Post Feb 1st 2013 - bugs - Single Player

[KMRblue1027]

In the past 20 turns or so I have experienced a massive slowdown (going from about a 30-40 second turn to over a minute).

https://dl.dropbox.com/u/25874166/Reardon%20BC-7091.CivBeyondSwordSave

 

[ls612]

In the past 20 turns or so I have experienced a massive slowdown (going from about a 30-40 second turn to over a minute).

https://dl.dropbox.com/u/25874166/Reardon%20BC-7091.CivBeyondSwordSave

On the SVN or with vanilla V28?

 

[Koshling]

1. The wonder "Famen Temple" makes your missionaries always successfull. A small bug in the text messages, when a religion spread would fail, then it is first showed "Your missionary failed.." and afterwards "Hinduism has spread..."
2. AI workers with the same cycles from the normal civ IV: first farm, then workshop, then farm on the same tile... My example has one building a cottage over some grains, another farming a forest nearby, and the third burning a second forest without ressources (no hammers from burning, right?)
3. When using viewpoints, occasionally the resources and improvements are drawn higher on the map.
4. Not a bug, but is there any reason not to build all the religions I get in my capital, since the prophets create holy cities?
Save and screenshot for No. 2 attached if someone's interested.

Sadly, if you force the workers in this save to recalculate from the save position hey don; start on the same builds, so the decision point (when they started that build) is lost, and not debuggable from this save.

What I need is save from a turn where an AI worker STARTS on such a build when you press end turn. Meanwhile I have added an AI log of the form:

City <NAME>, tile (<X>,<Y>) best build changed from <OLD BEST BUILD NAME> to <NEW BEST BUILD NAME>

in an effort to help track this down, or at least to see how prevalent it is. If you (or anyone) has AI logging on, occasionally scanning it for these messages (with the latest SVN once I push the change to add this log later today) we'll at least get an idea of how much changing of the AI's view of the best build is and hopefully a pattern will emerge that will help track this down.

 

[sp00n]

The "No Storms" checkbox in the C2C options doesn't seem to be working correctly. I'm still losing ships to storms.

 

[KMRblue1027]

On the SVN or with vanilla V28?

SVN updated just yesterday.

 

[ls612]

SVN updated just yesterday.

I noticed the same thing but profiling showed that nothing had actually changed. Koshling did though yesterday fix a bug with AI unit values which saved about 10% of a turn.

 

[Dancing Hoskuld]

In what sense? Do you mean if you ask what their cost is from the BuildingInfo? If so then I doubt it - it'll just return whatever value is in the XML.

There is a python method that asks if the building is free to the player. the actual test I am using is

if (not pPlayer.isBuildingFree(iType) and gc.getBuildingInfo(iType).getProductionCost() != -1):
	[B]buildingList[i] = (gc.getBuildingInfo(iType).getDescription(), iType)[/B]
	i += 1

It is falling over on the line highlighted and has only started doing so since the barter post became auto build. Edit I have now set up some trace prints so I may be able to find out exactly what is happening.

The "No Storms" checkbox in the C2C options doesn't seem to be working correctly. I'm still losing ships to storms.

The No Storms option means that storms do not appear on the map, grow and spread and cause damage to ships. It has nothing to do the the storm event. It would be possible to remove that event if the option is on tough.

 

[Koshling]

In the past 20 turns or so I have experienced a massive slowdown (going from about a 30-40 second turn to over a minute).

https://dl.dropbox.com/u/25874166/Reardon%20BC-7091.CivBeyondSwordSave

I'm about to profile this save to see if anything leaps out...

Edit - do you happen to have a save from 20 turns or so ago when it was playing faster? It would be useful to determine if the slow down is due to an SVN update, or just to a change in your game (e.g. - AIs starting to make contact with one another, exchanging those contacts, and diplomacy time going up a lot as a result)

Edit 2 - nothing particularly stands out in the profile of the turn you posted. Time to do the auto-save is about 12 seconds, so setting your auto-save to be less frequent might help, assuming you currently have it set to every turn. Apart from that there are a few things that can are probably worth optimizing, but I don't see any particular hot-spot that would explain a sudden jump. If you have the earlier save for comparison that might help pin-point it. You could also try using a DLL from whatever earlier revision you were using before to see iof that makes THIS save go faster (again to help determine if the issue is a change in the game, or an SVN change)

 

[KMRblue1027]

Here's from where I started last night (Not sure how long ago it was). What I can note is I got 2 messages about civs spawning close to me from barbs.

https://dl.dropbox.com/u/25874166/AutoSave_BC-7329.CivBeyondSwordSave

Also I do not have auto every turn because it was causing desync's in multiplayer a few days ago in my v28 game with a friend.

Edit: Just in case you need it here's a save about half was in between the two
https://dl.dropbox.com/u/25874166/AutoSave_BC-7161.CivBeyondSwordSave

Edit 2: If it helps I figured out I was on SVN 4787 before and the revision that removed the brick mason's obsolete tech last night.

 

[ls612]

So overnight I was testing some progression stuff with AI autoplay, and found this really wierd bug. It may be related to Koshling's changes yesterday to Max Compat saving.

Basically, the game had a MAF because it tried to save a game bigger than 10,000 KB (at least that is what I'm assuming, the file was 9999 KB). However, when I loaded that game and saved it manually it was about 4000 KB, which is far more reasonable for a game in the late Classical era. Anyways, I've attached both saves (they load the exact same, but one is way bigger). They compress to the same size though oddly.

 

[Koshling]

So overnight I was testing some progression stuff with AI autoplay, and found this really wierd bug. It may be related to Koshling's changes yesterday to Max Compat saving.

Basically, the game had a MAF because it tried to save a game bigger than 10,000 KB (at least that is what I'm assuming, the file was 9999 KB). However, when I loaded that game and saved it manually it was about 4000 KB, which is far more reasonable for a game in the late Classical era. Anyways, I've attached both saves (they load the exact same, but one is way bigger). They compress to the same size though oddly.

Manual and auto save you the exact same code, so I don't think it can be that

 

[ls612]

Manual and auto save you the exact same code, so I don't think it can be that

Well, those files are identical except one was auto-saved by Beyond the Sword and one was saved by me hitting Ctrl-S. They both load to AD 0045 in the same game.

 

[Koshling]

Here's from where I started last night (Not sure how long ago it was). What I can note is I got 2 messages about civs spawning close to me from barbs.

https://dl.dropbox.com/u/25874166/AutoSave_BC-7329.CivBeyondSwordSave

Also I do not have auto every turn because it was causing desync's in multiplayer a few days ago in my v28 game with a friend.

Edit: Just in case you need it here's a save about half was in between the two
https://dl.dropbox.com/u/25874166/AutoSave_BC-7161.CivBeyondSwordSave

Edit 2: If it helps I figured out I was on SVN 4787 before and the revision that removed the brick mason's obsolete tech last night.

I'm not measuring any significant difference between the turn times of the earliest and latest save (about 2 seconds) - does it still run faster for you if you reload the older autosave and play a turn? If it doesn't that would imply it must be due to SVN changes - do you know what revision you last saw running faster?

Edit - I then backed off to a version of the DLL from before all the recent changes (rev 4756) and tried it there. That took about 10% LONGER then the current version does, which is what I would expect, since there have been a number of optimizations since then.

Net result is that I can neither reproduce nor explain any increase in times you have seen.

 

[ls612]

Manual and auto save you the exact same code, so I don't think it can be that

This issue is repeatable. I autoplayed 100 turns from this and it started getting much bigger again in the autosaves.

 

[KMRblue1027]

I'm not measuring any significant difference between the turn times of the earliest and latest save (about 2 seconds) - does it still run faster for you if you reload the older autosave and play a turn? If it doesn't that would imply it must be due to SVN changes - do you know what revision you last saw running faster?

Edit - I then backed off to a version of the DLL from before all the recent changes (rev 4756) and tried it there. That took about 10% LONGER then the current version does, which is what I would expect, since there have been a number of optimizations since then.

Net result is that I can neither reproduce nor explain any increase in times you have seen.

I did another update and it seems to be running faster again. Maybe it was the fact that I had the game open for hours before actually playing (RAM running low or something of that nature). Well thanks for the help anyway. Your work with the AI has made them legitimately scary to fight against.

 

[Koshling]

This issue is repeatable. I autoplayed 100 turns from this and it started getting much bigger again in the autosaves.

I've looked through the load sequence of your 10M save, and there is nothing odd about the dictionary entries (if it was related to that change I would expect lots and lots of spurious repeated dictionary entries being the extra space). In fact, having spent several hours looking into it, I can find nothing odd about it at all, so my only theory is that (for some reason) the game engine is not compressing the stream. However, I'm pretty sure it has nothing to do with the dictionary change I made, so it remains mysterious.

If you can find reproducible recipe then let me know (i.e. - one that can be deterministically run in 10-15 minutes, not one that requires an afternoon of autoplay for each test run)

 

[ls612]

If you can find reproducible recipe then let me know (i.e. - one that can be deterministically run in 10-15 minutes, not one that requires an afternoon of autoplay for each test run)

Based on my analysis of that save it is getting very slightly too big every turn, but not enough to be noticable until you've had it on all night. But yeah, the only way I've reproduced that dramatically it is to have it save a lot over a long period of time.

 

[Koshling]

Based on my analysis of that save it is getting very slightly too big every turn, but not enough to be noticable until you've had it on all night. But yeah, the only way I've reproduced that dramatically it is to have it save a lot over a long period of time.

Does its size (the autoplay) 'reset' to normal on each session? i.e. - is it something that is just caused by long sessions maybe? (which would probably not be new)

 

[ls612]

Does its size (the autoplay) 'reset' to normal on each session? i.e. - is it something that is just caused by long sessions maybe? (which would probably not be new)

I'm not quite certain what you are asking. If you mean does it save normally if it is loaded and then re-saved then yes. But if I go in at the end of a session and save manually it is also the normal size, so session length can't be the only factor involved.

 

[strategyonly]

Had this error message show up and i added the NEW schema to it??? (Sub Combat Types???)