Game crashes when AI tries to dismantle conquered buildings

[wl-zocker]

After a while, the game crashes and I am back to Windows. There is no error message, but it seems only to happen at war. I tried it twice, here are the stdout.txt:

Cmd_EnemyFlagAction::execute player(2): flag->owner(4) number=31
...
Cmd_EnemyFlagAction::execute player(2): flag->owner(4) number=6
...
Forcing flag at (105, 135)

--------------------------------------------------------------------------------------------------------

Forcing flag at (117, 87)
TW(99141): destination disappeared or economy mismatch -> fail
...
TW(99000): destination disappeared or economy mismatch -> fail
...
TI(98438): destination disappeared or economy mismatch -> fail
...
Forcing flag at (59, 121)
TW(100372): destination disappeared or economy mismatch -> fail
TW(100373): destination disappeared or economy mismatch -> fail
TW(100374): destination disappeared or economy mismatch -> fail
Cmd_EnemyFlagAction::execute player(4): flag->owner(2) number=4

-------------------------------------------------------------------------------------------------------

The ... stands for things like
 get atk_ok_e
 get eva_ok_w
 get eva_fail_e
 get atk_ok_w
 get eva_ok_w
 get atk_ok_w
 get atk_ok_w
 get atk_ok_e
 get eva_fail_w
 get eva_ok_e

I did not consider them to be important.

I am on Windows 7 and tried that with the build17-rc1. The game did not crash with bzr6334 which I had before.

I have not yet found a way to reduplicate it in every case, just start a game and wait until there is a fight (I did not fight, I just watched the others). Sometimes, the game crashes

[Hans Joachim Desserud (hjd)]

Ouch. This sounds really bad. Could you please provide a savegame from just before it crashes or a replay of the game?

> I am on Windows 7 and tried that with the build17-rc1. The game did not crash with bzr6334 which I had before.

Thanks, this is valuable information. I wonder if the release candidate is built differently (release vs debug mode?), but I don't know.

[wl-zocker (wl-zocker)]

Here is a save game. You will probably have to let it play several minutes before it crashes. It was played on a Riverlands map I have changed a bit (added an eighth player to watch the others). Is it important whether the map is an official of a changed one?

[wl-zocker (wl-zocker)]

I have just tried it out with an official, non-changed map. It took some time, but finally the game crashed.

[Hans Joachim Desserud (hjd)]

>Is it important whether the map is an official of a changed one?
Shouldn't really matter as we obviously want all maps to work.

Thanks for providing the save game. I loaded it here and got a crash (see the attached backtrace). However, this indicates that the problem somehow resides in dismantle site. This may explain why it is so hard to trigger and also why I got it the first time I loaded the game, but not the second time (waited longer even). I hope this is the same issue and that someone will be able to look into it.

[Venatrix (elisabeth)]

Could be the same error message as here. I just read "floating point exception" (don’t have the debug mode atm).

[SirVer (sirver)]

The problem is that the computer player tries to dismantle a conquered building which is forbidden. There used to be a bug which had this problem for human players - the button is no longer shown now - but the AI does not seem to check this.

This is indeed a critical bug which will make a rc2 necessary. Sad story but lucky that you found this!

[Hans Joachim Desserud (hjd)]

>There used to be a bug which had this problem for human players - the button is no longer shown now - but the AI does not seem to check this.

Bug 886808 / r6083.

[SirVer (sirver)]

Fixed in r6345. I let the savegame run for 38 game hours and the game didn't crash anymore.

[SirVer (sirver)]

Released in b17 rc2.