Big Bug thread for 0.82.2/3 - Report bugs here - Page 3

Big Bug thread for 0.82.2/3 - Report bugs here

Discuss the source code and development of Spring Engine in general from a technical point of view. Patches go here too.

Moderator: Moderators

User avatar
Carpenter
Posts: 216
Joined: 10 Jul 2009, 16:07

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Carpenter »

http://www.uploadgeek.com/image-0ECA_4C6D199F.jpg

What's that blue square that is flashing constantly ingame? I managed to capture this pic because I rejoined a ffa game that was still running and I saw the whole game in pictures with 2 fps.

Anyway this blue square appears for me in every game, same for couple of other players, dont know if for everybody.
Last edited by Carpenter on 19 Aug 2010, 16:45, edited 1 time in total.
User avatar
JohannesH
Posts: 1793
Joined: 07 Apr 2009, 12:43

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by JohannesH »

I'd be interested to know, with the new heatmapping+whatever else pathing change, what was the exact goal?

How should the units behave, if it worked right?
imbaczek
Posts: 3629
Joined: 22 Aug 2006, 16:19

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by imbaczek »

master doesn't link (possibly only spring-multithreaded is affected) under windows, due to windows' limitations (command line length exceeded.)

possible solutions: split spring into one library per subsystem.
thelawenforcer
Posts: 106
Joined: 20 Nov 2008, 18:00

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by thelawenforcer »

the pathing is just absolutely horrendous. NOTA ground wars are pretty much unplayable atm.

What i dont understand is how something like this got missed in testing, its painfully obvious that something is wrong with the pathing as soon as u try to move a group of units...
User avatar
hoijui
Former Engine Dev
Posts: 4344
Joined: 22 Sep 2007, 09:51

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by hoijui »

imbaczek, maybe it works if you use msys, cygwin or something like that?
sorry but.. i .. have little mood to test on my windows machine and look for a solution for that. :/

one library per subsystem, do you mean: Sim, System, Rendering, ... ?
if so, that is planned in the long run would it be possible already now, even doh the subsystems are completely woven together?
guess yes, with static libs... i put it on my todo list. should not be too much of an effort.. will use a branch at first. too tired today.

thelawenforcer, two explanations:
  • you did not test any of the RCs!?!?!
  • you are testing it right now!
there are no other options, and the second one is already only me being nice.
User avatar
JohannesH
Posts: 1793
Joined: 07 Apr 2009, 12:43

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by JohannesH »

hoijui wrote:thelawenforcer, two explanations:
  • you did not test any of the RCs!?!?!
  • you are testing it right now!
there are no other options, and the second one is already only me being nice.
  • Did the dev who made the pathing commit test it?!?
  • If they did, and didn't notice anything wrong, are they experienced enough with the subject to yet alter the pathfinder of a major release?
  • I did not see big pathfinding changes marked in RC changelogs (might be fault of my eyes though), to test it
Sorry but I simply can't understand why such a change would be made... It does not give responsive or predictable unit behavior at all

Btw where could I get some info on the basic idea behind implementing heatmapping, I mean I understand how it works, but not why it'd fit into RTS pathfinding. Why would less path overlap make units behave more intelligently here? Is this sort of "intelligence" desirable for an RTS unit?
User avatar
BrainDamage
Lobby Developer
Posts: 1164
Joined: 25 Sep 2006, 13:56

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by BrainDamage »

JohannesH wrote:Why would less path overlap make units behave more intelligently here? Is this sort of "intelligence" desirable for an RTS unit?
a pathfinder usually searches for the shortest path, while that's nice and dandy with a single unit, it can be a problem when multiple units travel together, if the pathfinder is oblivion to multiple units ( left case ) and fetches each unit's path individually, they'll avoid eachother in the chokepoint only by collision detection ( read: after bumping into eachother ), while with a subtile usage of heatmapping, they'd avoid eachother paths and walk in parallel across the chokepoint ( right case ), which gives quite evident advantages

Image

in this case in spring it's broken so that they avoid eachother even too much, so end up getting silly curled paths even if there's no apparent obstacle, the hidden obstacle would be another unit's path.
heatmapping is only a mean to achieve right's result over left, there's other algorythms such as crowd continuum, but I will not discuss the differences unless requested
Attachments
heatmap.png
(3.26 KiB) Downloaded 2 times
User avatar
JohannesH
Posts: 1793
Joined: 07 Apr 2009, 12:43

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by JohannesH »

But the thing is, we already have customformations and the engine line formation and sometimes ctrl-click to make for that kind of maneuver when the player wants that. But when you want your units to clump as tightly as possible, wouldn't heatmapping interfere with that in many cases?

The thing is that units shouldn't be so "intelligent" to always presume something, like that keeping a loose formation is good, if it is only true most of the time. Most important thing is that you can predict how units will respond to your orders, that they act intuitive.

In that picture for example, imagine if there is an enemy turret that you're evading on the lower left, imagine another turret on the right of the designated path for good measure. You just click barely evading the turret range. If they keep this smart formation some will maneuver into the turret range very likely, but if they all scramble for the straight route to the waypoint they will probably form a bit of limbo line but will stay within the preferred path which is more important.
User avatar
BrainDamage
Lobby Developer
Posts: 1164
Joined: 25 Sep 2006, 13:56

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by BrainDamage »

JohannesH wrote:But the thing is, we already have customformations and the engine line formation and sometimes ctrl-click to make for that kind of maneuver when the player wants that. But when you want your units to clump as tightly as possible, wouldn't heatmapping interfere with that in many cases?
if it's done in a subtle way, as I said, not much/not really, also customformations, ctrl click and spring line formation themself queries the spring's pathfinder to manage the routes
JohannesH wrote:The thing is that units shouldn't be so "intelligent" to always presume something, like that keeping a loose formation is good, if it is only true most of the time. Most important thing is that you can predict how units will respond to your orders, that they act intuitive.
.
you could probably ( I did not check so don't take my word for it, perhaps it's even already done ) simply add a parameter to chose whenever to ignore or not the heatmap for your units, this way you could easily disable it when you need complete control over your units and leave it on for the rest of the cases
User avatar
Hobo Joe
Posts: 1001
Joined: 02 Jan 2008, 21:55

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Hobo Joe »

Yeah, I'm a little confused why that was implemented as well. I know it's obviously bugged in this release, but just the concept seems counter-productive to me.

For example, lets say there's a bunch of samsons, and you have a bunch of flashes. The obvious desired location for your flashes in right up in the face of the samsons, preventing them from retreating and nullifying their range advantage, so, in this situation, I draw a line on the backside of the samsons. My flashes move foward, and get right in the middle of samsons exactly where I want them. To my understanding, heatmapping would make the flashes attempt to go AROUND the samsons to get to their move order, rather than through like the order would obviously suggest. This is what happens in the new release, when the units aren't being retarded and going in circles or the completely wrong direction.

Since we have things like ctrl-click and line formations in spring, this functionality is totally pointless. The way I see it, units should attempt the most STRAIGHT possible route to their destination, regardless of what is in the middle. If I give a tight move order, I want them to move tightly, if I give an order in the middle of enemy units, I want them to go to the middle of the enemy units, not attempt to take the path of least resistance.

Units jumbling together is only a problem if you give a bunch of units a move order to the same spot, or a very undersized line, and speaking for myself, when I give a move order to a single spot, I want my units to jumble together, otherwise I just use lines.

Like Johan said, if there's two towers spread out, but there's a path of safety between them, I want my units to follow MY lines and stay outside of the range of both towers, because that's what I ordered them to do, not to try and spread out for more 'streamlined' movement.

If we didn't have line formations, I'd agree, but since we do, seems pointless to me. That's not to say I don't want improvements to the pathfinder, we all know Springs isn't perfect, but changes that take away control from me and try to give my units their own individual brains is a step in the wrong direction, imo.
User avatar
Carpenter
Posts: 216
Joined: 10 Jul 2009, 16:07

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Carpenter »

The biggest question for me is, why this spring version was released in the first place when the pathing is pretty much fucked up. :D
I mean when testing the new version, isn't it very obvious that this kind of a problem comes up very perceptibly? Just thoughts...
Kloot
Spring Developer
Posts: 1867
Joined: 08 Oct 2006, 16:58

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Kloot »

If we didn't have line formations, I'd agree, but since we do, seems pointless to me. That's not to say I don't want improvements to the pathfinder, we all know Springs isn't perfect, but changes that take away control from me and try to give my units their own individual brains is a step in the wrong direction, imo.
The thing is that units shouldn't be so "intelligent" to always presume something, like that keeping a loose formation is good, if it is only true most of the time. Most important thing is that you can predict how units will respond to your orders, that they act intuitive.
In AI, you actually can not make much progress in any direction if you never presume anything. And just for fun, try defining what "acting intuitive" means under a variety of battlefield conditions. I'd put some money on "avoiding unit collisions" being part of the definition all the time. In an RTS engine as much as in the real world, that implies leaving some distance between group members, which is the intended purpose of heatmapping.
In that picture for example, imagine if there is an enemy turret that you're evading on the lower left, imagine another turret on the right of the designated path for good measure. You just click barely evading the turret range. If they keep this smart formation some will maneuver into the turret range very likely, but if they all scramble for the straight route to the waypoint they will probably form a bit of limbo line but will stay within the preferred path which is more important.
What is more important *always* depends on context, and either way people will complain, as evidenced by:

"If you try to send units as group they seem to form a long line and get owned 1 by 1"
"A group of 100 lighttanks in evo told to go in a looong line"

(which happens to be exactly the result of "units should attempt the most STRAIGHT possible route to their destination" when a single destination is shared by many units)

So when any change is inevitably going to meet resistance, the ones that make most sense are those which improve the situation that is "true most of the time". If that's not what you're after (and prefer direct control to intelligence), this is possibly the wrong engine for you.
thelawenforcer
Posts: 106
Joined: 20 Nov 2008, 18:00

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by thelawenforcer »

Hmm, im sorry that the thread got derailed like this :s

i look forward to trying out the new pathing (when it works properly) as you guys seem to know what your talking about. Its a shame though that this new version is so obviously bugged though. I hope a fix is released very soon.

@hoijui, i wasnt aware that spring had public testing.
User avatar
very_bad_soldier
Posts: 1397
Joined: 20 Feb 2007, 01:10

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by very_bad_soldier »

I think the heatmapping stuff is a good concept (there may be exceptions where it is not appropriate), it just needs some tweaking (hopefully).
User avatar
JohannesH
Posts: 1793
Joined: 07 Apr 2009, 12:43

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by JohannesH »

Kloot wrote:In AI, you actually can not make much progress in any direction if you never presume anything. And just for fun, try defining what "acting intuitive" means under a variety of battlefield conditions. I'd put some money on "avoiding unit collisions" being part of the definition all the time. In an RTS engine as much as in the real world, that implies leaving some distance between group members, which is the intended purpose of heatmapping.
Avoiding unit collisions with heatmapping is not very intuitive, since you cannot always easily tell in which way they will maneuver to achieve that. Making a line formation is much more intuitive way to spread units when you know how it works.
So when any change is inevitably going to meet resistance, the ones that make most sense are those which improve the situation that is "true most of the time". If that's not what you're after (and prefer direct control to intelligence), this is possibly the wrong engine for you.
But in this case the result that is wanted "most of the time" is easily achievable by other means too. While the alternative, making units stick closely together, is already a pain in the ass to achieve oftentimes (tightly packed units will spread themselves out, when they stop moving already in old engine :/). If you add an extra switch, more complexity to provide that possibility, dunno if that will make units easier to control either when youve got extra switch to worry about instead of just sticking to lineformations.
Last sentence is just horrible, ask almost any competent RTS player and they will prefer crisp control of their units more... And bad players might not notice much difference either way


But then again it shouldn't be a problem when heatmap can be disabled, then units will work like they used to right? Or was there other changes as well?

Btw, would be nice to have a concept mod just to test pathing. A few different unit types with varying turnspeeds, speed, and footpring etc. Then have heatmapping and non heatmapping version of each unit, so good players could figure out better which ones behave nicer for different situations. When it works as it should first, ofc.
echoone
AI Developer
Posts: 150
Joined: 16 Nov 2009, 18:26

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by echoone »

Any ETA on when the heat mapping will be fixed or disabled?

I played yesterday. The game is borderline unplayable because the pathing is so badly broken. Even constructors seem to constantly get stuck on units and refuse to move. This effectively means queuing commands for constructors is pointless.

I was steamrolled by an AI because I could simply not keep up with construction, one click at a time. Constant moving back and forth takes too much attention away from front line units.

And its flat out enraging when you direly need troops on the front line and yet they will sometimes walk in any direction other than the direction they should be traveling. So on and so on...
User avatar
zwzsg
Kernel Panic Co-Developer
Posts: 7049
Joined: 16 Nov 2004, 13:08

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by zwzsg »

Hobo Joe wrote:samsons and flashes
I heard enemies units are not taken into account into your units heatmap.

echoone wrote:Any ETA on when the heat mapping will be fixed or disabled?
It can already be disabled mod-side.
User avatar
Thor
NOTA Developer
Posts: 291
Joined: 05 Mar 2006, 10:26

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Thor »

Collision spheres seem to have increased in size. It's more obvious for bigger units but seems to affect all units. Can we please get the old collision spheres back? This seriously affects balance for all mods.

old spring:

Image

new spring:

Image

old spring:

Image

new spring:

Image

There's also some other weird behavior. This hovertank is shooting way above the floating metal extractor. For some reason Spring seems to think the hitbox should be centered there, and moving it down doesn't change where the unit targets.

Image
User avatar
Hobo Joe
Posts: 1001
Joined: 02 Jan 2008, 21:55

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by Hobo Joe »

zwzsg wrote:
Hobo Joe wrote:samsons and flashes
I heard enemies units are not taken into account into your units heatmap.

echoone wrote:Any ETA on when the heat mapping will be fixed or disabled?
It can already be disabled mod-side.
Well I'm not sure what you heard or how accurate it is, but my units are most definitely trying to go around enemy units, rather than through them.
User avatar
CarRepairer
Cursed Zero-K Developer
Posts: 3359
Joined: 07 Nov 2007, 21:48

Re: Big Bug thread for 0.82.2/3 - Report bugs here

Post by CarRepairer »

Thor: jpg.
Post Reply

Return to “Engine”