Merchant ships don't want to trade on one specific map

[sanderd17]

As you may know, I was working a bit on a map. It had a big sea part, which I wanted to use, so I thought I'd create a dock on the other side of the sea, where the players can trade with. That's how I found most bug I reported lately.

Anyway, there's one particular bug that I don't even understand. I don't succeed to create a minimal example of it, as it only happens in the map I created, and only when you follow the right steps.

I've added the pmp and xml file in a zip to the post, so you can also test.

Steps to reproduce:

• start playing as player 1 (Britons), you can set the bot to unassigned
  
• upgrade to town level
  
• Build a dock on the british side of the ground, about as far as you can from the Gallic traders (to gain more profit)
  
• Set the rally point of the dock to the ally gallic dock, on the other side of the map
  
• Create one mershant ship
  

Normally, when the merchant ship is created, it should automatically start trading with the Gauls, but this doesn't happen. The ship just stays where it is.

• Try to move the ship around, that works without problems. So I think it's not a pathfinding problem
  
• Try to manually assign a trading route between your dock and the Gallic dock
  

The ship starts trading, but when it first arrives at the second dock, it just stops. I've included a screenshot with the state when the ship arrives at the second dock. No matter what I do,I can't get the ship to deliver its metal.

As said, I've tried creating minimal examples, but that doesn't work. Even if I place the dock and merchant ship in Atlas instead of letting player 1 build it, the ship wants to trade. But other simple things I do, like saving the map again don't change these buggy results.

Is there someone willing to look after such a strange bug? I'm using a build from yesterday btw.

HadriansWall.zip

Edited May 31, 2013 by sanderd17

[historic_bruno]

Actually I would say it is a pathfinding issue. It's stuck in the "approaching" UnitAI state, so that's an indication. One of the reasons for redesigning the long pathfinder is that units get stuck in certain situations, places they can get into but not out of with the long pathfinder, though the short pathfinder can sometimes get them out, this could be one of those bugs. I would advise altering the terrain if possible to make pathfinding easier on your map.

[historic_bruno]

The other thing you can do is view the pathfinder overlay, that can reveal how tricky the terrain is and also what the pathfinder is doing.

[sanderd17]

I thought about a pathfinding issue, but why doesn't it deliver the metal when it arrives at the second market? Is that also because it doesn't find a global path back? And then how does it find the first route (from the first markef to the second one) as a beginning? It all seems a bit strange. Maybe I should just wait for the new pathfinder.

[alpha123]

I thought about a pathfinding issue, but why doesn't it deliver the metal when it arrives at the second market?

I think because it never actually reaches the second market, according to the pathfinder. It just gets stuck really close, even though the model is at the second market the pathfinder thinks it isn't quite there yet.

[sanderd17]

I looked at it a bit closer, and in the unit status position, I saw that the x and z coordinates were still changing (should z change for something floating? Or is y pointed upwards?). But it only changed with about 0.0001 units per second. Meanwhile, I've also had some ships that started the trading route quite well (going up and down a few times), but then got that same problem of being stuck, but on the market on the other side.

[FeXoR]

I'm not as sure as I thought first x)

But I think it's:

Positive X is right, positive Y is bottom (or top, not sure!) and positive Z is backwards (at default rotation) in the engine.

(If Y is top that would be a left handed coordinate system which would be strange...

On the other hand Y not beeing top seams strange to me as well...)

I'm quite sure about the general directions though not so sure about the positive directions.

If this is confirmed I'd like to add that to http://trac.wildfire...oordinateSystem

That way it can be found at least somewhere.

An engine documentation would be better ofc. but I couldn't find a fitting one (only the doxygen documentation and the quite outdated one not saying anything about the coordinate systems directions).

Edited June 21, 2013 by FeXoR