Podbot Beta 3.0 Updates? ^^

[KWo]

That is really weird, because I can see what I have added in code already - when the server is stopped - at ServerDeactivate the visibility table is saved if it wasn't before (if it has been recalculated). Wait for V3B21b - then we will see what does it say.
BTW - does Your OS allows to write to the podbot\wptdefault folder any new file? If it's a linux server - don't forget to allow it to write there.

About weapon_maxClip array - it was me who has created that table (it wasn't originally there). I have to test tmp again to see how many bullets it gives at buying.

[Serious Sam]

My OS is your standart win xp sp3 32bit and it has no problems writing pxp files there, it also writes pvi files, just not right away, but rather a few minutes after the analyzing has finished. That is, in the hlds, on a listenserver it only saves pvi when the map changes, if I click diconnect or quit notting is saved at all...

[KWo]

Update - V3B21b:

+FIXED max clip for tmp (now is 30 like it should be).
+Added code to prevent bots loose too much ammo while shooting at the wall (they should fire at the wall less now).
+Changed - when the bomb is planted, CTs shouldn't be so much scarried now (so they shouldn't overuse the task seekcover) - thus they are more hurry in defusing the bomb.
+Added more info to the log file about saving (or not) pxp and pvi files.

As usually, please, leave a feedback.

I have tested and - now I can confirm - the pxp file is saved at closing the listenserver. If it's not - it should write now the clear info about that to the log file. It should also write more clear now why the visibility table needs to be recalculated (I need the feedback why it is recalculated - if it was approved - now this should be clearly written).

[sacripan]

New update on my 4 servers and my listereserver.

[Serious Sam]

I was busy these past 2 months so I didn't have time for cs. I'll test the new version soon, in the meanwhile I want to add that the bots don't follow the waypoints very accurately, and it gets real obvious on cs_mansion. I made the waypoints myself and took a lot of hours to make them perfect, but the bots inside the house keep getting stuck at the walls, and I'm confident it's not because of my waypaths and wayzones.

[Serious Sam]

I did test them just now. From what I saw, they still did shoot at random directions when missing sight of the enemy, so I guess the fix didn't work. They also still keep not following the path correctly thus bumping into walls on cs_mansion, and maybe it was just me but I think they stoped buying pistols after the first round, which might be a new bug or maybe just coincidence.

[SamPlay]

hi,
as regards bots not following the path defined by your waypoints, there is nothing you can do but anticipating the inertia of the moving bot and adjust your waypoints accordingly.
This a flaw in the way all bots I know of use waypoints for trajectory control: they move straight towards the next waypoint WHATEVER THEIR CURRENT POSITION relative to the SEGMENT linking the 2 waypoints; that means that depending on when the bot switches from one waypoint to the next, it may "overshoot" ( most of the time) or "undershoot", especially with big angle differences and high speed.
The ideal position/time for switching is easy to compute and code thanks to how the Half-life engine computes moves, but the problem cannot be solved for all cases without resorting to speed control.
I have done it with a nav system that strives to nullify the distance to the segment: it works very well, but it is requires a bit more computations and having to use available waypoint graphs where you have no idea of the width of the "free" path between waypoints is a bit of a problem.
I hope this clarifies the issue.

[Serious Sam]

I see, but I have recently switched to pbmm from the old yapb 1.0 because they couldn't press many buttons on the maps and couldn't work with higher sys_ticrate than the dafault 100, and I haven't had this problem with them. Basically I use now the same waypoints as I did with them, with just a few changes to use the use_button flag now, and I see this big difference in them following the waypoints. So, if their navigation works fine, isn't it possible to implement it in pbmm

I'll add 2 more minor bugs I noticed on cs_mansion (btw I'm using the old version of the map where you have to open the doors with the use key) - when they try to enter the house, they stand in front of the door and repeat opening it and closing it until they sense an enemy, then enter. Maybe is because the doors take about a second to open and they think its still closed and press use again, I don't know. Also when they go to the ladder on the side of the house, they stay in front of it and look up as if they can't climb it, that's again until they sense an enemy, then they climb it. And I have confirmed that they're not buying pistols anymore after the first round, I see them always switching to usp and glock when they run out of ammo, I wonder if it was like that in the previous version...

[KWo]

Update - V3B21c:

+Added - bots are buying also the secondary weapon now (especially when they have a sniper or a shield).
+FIXED bots are hunting the teamnate when he was attacking him - now the TA revenge takes only 1.5 sec after TA (the bot gives some damage only instead fire until kill the team attacker).
+FIXED team specific waypoint are now really team specific (before it was fixed for offensive and deffensive goals, now bots should really take care about their team specific waypoints).
+FIXED planting and defusing the bomb problem (bots sometimes were stopping to plant / defuse the bomb when they have seen an enemy, but they didn't switch to better weapon to fire at the enemy).
+FIXED infinitely switching the weapon between the sniper and the pistol.
+FIXED chicken and zombie bots camping (now they will no longer camp - except they need to hide because of low health).
+FIXED invisibility not working for an invisible player using a knife.

Note:
I have checked the flashlight problem on a linux DS reported by addam here. I do confirm this bug and I can add even one more - bots are not using nightgoogles on a linux DS. But the bug exists is HL engine (pfnGetEntityIllum function seems to be not working the same way on a linux and on a windows DS), so I cannot fix it now. I may need a help from any coder able to test more the function pfnGetEntityIllum, to figure out what the problem is with it on a linux DS. I just don't have time to figure it out.

As usually, please, leave a feedback.

[mattisbogus]

Thank you so much KWo for your continued support on this project. I will start testing immediately. I will report back with some feedback within a week or so.

Best Wishes