[Freeciv-Dev] Re: (PR#3928) Convert client to use PF

["Gregory Berkolaiko" <Gregory.Berkolaiko@xxxxxxxxxxxx>]

On Tue, 8 Apr 2003, Raimar Falke wrote:

> On Mon, Apr 07, 2003 at 08:43:28AM -0700, Gregory Berkolaiko wrote:
> >  /********************************************************************** 
> > -...
> > +  PF callback to get the path with the minimal number of steps (out of 
> > +  all shortest paths).
> >  ***********************************************************************/
> > +/* Glip: can we have a better name please? */
> > +static int my_get_EC(int x, int y, enum known_type known,
> > +                struct pf_parameter *param)
> 
> You can create one. I always use "my_". If I would have anonymous
> function I make give them no name at all.

Well, even client_goto_EC_callback is nicer than "mine".

But I would do EC_always_one...

> > +/* Glip: dir is enum, in many other places too. */
> 
> The use of direction in the other cases is more or less buggy. You
> loose for example if dir>=4 && reverse(dir)>=4. There is now code to
> test for this. So some kind of "enum direction8_folded_to_0-3" is
> needed. With the corresponding functions for the conversion.

Well, maybe not then ;)

> > I didn't change any of your code, only comments, but I added one fix to 
> > prevent client crashing (also marked by "Glip:") and threw in a couple of 
> > fixes of the path_finding, for good measure.  These were the bugs I 
> > discovered while doing rampage, we can commit them separately or as part 
> > of your patch.  The bugs are:
> > -- construct_path tried to step back beyond the initial position of the 
> >    path
> > -- zoc-handling prevented paths ending on a military target.
> 
> It was to be expected that we find errors in PF. These should be
> commited separately.

Can you check and commit them please?
Nobody but you can check it anyway.

> > On Sat, 5 Apr 2003, Raimar Falke wrote:
> > 
> > > > > Open issues are:
> > > > >  - the network interface doesn't support dangerous tiles
> > > > >  - the path execution at the server doesn't support dangerous tiles
> > > > >  - the network interface is quite weird (for example it divides the
> > > > >  path into 20-positions chunks while a chunk could holds 1500/2=750
> > > > >  positions)
> > 
> > So this means that if I attempt doing goto with a trireme I will get a 
> > core?  I think it should be fixed before this is committed.
> 
> Untested: the server will just go the steps and will not wait at the
> steps at which the unit should wait. So you don't get a core but a
> sunken trireme.

Not as bad as the core.  Any reason why you want to leave it for later?

> > But have you fasted?
> 
> Can I do more sport in exchange? This is also a burden.

Okay.

G.