Sadly your repeatable CTD (which was repeatable for me too) wasn't the long-standing city razing case (check your last upoad - the minidump was from a day earlier than the save game - they didn't correspond). However, it WAS a repeatable crash related to the recent memory changes, so still a good catch. Fix is now up on the latest SVN.
v22 bugs and crashes
- Thread starter strategyonly
- Start date
Sadly your repeatable CTD (which was repeatable for me too) wasn't the long-standing city razing case (check your last upoad - the minidump was from a day earlier than the save game - they didn't correspond). However, it WAS a repeatable crash related to the recent memory changes, so still a good catch. Fix is now up on the latest SVN.
Fixed the double unit highlighting (latest SVN).
Slowdowns are probably due to the new AI work I pushed last week, and haven't yet had a chance to tune due to the memory issues that swamped me last week. If you have a particular save where you feel end-turn processing is goign slower post it and I'll analyse it.
I can't reproduce this. If I select this stack and click the target plot in your images it takes the short path. Precisely (down to left vs rigth clicks, clicks and holds, whatever) did you do to get the picture that shows the long path. I've tried right click-and-hold, and just right click alone, without reproducing it...
strategyonly
C2C Supreme Commander
As long as the double unit thing has been solved, no rush on the others right now, i know you are swamped like me, so no biggy, right now.
Glad you got that fixed, i was going crazy, trying to figure out what unit i was using
Partly this is just AI dumb-assedness that needs some serious work to correct, but partly it was a plain old bug, which I have corrected.
The dumb-assedness lies in how it evaluates if the city is threatened (and basically it cannot see your 3-tile away horsemen with 3 movement points for the purposes of that check).
The simple bug lies in its passive distribution of troops between local defensive territory and the city itself. It's never supposed to take more than half the total defensive force it has for that city out of the city itself, but a bug means that troops 2 tiles away north or east, which themselves think they are part of the city defense are not seen in totalling the available defenders, and so don't count towards the total, half of which must be in the city!
Current SVN fixes this, and should cause the AI to deploy more of its defenders in the city than before.
Dancing Hoskuld
Deity
For the longer path I just clicked on the target plot. Even when I selected the shorter path by hand it still took the long path. It happens fairly often with stacks which contain units with different travel speeds. I will try and capture the next one.
...and it reproduces for you from that save game?
Sleeeper
Warlord
Very long turn times (@ Stardard Earth Map), around 5-10 minutes of WfoC each turn with animation halting (like it loops somewhere), and after that it finally passes.
But sometimes (rarely) a turn passes fast as usual (less than a minute).
But sometimes (rarely) a turn passes fast as usual (less than a minute).
The posted save is presumably one of the turns that goes slowly on end-turn...?
Sleeeper
Warlord
Yes, just hit end of turn. Took 5.5 minutes for me, also checked that dynamic entities setting doesn't affect the delay.
Ok, this has definately identified a scaling issue. The time is being consuming in stack attack odds caclulations when one very large stack attacks another very large stack. Basically this is because the odds calculations pretty much simulate a mock battle to assess probable casualties on both sides. In this process, at each stage it has to choose the best attacker, which involves testing each one in turn. For each attacker it then has to select the best defender, so the unit on unit comparisons (which involve taking account of all the promotions etc., so are not just a direct strength number comparison) happen O(N x M) times where N and M are the stack sizes. Furthermore all possible attacks (in a close by area) will be tested, so each large stack is probably involved in several test-battles before a decision is taken on what to actually do. If there are more than 2 large stacks in the area the combinatorics of different stack combinations for different potential target choices will also come in.
Furthermore, the AI remakes its decision for each move, so in the case of a large stack once the top unit has decided it's a good idea and used its turn in the attack, the (rest of) the stack will go through the decision process again, and so on so the entire process is effectively multiplied by the stack size again!
None of this is cached.
In your game there appears a SoD of 65 units attacking an area that has defender stacks of 45 units, 28 units and a few smaller ones.
I should be able to resolve this with some judicious caching.
Dancing Hoskuld
Deity
Sleeeper
Warlord
When sea creatures get XP they now have real sea creature promotions rather than getting land based ones. This can make them very strong very fast. i.e it is a design feature.
Really, 540% against wooden ships, that's like 19 strength, i.e. unbeatable. Also I've seen several sea animals with the exact same strengths. If it's meant to be like that, perhaps it could be toned down a little.
that means that you have to bring stack of 5-6 ships to take it down.
in my previous game i was shocked when i saw my caravel destroyed by some shark, then i closely looked at promotions
In my current game i have war canoes and Great White Shark, who ate couple of my canoes and promoted to +360%, so i have to very carefully plan my sailing now with it lurking around Please don't lower it, it's really a challenge. I would like to see same for ground animals so my uber hunter with 50 promotions would have something challenging to kill.
This doesn't cause any problems for me, but I'm using assets that are about a week out of date, so it's almost certainly an asset issue (which I'm goign to lave for someone involved in asset changes to look at). The fact that your report graphical issues for the new bonus type suggests a good place to start looking....
Thanks, this explains some fast turns when AI is bored to fight
Fix now pushed to SVN (reduces total odds calculation time in your save game from 140 second to 1.7 seconds). Also fixed a couple of performance bottlenecks (processing missionaries and processing spies) that were highlighted by Sgtslick's game (you should fidn that is now running about 25 seconds per turn faster).
