Spring 89.0 - bounce over the cliff and other pushing

[PepeAmpere]

Some examples of the same kind of bug

1) Units can bounce each other to much, that they can be pushed on very steep places with no chance to get from there. (example image)

2) Sometimes "very interested" builder pursue too much his task, that he ends on place where cannot come from

3) Tanks moving via friendly lines, doesnt matter friendly bots in the way, they bounce them away (dont try to avoid the touch, dont try to avoid one small thing in their way, they only drive over it and try to push it away)

4) Tanks trying to get over obstacles by bouncing way... still pushing the obstacle like with cocain in their fuel.. until they are pushed/bounced away from obstacle, after that they continue in their way

...

all exaples from Spring 89.0 + NOTA 1.68 games

[Google_Frog]

I have an image about 1). These units moved from one side of the field full of holes to the other and over 1/4 of them ended up stuck in a hole.

But I haven't played many games with this pathfinding and there is a big difference between testing pathfinding and trying to use it in game. Units getting stuck may be rare in practice and be outweighed by better terrain handling.

[Kloot]

3) is a feature.

1a) the slopemap resolution is too low to prevent this near steep cliffs
1b) GF's image above shows discrete large jumps in the heightmap between two adjacent squares, yet within each hole terrain is still flat/passable --> units can easily be pushed over near-infinite slope edges without the transition being detected

[PepeAmpere]

3) is a feature.

I thought you say this is feature :) - so these guys ignore everything? theres no little penalization for way with obstacles, or with allied units? Just asking :)

1a) the slopemap resolution is too low to prevent this near steep cliffs
1b) GF's image above shows discrete large jumps in the heightmap between two adjacent squares, yet within each hole terrain is still flat/passable --> units can easily be pushed over near-infinite slope edges without the transition being detected

= "bad map" ?? the same happend to me on TheRockFinal, KappaBasin, TheRustyDelta... are these maps bad, too?

[Kloot]

so these guys ignore everything? theres no little penalization for way with obstacles, or with allied units?

Obstacles obviously do count as penalties, units (allied or enemy) do not.

If you do not want that last part, look into the avoidMobilesOnPath / speedModMults MoveDef keys described yet again in the changelog and the wiki.

= "bad map" ??

Read again, this time without assuming things I did not say.

[PepeAmpere]

Obstacles obviously do count as penalties, units (allied or enemy) do not.

Ok, thank you.

= "bad map" ??

Read again, this time without assuming things I did not say.

Im not assuming things you didnt say. It was hyperbolic question. I only feel you answer incomplete or not adequate to person, whos tagged "spring developer".

Maybe you answered why things happend this way, but you didnt say why it is happening this way in NEW version of engine, when previous version had not such (i can call it) buggy behavior. OR when or how you plan to change it, if you plan to change it.

I know short answeres are cool, but some bullshit around can help me expect what Spring guys plan. A bit. I dont want fix from you in few hours, i only want to know your intentions.

Please.

Thank you.

[varikonniemi]

Oh i have noticed the tendency from some devs that you have to ask the exactly right question to get anything remotely sensible as an answer.

Ie. you have to be precise and state it syntactiically correctly, like when talking to an AI. Maybe this is what is going on here after all.. the skynet iterating and improving over itself but having trouble deciphering the human emulation part

ps. short anwers are seriously not cool, if they have only barely touched over the subject at hand when writing a few lines more would have opened the book and pointed to the page you should take look at to understand what you are actually asking.

[hoijui]

what happend:
Kloot said the slopemap resolution is too low.
PepeAmpere assumed, the slopemap is something stored in the map archive, but it is actually soemthing calculated from the heightmap at runtime.
we can not just use a higher resolution slopemap, because it would be too slow or use too much ram or .... maybe it would need a much more refined algorithm and much mor this and that evyerwhere and not worth it. i don;t really know.. maybe only Kloot (and possibly jk and/or tobi do). klloot woudl have to explain a lot in much detail for us all to understand, and do the same with stuff touching the issue, and in the end explain too much of the engine. it would end up in a .. most likely useless conversation, which would make kloot angry.

it is not as simple as .. "just explain what we want to know!"

think of it like that...
try asking god for the meaning of life, and expect to get a simple to understand, practical explanation, just like when you ask yo daddy fo how to play baseball.

[Google_Frog]

To clear up 1b) those holes could be as little as 6 elmos deep and have the same effect.

[gajop]

To clear up 1b) those holes could be as little as 6 elmos deep and have the same effect.

still a near infinite slope if angles are used in that case

[SinbadEV]

I'm probably an idiot but... it could be cool for the cliff issue if you had a Lua gadget that kinda "shoved" them down the hill the rest of the way... clearly if the unit's slope tolerance or whatever is too low for the slope it is on for it to climb up the hill this shouldn't prevent it from falling down the hill.

[zwzsg]

That's typically the kind of solution that would create a dozen bugs in the attempt to solve one, SinbadEV.

[PepeAmpere]

what happend: ...

Thats not the point, not precisely. I hope hoijui will be one time as accurate in understanding of written text as he is in coding.

Its hard to be objective impartial judge commenting discussion when you have the same tag as kloot near your name hoijui, esp. when you start giving fathers advices.

-----------------------------------

Only thing i was expecting from any of spring developer:

yes, this can be bug, yes, this behavior is not wanted, i plan to fix it: ETA: 3 weeks.

OR

no, you dont understand it, thats feature, we dont plan to do anything around that in next few months

Thats all i need as answer... Such info is enough for content creator or maintainer of project to do some meaningful/relevant decision.

------------------------------------

That's typically the kind of solution that would create a dozen bugs in the attempt to solve one, SinbadEV.

yes :)

[FLOZi]

I'm probably an idiot but... it could be cool for the cliff issue if you had a Lua gadget that kinda "shoved" them down the hill the rest of the way... clearly if the unit's slope tolerance or whatever is too low for the slope it is on for it to climb up the hill this shouldn't prevent it from falling down the hill.

Engine used to do this, code was disabled

[hoijui]

@pepe
neither of the two things you want(ed) to happen make sense here.
it is neither a bug that will be fixed soon, nor is it a feature.
it is a side effect of how spring works, and that is pretty evident from kloots post; and the post contains enough info even.
it would need very hacky stuff to "fix" this. practise has shown, that ... exactly what sinbad suggested, and the reactions to it in this thread. such "fixes" create more problems then they solve, and for this one especially, it was even proven, as Z said.

i gave a coherent explanation for how things seem like, to me.
you said i was wrong, without giving any explanation where i was so, which is fail, and obvisouly does nto allow me to get better at such stuff, and reach perfection, like in coding.

i made very clear links between kloot and god. now... i guess you just don't know me well enough, but that was irony.
i can assure you, that i had more problems with kloot then you will probably every have, because of our shared "tag next to the name".

short: you got served well by kloot, you just do not want to hear the truth.
if you get an answer you don't like, or you don't believe, you may proove it wrong by doing the aledged impossible, like jamerlan did with his intel fixes recently.
you'd be a hero!

ouh and.. props for propper (minimal) quoting!

[Johannes]

So it's a bug that won't be fixed soon, if at all, why not say that clearly up front?
There was no way someone with little beforehand engine knowledge could deduce that from the initial reply, that just elaborated further what causes the issue - a course of action (in this case, doing nothing) can only be assumed from that.


Units getting in holes they shouldn't get to is nothing new, happened for quite a while already (if not always, not sure of that either way). Maybe it's more prevalent in newest version but hard to say about that since I haven't really played it, or even if I had.

But was the slopemap resolution always what it is now? For example on Cooper Hill, BA vehicles used to be able to smoothly pass through a gap on the south side of the hill, but a good while ago they stopped registering the passable spot and treated the whole south side of the hill as impassable. You can still fps single units through the gap if you know the spot.
Or was it another thing that affected this problem?

[knorke]

I'm probably an idiot but... it could be cool for the cliff issue if you had a Lua gadget that kinda "shoved" them down the hill the rest of the way...

http://www.youtube.com/watch?v=Q0pYfDydKPE \o/
There is even inertia so if you slide down one cliff, you can go up another!