Jump to content

suma

BI Developer
  • Content Count

    3202
  • Joined

  • Last visited

  • Medals

  • Medals

Posts posted by suma


  1. Otherwise I'd say it's time for 1.60. Then we can get on with MP interpolation.

    Yes, this is exactly the plan now. I think we are all setup, as the server side fix already done should be enough to allow further MP interpolation to be done client side only.

    Now all we need is to verify there are no significant regressions, which might take a few days, and then the new stable patch could be released.


  2. http://www.arma2.com/beta-patch.php

    [84057] Fixed: Possible crash after calling joinSilent and deleteVehicle.

    [83810] Fixed: Rare problem of AI fleeing when all enemies mounted vehicles.

    Note: this time there is a server-side fix, which should partially reduce unit lagging in MP (sometimes the same position was broadcasted twice erroneously). The fix by itself is not very significant, but it should reduce the artifacts caused by the interpolation.


  3. Amazing how I've been playing this game since 2009 and just now I find out about this.

    On my Gateway p7805u FX gaming laptop p8400 2.26ghz dual core/ 9800m GTS (at stock speeds):

    before using -exThreads=3 -cpucount=2

    fps on bechmark 1 = 25

    after using -exThreads=3 -cpucount=2

    fps on bechmark 1 = 32

    Thanks for sharing your experience. Still, what you write sound a bit strange, as -exThreads=3 -cpucount=2 should be the default for dual cores since build 76122. I would like to understand more about what is happening here. Are the results repeatable? What is the exact CPU model you have?


  4. This makes sense.

    Just a slightly OT question: within danger.fsm, the CANFIRE event is still triggered when a AI unit is fleeing, out of ammo or ordered to hold fire (combatMode "blue")? I would like to know, because in these cases it should probably not be triggered ( it would "stop" intermittently the unit without any real reasons). Asking before checking it, thanks ;)

    It should be triggered only once the unit can fire - i.e. has ammo and is allowed to fire.


  5. AI with heavy AT seems to have difficulties to reveal + fire at targets with clear LOS:

    Simple test in editor (desert map) - just place BLUFOR + OPFOR groups with move waypoint.

    AI with Javelin/Metis don't fire at tanks even if those vehicles are driving 500m in front of them.

    AI with sniper rifles suffer from the same problem. Note - the AI are standing and have clear LOS on targets and reporting "Tanks 500m".

    This is probably location dependent, I was unable to reproduce so far. Please, always attach a repro, even if it looks trivial, it makes sure we are not wasting time here.

    ---------- Post added at 13:06 ---------- Previous post was at 12:56 ----------

    AI in Tunguska dont use missiles against air targets.

    Simple test in editor (Takistan) - just place 1x Tunguska and 1x Blufor plane (player) with move waypoint.

    Blufor (player) plane can even circle above the Tunguska and nothing happen unless the pilot attacks the Tunguska.

    I can confirm only partially - the result was dependent on Tunguska skill. With low skill no missile was fired on me, but there was a significant MG fire (I was flying Av-8). With high-skill Tunguska I was shot down by a missile after a while. However, even with high-skill it took quite long before the missile was launched.


  6. AI without ammo has a very high chance to flee from my experience

    Exactly. Weaponless or ammoless AI fleeing whenever it encounters an enemy, unless fleeing is completely disabled for it by allowFleeing 0.

    ---------- Post added at 08:17 ---------- Previous post was at 07:37 ----------

    However, this does not explain what we see here, as units do have ammo here in the repros.

    ---------- Post added at 09:26 ---------- Previous post was at 08:17 ----------

    This time the repro revealed a rare corner case, caused by the way the repro is constructed. The repro starts with a soldier out of the car, which is later moved into the car by the trigger. This has confused the target assignment function and it started fleeing as there was supposed to be an enemy target, but it was enable to engage any enemy target. I have fixed this in 83810. You should not see this problem if you move "man moveInGunner uaz1" from the trigger into the init line.

    As for why they are not engaging the car, it is because they do not recognize it as a dangerous enemy. He has no ammo and he is not firing at them. If you add an ammo to him (just a small amount will do), they will engage. There is one more problem with M2s: the view is obstructed by a grass on a small elevation in between, but the grass in not rendered at that distance. This is even more visible because the UAZ gunner is very high, the AI is aiming at the car body, which is almost hidden by the grass. This is something I will check (I will try to consider some point in the upper half of the target for the purposes of grass visibility testing).

    Edit: it shows the car body is hidden behind the terrain and even before the grass check was introduced the M2 would not fire on the car.


  7. I have seen it just yesterday, with 83759 - you can see it with 83780 now if you want. :) I have tested a Chernarusion general store (a repro provided on CIT) and in addition I have tested AI shoothing through UAZ windows at the unit standing behind the UAZ. If you have some other interesting cases which work or do not work for you, let us know here.


  8. AI tanks seem to spot other AI tanks really poorly now. Sometimes they even drive next to each other without reaction. I saw M1A1 blindly standing still on takistan sand land (with sparse grass), T34 noticing it finally few meters away. I was also laying still on takistan sand land. 3 enemy tanks moved only under 100m away, but they didnt spot me until I started to crawl. Also then the very nearby T34 didnt spot me until after about 20 seconds.

    The grass visibility was incorrectly done on a line to a bottom of the target, instead of to its aiming center. Fixed in 83769, tanks seem to fight normally now.

    ---------- Post added at 22:26 ---------- Previous post was at 22:25 ----------

    Try playin with the USMC character first, and see how the AI is reluctant to fire, even when standing up and the player is standing as well. When i tried with the latest official patches the AI rips me to shreds.

    This seems to have the same cause as the non-fighting tanks, and it should be fixed in the same build.


  9. Regarding the delay it seems when my character puts his weapon to the ground to start healing the weapon falls through the ground texture and keeps falling at a distinct speed while healing so it really takes my character a while to retrieve his weapon from far down under after finishing healing :j:

    Plz try n fix asap :bounce3:

    This time I would really like to have a repro (preferrably on the CIT), as it is not obvious to me how to get into such situation.


  10. Great repro. I have run it, I confirm the issue, and I have no explanation so far - both collision avoidance and collision response (which I would assume as first candidates to cause the problem) are never hit. Stay tuned ...

    Fixed now. There is one more collision testing specific for cars, which is forward-reverse sensor (something like a parking sensor) which triggers forward-reverse way of tight turning. This sensor was detecting the cranes. I have fixed it low-level in the collision detector, which at the same time fixed the issue of 3rd person camera failing because of cranes (and perhaps even more similar issues, if there were any).

    Stay tuned for 83597


  11. Suma, here is an old AttachTo example mission which shows the issue F2K Sel is talking about

    http://norrin.org/downloads/ArmA2/attachTo/heliDoor.utes.rar

    Enter the chopper, tell your pilot to go somewhere (this shows the issue better), select 'move to the right (or left) door' action. Watch it lag

    I think this is the issue fixed in http://dev-heaven.net/issues/22863#note-13, as the situation is the same - camera on a soldier which is attached to the vehicle (I assume you should be able to verify within a few days).


  12. Here:

    http://www.mediafire.com/?cc4229oycu9h4c9

    You're a passive passenger of a truck. The truck will drive very badly while it has boxes attached to its back, but when you delete the boxes, it will drive normally (although still toppling over a couple of objects at the first intersection).

    Great repro. I have run it, I confirm the issue, and I have no explanation so far - both collision avoidance and collision response (which I would assume as first candidates to cause the problem) are never hit. Stay tuned ...


  13. I see problems with cargoed AI groups unable to completely disembark from vehicles (both helicopters and ground vehicles ).

    Some units won't disembark at random, so both the transported group and the vehicle get stuck endlessly.

    If you provide a repro, chance is it might be fixed very quickly. Without a repro is can take very, very long.


  14. I'm not sure it's a bug, but probably due to the scan horizon changes.

    I quite like it, but it would be nice to be able to control this option.

    Changing the behaviour has some effect.

    Yes, it is a bug, and it was due to scan horizon changes. Fixed in 83559 (no turrets scan any more).

    As for some automatic scanning by commanders or even gunners, I agree it would be nice to see, but it was not my intention this time, and if it is to be implemented, it should be implemented properly.

×