[Freeciv-Dev] Re: {gen,spec,sort}list stuff

[Justin Moore <justin@xxxxxxxxxxx>]

> > I've started to notice some inconsistencies in games using the same random
> > seeds (map and game) between games that use the old genlist and ones that
> > use my genlist.  Yet I can't find any problems in my implementation of
> > genlist (yes, I realize that by saying this I'm opening myself up to
> > someone finding incredibly obvious bugs ... which is almost the point).
> > I'd appreciate any extra eyes looking at the code.
>
> Is it possible to have both lists in parallel and compare their results.

   Yes, I'll get back to everyone on this.

   In some more recent profiles I've been able to cut down the number of
old iterations by about 66% for a speed savings of around 5%.

> >    There are currently two kinds of list iteration macros: one (STABLE_*)
> > for lists where the number of elements doesn't change and another (REAL_*)
> > for those that do (REAL_ because that's what most use ATM).  The REAL_*
> > macros use the iterators, which are painfully slow but are very tolerant
> > of one or more elements being removed from underneath them.
>
> I'm not happy with the name REAL_.

   SAFE_?

> Please people test it. Looking at it I think that you should be more
> restrictive with your arguments:
>  - pgenlist->capacity should be set for every list and yet you test
>  for it in genlist_resize

   I had, at one point, thought about reclaiming memory if the current
list size went below 1/4 of the allocated size.  The check was meant to be
a safeguard against some weird math problem that may have happened
somewhere along the line, but it's really not necessary.

>  - is this mess with the current genlist_sort and NULL needed? It is
>  used.

   Could you be more specific?  I'm just replicating the external behavior
of genlist_sort when given a NULL compar pointer: nothing is done to the
list.

>  - which code unlinks none items which aren't in the list (assert,
>  assert,...)

   ??? Ya lost me.

-jdm

Department of Computer Science, Duke University, Durham, NC 27708-0129
Email:  justin@xxxxxxxxxxx