Spring RTS Engine

Well, I didn't notice any significant differences but I suppose it's hard to tell just like that.

I tested with a few different units in BA7.19. Generally, I found pathing to be alright but every now and then, units got stuck when faced with a target destination very near to them but in reality, would take a lot longer to get to because they can't directly travel to their destination.

I took a screenie:

EDIT: These observations are all pre test build. As in, the current version that's up on the server. I could probably help you guys test once I'm done with classes for the day...

As far as I can remember, Jaz, that problem you described has always been a problem in spring's pathing (and, as a matter of fact, in a lot of other games too). I habitually and manually tell units to go around cliffs like that.

The main thing to fix, IMO, in the current pathing setup is units getting stuck in open plains. I've often told units to just nagivate across an open terrain, and they'll kinda get half-way stuck in the ground. I then have to tell them to go backwards to get unstuck, and then they can move again (usually). Sometimes, though, they'll just get restuck again in the same spot when I tell them to move forward again. I've been nuked a few times now due to my anti not moving into the area I told it to because it got stuck in the middle of open ground (or stuck on the corner of a building sometimes).

Oh, and I have easily reproducible pathing bug. In BA, make a T1 vehicle factory. Make the factory's waypoint pretty far from the factory. Then, make 1 stumpy followed by 1 jeffy. The stumpy will go to its destination normally, but the jeffy will attempt to use the exact same path as the stumpy, and it'll bounce off the back of the stumpy forever until it reaches its destination.

Also, if you get a large (~40) group of units like AKs and tell them to go somewhere, sometimes a large portion of the units will stay in place while the units on the edges will move to the selected location. You have to replace your waypoint a few times before they'll all move. It seems to be related to the bug where a unit gets stuck in an open plain, but apply it to a large group of units.

Hmm, another thing I noticed on this build is that some units freeze for maybe 0.25-0.5 seconds every now and again when moving. It's like they keep getting tasered.

Jazcash wrote:Hmm, another thing I noticed on this build is that some units freeze for maybe 0.25-0.5 seconds every now and again when moving. It's like they keep getting tasered.

That would seem to me that the unit was following a path, but then pathing was recalculated for whatever reason and the new path required the unit to pause and turn slightly or accel in a different direction before continuing, hence making the unit freeze.

gonpost wrote:As far as I can remember, that has always been a problem in spring's pathing (and, as a matter of fact, in a lot of other games too). I habitually and manually tell units to go around cliffs like that.

It didnt use to be a common problem at all, compared to how everyday happening its now.

Havent tested .82.6.1 yet, will do so soon. But 1 thing, that maybe hasnt been said yet, that apparently is unchanged in .6.1., when labs now automatically assign a waypoint in front of them that you cant remove, thats not good. Since not all labs send units only to the direction they're facing, for example TA airlab (unit spins and leaves in any direction), or TA bot lab (unit can exit thru the back too), in those cases it can be very undesirable for them to take the time to reach that point before player-assigned waypoint.

Carpenter and I tested the pathing on Talus (very hilly and hard to navigate) and unit response looked very good, I highly recommend testing it before you post here, as units getting randomly stuck on ground or features was not encountered.

Edit: Forgot to ask, alt+b no longer displays the pathing, how to change this?

Ok, so there is a map with very error prone pathing: LLTAComplexV2

Sending fleas or peewees across the top will result in them getting stuck in rocks, even in this version hoijui posted.

My own so far limited testing in S44, for problem 3 that Godde identified:

http://springrts.com/phpbb/viewtopic.php?f=11&t=24312

Is that the new code is an improvement, but not yet a return to the correct behaviour of 82.5

Edit: hadn't seen the new 'Godde-III' fix. Will test immediately!


Edit2: Some improvement to previous, units tend now to get stuck on corners of obstacles instead of driving right into walls.

Infantry pathing around features generally seems improved.

my own tests with Godde-III fix (== 0.82.6.1-17) and S44 1.5 on Brazilian Battlefield: i built about 50 infantry and 20 tanks. after about 5 minutes of testing, i only managed to get the tanks stuck once. moving all of them around corners and through narrow passages (only 1 tanks at a time fits through).

I think it may be partly due to the strange yardmap of our tanktraps.

Tested and all seems to be okie dokie. Nothing got stuck with me ordering 100 tanks in line moves randomly across the entire map on greenhaven.

This is with 0.82.6.1-17

TurnInPlace = 0 is completely broken. Units will turn and accelerate at the same time but when they are close to pointing in the correct direction they will slow to a stop and complete their turning.

Heatmapping still does not work. Units often walk backwards and spread out too much.

Pathing through gratings and around walls is better than in 0.82.6.1.

On an unrelated note, in BA, units created in factories turn to the left slightly as soon as they're built which makes them sluggish when they try to exit the lab. Is this anything to do with pathing?

Jazcash wrote:Hmm, another thing I noticed on this build is that some units freeze for maybe 0.25-0.5 seconds every now and again when moving. It's like they keep getting tasered.

I've noticed that in what I'm working on too, but in my case its due to weapon1 having a shorter range than weapon2 (or 3 or 4 or whatever) and it only happens when the unit-AI is navigating the unit to the target on its own (ie, it spots an enemy unit, and is moving to engage without the player having put an attack order on the target) this is not new.

Google_Frog wrote:This is with 0.82.6.1-17

TurnInPlace = 0 is completely broken. Units will turn and accelerate at the same time but when they are close to pointing in the correct direction they will slow to a stop and complete their turning.

Heatmapping still does not work. Units often walk backwards and spread out too much.

Pathing through gratings and around walls is better than in 0.82.6.1.

Hmm, I didn't notice anything like this happening however perhaps it was just on a scale small enough that I didn't notice. I'll maek moar test.