[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Follow-up Comment #31, bug #22567 (project freeciv):

> I raised patch #5231 for handling that.

Forgot to mention that this replace "assess_danger_unit.patch".


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Update of bug #22567 (project freeciv):

  Depends on: => patch #5201


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Update of bug #22567 (project freeciv):

  Depends on: => patch #5231


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Update of bug #22567 (project freeciv):

 Assigned to:None => pepeto 

___

Follow-up Comment #30:

I have found another cause of this bug: See comment 6
 for patch #5201.

> It makes more sense to me that the counter changes from 0 to 1
> when the unit actually needs one extra turn to arrive, not when
> the unit is going to run out of move points.

I raised patch #5231 for handling that.


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[David Fernandez]

Follow-up Comment #29, bug #22567 (project freeciv):

I have opened a new ticket bug #22621 for the issue specific to civ2civ3
ruleset.

>But if the unit can exactly reach the tile spending all its moves left, then
the turn counter is increased by 1.
I do not know about the AI, but as human player, I never liked this way of
showing the estimated turns when moving a unit with goto interface.
It makes more sense to me that the counter changes from 0 to 1 when the unit
actually needs one extra turn to arrive, not when the unit is going to run out
of move points.

___

Reply to this item at:

  

___
  Mensaje enviado vía/por Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Update of bug #22567 (project freeciv):

  Status:None => Ready For Test 

___

Follow-up Comment #28:

I have found the probable biggest cause of the bug. pf_position::turn for
pf_reverse_map is now like other pf_map accessors. But if the unit can exactly
reach the tile spending all its moves left, then the turn counter is increased
by 1. This behavior is present since revision 10626. Personally I don't like
it. I maybe will try to change it. assess_danger_unit.patch should make the AI
defending again.

In assess_danger.patch:
* check for utai->carries_occupiers instead of utype->transport_capacity ;
* removed duplication of code for grave_danger, including the removal of extra
"break" (bug that came from patch #4831, probably another cause of this one)
;
* removed useless operation.


(file #22170, file #22171)
___

Additional Item Attachment:

File name: assess_danger.patchSize:1 KB
File name: assess_danger_unit.patch   Size:0 KB


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Follow-up Comment #27, bug #22567 (project freeciv):

Tested with revision 25715 shows the AI to build units for attacking and for
defending. So the cause is probably something in revision(s) > 25715.


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Follow-up Comment #26, bug #22567 (project freeciv):

I have observed:
* The AI doesn't make units for attacking other players (with or without
reverting) ;
* The AI doesn't make units for defense (with or without reverting) ;
* The AI sometimes (but not always) defend against barbarians with reverting
the patches, but doesn't defend at all without.

Also noticed that patch #5201 doesn't help.


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[Marko Lindqvist]

Follow-up Comment #25, bug #22567 (project freeciv):

> I have rebuilt an autogame with reverted patch #4970 and
> patch #4971. Attaching autosave at turn 800 shows that
> barbarians nearly invaded all the world.

There's several problems in AI vs barbarians equation, but reverting those
patches makes AI to at least to attempt defending. I recommend testing as
global observer to autogame with timeout 10-20 to see how AI does not get
defenders even when barbarian unit is standing next to the city to conquer.
Low 'onsetbarbs' helps.

I've been using my own rulesets, which is almost identical to variant2
(available via modpack installer).
I'll create an autogame to demonstrate this unless you can make one yourself
sooner.

As for my last night comments about patch #4970, I probably read the patch the
wrong direction (reading the reverting patch -> what I thought was removed,
was actually added)

___

Reply to this item at:

  

___
  Message sent via/by Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Follow-up Comment #24, bug #22567 (project freeciv):

>> Umh, what else there was in patch #4970 and patch #4971 to
>> explain this? Or do you think it's dormant issue that pf
>> changes activate?
>
> I don't know. I have posted the reverse map (new version) at
> patch #5201. If it really fixes the symptoms, I would say I
> have done a mistake testing without patch #4970 and patch #4971
> (forgot to compile?). 

I have rebuilt an autogame with reverted patch #4970 and patch #4971.
Attaching autosave at turn 800 shows that barbarians nearly invaded all the
world.


(file #22159, file #22160)
___

Additional Item Attachment:

File name: TEST-T0800-Y02375-auto.sav.bz2 Size:87 KB
File name: TEST.serv  Size:0 KB


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[pepeto]

Follow-up Comment #23, bug #22567 (project freeciv):

> Old pathfinding code used to consider effects of veterancy and
> bonuses (including tile and player specific requirements) to
> the attacker speed. New code just assigns unit type's base
> move_rate.

This is not right. Old pathfinding code for reverse map was ignoring movements
effects. I forgot it in the first implementation. This was one of the higest
reason I did patch #4970.

I forgot however to change the cost adjustment line in
pf_reverse_map_get_costs():

if (cost > param->utype->move_rate) {
  cost = param->utype->move_rate;
}


Another change in patch #4970 is the use of utype->unknown_move_cost when the
map is not known.


___

Reply to this item at:

  

___
  Message posté via/par Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev

[Freeciv-Dev] [bug #22567] AI does not build/buy defenders as enemy approaches

[Brandon J. Van Every]

Follow-up Comment #21, bug #22567 (project freeciv):

Civ4 had Jaguar units as the Aztec special unit.  It was a warrior that moved
2 spaces instead of 1.  Solving the problem by making specific units might be
a better approach as far as getting an AI to understand it.  Allowing all
units to move faster seems rather overpowered.  There's some precedent in
SMAC, where if you achieved Elite experience you got bonus movement.  Typical
Spartan strategy was to massively train units so that they would be Elite. 
Such units are extremely powerful.  I don't think having an option to just
"flip a switch" to get that kind of benefit is a good game design at all. 
There should be some kind of significant cost or delay to get that kind of
capability.  Otherwise what you'll have is a goofy game where the Tribal
faction always wins.

___

Reply to this item at:

  

___
  Message sent via/by Gna!
  http://gna.org/


___
Freeciv-dev mailing list
Freeciv-dev@gna.org
https://mail.gna.org/listinfo/freeciv-dev