[Freeciv-Dev] Re: core file on civserver, http://civserver.freeciv.org/games/39967

[Gregory Berkolaiko <Gregory.Berkolaiko@xxxxxxxxxxxx>]

On Sun, 1 Sep 2002, Per I. Mathisen wrote:

> On Sat, 31 Aug 2002 freeciv@xxxxxxxxxxxxxxxxxxxxx wrote:
> > #0unit_type (punit=0x0) at unittype.c:94
> > 94  return get_unit_type(punit->type);
> > (gdb) #0unit_type (punit=0x0) at unittype.c:94
> > #10x0805c659 in search_best_target (pplayer=0x8130ae0, pcity=0x834ac38,
> >   punit=0x82d4500) at autoattack.c:98
> 
> Let's see. We found a tile which
>       - has units
>       - is_enemy_unit_tile(map_get_tile(x, y), pplayer) == TRUE
>       - but where get_defender(punit, x, y) == NULL
> 
> We don't check for NULL return value from get_defender, so we core.
> Obviously we should check this, problem solved and patch attached. But why
> does it happen?

Another question is why the autoattack code was executing at all?  Was it 
a human user using auto-attack (unlikely) or some recent patches made AI 
use it?

G.


> 
> Only case I can think of is a transport of one of your allies carrying a
> unit whose owner you are at war with getting within range of the
> autoattack.
> 
> That should have triggered the following LOG_ERROR:
> 
>     freelog(LOG_ERROR, "Get_def bugged at (%d,%d). The most likely course"
>             " is a unit on an ocean square without a transport. The owner"
>             " of the unit is %s", x, y, unit_owner(debug_unit)->name);
> 
> Is there any way to get LOG_ERROR and LOG_FATAL output from a game put
> into the core bug report?
> 
> Yours
> Per
> 
> "I don't see why people are so upset about cloning sheep. American
> television networks have been doing that to their audiences for years."
> -- Jello Biafra
>