Hasty Aliance - 1st turn never ends

dkulis · Post by **dkulis** » May 21st, 2004, 5:56 am

I have a problem in Hasty Aliance - 1st turn never ends. This is ONLY with with v0.7.7 & v0.7.8-CVS, v0.7.6 works fine.

Symptoms:
- I do my moves for 1st turn
- Li'sar does hers
- orcs do their moves and then nothing more happens. I can scroll etc. (i.e. normal operations during AI turn)

Meanwhile wesnoth.exe process has around 50% CPU usage and keeps on grabbing more memory.

miyo · Post by **miyo** » May 21st, 2004, 7:27 am

Your bug report lacks all the needed details.

What operating system are you using. If GNU/Linux what distribution? Did you install some pre-packaged binary or did you compile yourself? If you compiled yourself what versions of dependency libraries you have? What optimizations you used when compiling?

- Miyo

dkulis · Post by **dkulis** » May 21st, 2004, 9:00 am

miyo wrote:What operating system are you using.

Win2k Pro

miyo wrote:Did you install some pre-packaged binary or did you compile yourself? If you compiled yourself what versions of dependency libraries you have?

- v0.7.6 (the Ok version) - built with VS2003.NET
- v0.7.7 (problem version) - prepackaged Win executable
- v0.7.8-CVS from 2004.05.20 (same problem) - built with VS2003.NET with

- SDL v1.2.7 devel_VC6
- SDL_image v1.2.3 devel_VC6
- SDL_mixer v1.2.5 devel_VC6
- SDL_net v1.2.5 devel_VC6
- SDL_ttf v2.0.6 devel_VC6

miyo wrote:What optimizations you used when compiling?

Default ones with VC (can't specify)

Anything else?

O'Dubhghaill · Post by **O'Dubhghaill** » May 23rd, 2004, 4:40 pm

I'm having the exact same problem on Wesnoth 0.7.7 on debian unstable using the current unstable binary package.

thk · Post by **thk** » May 23rd, 2004, 4:46 pm

Yep, same over here on Mandrake 9.2. There is a huge amount of text-output during the whole game in the terminal I start Wesnoth from and when the 'never ending turn' runs that output seems to be repeating itself, like in some infinite loop (ain't sure, cause it's coming so fast, but it _looks_ the same...)

miyo · Post by **miyo** » May 23rd, 2004, 7:27 pm

Start Wesnoth from terminal with 'wesnoth &> /var/tmp/wesnoth.log'

- Miyo

cipso · Post by **cipso** » May 23rd, 2004, 8:27 pm

I reported same problem.

https://savannah.nongnu.org/bugs/?func= ... em_id=9038

I found that Wesnoth is eating memory at that turn... it ate 160megs (then I quit that scenario).

-cipso

Dacyn · Post by **Dacyn** » May 23rd, 2004, 11:39 pm

Well this is happening to me too, with Debian Linux. Using debug I discovered that this happens right after the troll leader is moved, when there are only 2 trolls left to be moved.
This is probably an AI problem. Here is a screenshout including AI output:

jyrierik · Post by **jyrierik** » May 24th, 2004, 3:26 am

I'm having the same problem with 0.7.7 with the Mac OS X version running X.3.3 (Build 7F44). Goes into never-neverland after the visible Orcs/Trolls finish their move & uses up as much free processor cycles as it can grab according to "top". Previous versions ran fine (although I sometimes had some sluggishness from time to time). Looks like an OS independent problem.

Jyri

jbmesserly · Post by **jbmesserly** » May 24th, 2004, 5:13 am

It looks like the AI is caught in an infinite loop. It can't get out of ai::do_move(). In particular, it's looping in the acquire villages movement phase. That part calls ai::get_villages(...), which for some reason returns "got_village == true" endlessly.

I'll see if I can figure out why that's happening.

jbmesserly · Post by **jbmesserly** » May 24th, 2004, 5:50 am

Problem fixed in 0.7.8-CVS.

The fix is for get_villages(...) to return true if it *actually* makes a move, not just that it *tried* to make some moves. Otherwise, it got stuck "trying" to make a move forever.

Of course, the real question is why get_village_combinations(...) is getting caught in a loop returning the same invalid moves over and over again. I'm not sure if that needs to be looked into or not; it suggests that that function might not be working properly.

In any case, it's not looping now and the AI seems to be making reasonable moves. I'll save further investigation for another day (or night).

Dave, if you have any ideas what's going on with the new AI village code, let me know.

jbmesserly · Post by **jbmesserly** » May 24th, 2004, 5:53 am

p.s.-- see the comment of my CVS commit to src/ai.cpp, it goes into more gory detail about the problem and the fix that I added.

Post by **Dave** » May 24th, 2004, 1:36 pm

Hmm...I'm looking into what the problem is.

get_villages() should no longer have to cause the ai to be recalculated from scratch anymore, since it generally does all the moves it has to, instead of just one move.

The only time it should return invalid moves though, is if someone got ambushed, but that clearly isn't happening here, so I'll have to look into that.

David

Eponymous-Archon · Post by **Eponymous-Archon** » May 27th, 2004, 5:07 am

miyo wrote:Your bug report lacks all the needed details.

Well, a few details anyway.

Sith, any chance for a recompiled Mac version with the CVS bugfix in there?

I know you're just sitting around looking for something to do.

The Battle for Wesnoth Forums

Hasty Aliance - 1st turn never ends

Hasty Aliance - 1st turn never ends

same problem (bug 9038)