Massive slowdown in AI processing in 1.9.9.. not in 1.9.8

[PorkSol]

There is a massive and crippling AI turn processing slowdown in 1.9.9. This is not present in 1.9.8.

I've attached a turn 3, 1.9.9 save from THoT - Invaders.

In ThoT - Invaders, there are bandits that pop out of villages when you or the AI takes the villages. In the attached save game, the AI will trigger some of these ambushes if you hit end turn a couple times.

In 1.9.9, once the Orc AIs have triggered several large packs of bandits, each Orc AI will take several minutes to think about his turns. Wesnoth will start using far more CPU than normal during this time. The cursor will turn into an hour glass and nothing will happen for a few minutes, then the AI will move a few units.

Switching back to 1.9.8, loading the same save (even though it is a 1.9.9 save), the Orc AIs will process their turns normally with no slowdown, despite triggering the same ambushes and spawning just as many bandits. The Orc AIs will start making their moves immediately, as normal.

Versions used: 1.9.9, Windows, downloaded from the front page. 1.9.8, Windows, downloaded from the front page.

Computer: Pentium 1.7Ghz, 768 mb of RAM, Windows XP service pack 3, plenty of free HD space. (Wimpy computer, but it runs scenarios much bigger than this one without anything like this kind of slowdown. I've never see the AI take anywhere near this amount of time to think about its turns).

[zookeeper]

If it only happens on the orc turns, then there's the possibility that the bandit-spawning code goes into an infinite loop when an orc captures a village (which still terminates after a while since WML loops have a maximum number of iterations) due to some WML-related function having gotten broken in the engine. A WML bug sounds very unlikely since the scenario hasn't been touched in over a year.

Does anything similar happen in other scenarios?

[PorkSol]

Does anything similar happen in other scenarios?

Not entirely sure if it is related or not, but I do recall a similar but less reproducible and less severe issue in Delfador's Memoirs - Scenario 2: "This Valley Belongs to Me" under 1.9.9. I believe it just manifested as a turn or two slowed AI processing with the hourglass mentioned above. It was much slower than anything I'd seen under 1.9.8, but no where near as bad as in THoT - Invaders. I don't believe it showed up reliably every time like it does in Invaders either.

I've only played a few scenarios under 1.9.9. though. THoT 1 - At the East Gate, THoT 3 - Strange Allies and THoT 4 - Troll Bridge didn't have the issue.

[doofus-01]

For what it's worth, I've noticed this slowdown too, in several places, though I'm not sure yet that it is not WML bugs. However, it does seem to be new to 1.9.9.

[averyimaginativename]

May or may not be related, but this appears to happen when there are no good ai moves. Mostly in UMC, but the final scenario of DiD has a similar problem if you move Mal Keshar advanced to an ancient lich, with no recruits towards an Elvish foolish hero. The delay here was only six or seven seconds per unit, but noticeably slower than other scenarios. The elves also refuse to attack (though they will attack, and the slow down doesn't happen if Mal Keshar is just a level 3 lich).

[Crab]

Thanks for the report, I'll take a look

[Crab]

AI attacks involving leader were taking too long to calculate. Should be fixed by [rev]52543[/rev]