > > > Erik Sigra <freeciv@xxxxxxx> writes:
> > > > 
> > > > > Yes, my implementation requires a lot of recoding (I'm doing it
> > > > > right now). But it can be done in steps. It just takes some
> > > > > typecasting klugdes ("(struct locus({x, y}") in some places during
> > > 
> > > Should have been "(struct locus){x, y}" of course.
> > > 
> > > > > the transition.
> > > > 
> > > > Please don't; this is not portable, it's a gcc extension.
> > > 
> > > Really, I didn't know. Then how it is it done by ANSI?
> > 
> > 'on-the-fly' struct creation is not possible.
> > 
> > Instead use a 'constructor' function
> > 
> > struct locus make_locus(int x,int y) {
> >   struct locus l;
> >   l.x=x; l.y=y;
> >   return l;
> > }
> 
> AFAIK this is wrong. The variable l is created on the stack which is
> destroyed at the return. You have to do return a newly allocated part of
> memory. Using this you will get some problems with bookkeeping since C has
> no GC.

That's true (unless this is some weird special case)
The easiest way to do it probably is to have a variable in the calling
context and have a function like:

struct locus *
set_locus( struct locus *l, int x, int y )
{
    l->x = x;
    l->y = y;
    return l;
}

That way You don't have to worry about allocating memory and freeing it
later...

/Daniel

-- 
Now take a deep breath, smile and don't take life so seriously... :)