Home » MODDING HQ 1.13 » v1.13 Coding Talk » Experimental Project 7
Re: Experimental Project 7[message #352374 is a reply to message #352326] Mon, 12 February 2018 16:10 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Thanks for reply, appears i should not have tinkered with AI settings,
as i must have set uncover_in_combat=true myself ;)


Ok, next thing that made me think whether that works as intended:
The AI soldiers fire at tiles from where they were shot although its impossible they heard the sound,
and even if they would have seen the projetile, it was for less than 50% of its travel distance.
Ability appears to require that target survives the turn.

Long story:
snipers use silenced TRG22 and M25 EBR(all items @100% condition) with cold loaded ammo, = 21 tiles sound range.
Cheat+Alt+E shows distance of closest soldier(not the target) to closest sniper is 36 tiles,
no radio operator or soldier with radio in inventory among the enemies in sector.

If i kill the target(60 tiles away) in one turn, no matter whether 1_shot_kill or 2+ silenced weapons,
the enemy will not reply fire at all or even at the exact tile(s) from where i shot the victim(although 1 or 2 soldiers see the target die)
If the victim is still alive when i end the turn,
the enemies appear to "know" exactly the tile(s)from where my merc(s) fired.
Reliably reproduceable with r657&659 in AV1.10, save most likely still exists.



I get regular locks with r657&659 (prevous relases not tested) for specific path condiditions
locks do not occur with official 7609 exe, i.e. only the latter will skip the action after deadlock_delay time:

I completely blocked access to the inner compound in B13 by placing a line of blue flags behind the barriers,
and encountered many unresolveable halts when soldiers stopped directly at the barrier,
probably intending to jump over it, but realizing that there is no spot where they could land.

Stock 7609 exe will resolve the issue after deadlock_delay_time,
r657&659 left me no other option than ALT+X(waited 5+ minutes)

Now you may say i produced it myself with that exploit/cheat whatever and deserve nothing better,
respetively deserve no help to solve a selfmade issue, which i would be totally fine with
(btw i placed the blue flags in front of the barrier which resolved the issue :D
that is also the reason why i think its path related,
somewhat guided by the fact that more or less the same dev team also "produced" similar issues in Wiz8,
e.g. party has no path condition on arrival of statonary teleporter,
nearby pathnode behind physical wall allows bounding box to pass though, etc pp
;)
but a same or at least similar issue occurs with basic condition of map(s):

E.g. D15 map in AV 1.10:
The door in the NW wall blocks access to tile 11924(aka direction NW) for a soldier who opened it on tile 12088 and stepped on tile 12087,
and "he" appears uable to recogize that he has to close the door in order to proceed into the direction of my attackers.
This happened with stock 7609 exe and r657/r659,
and reliably locked r657&659(previous untested), while old 7609 exe manages to escape the loop(or whatever)

Save for rather reliable reproduction of game to become locked in B13 with r657+ exists,
save for similar/same issue in D15 was overwritten, sorry.
(On a hdd, so probably gone forever, but would check sectors if really important).


Pls let me know if you would want me to upld any of those saves.

[Updated on: Mon, 12 February 2018 16:11]

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352376 is a reply to message #352374] Mon, 12 February 2018 16:31 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
townltu wrote on Mon, 12 February 2018 19:10
snipers use silenced TRG22 and M25 EBR(all items @100% condition) with cold loaded ammo, = 21 tiles sound range.

You cannot calculate exact hearing distance from just gun's loudness, it's much more complicated, and first thing is that basic hearing distance is calculated as noise volume + soldier's hearing ability, which can be improved by skills or extended ear and can easily add 10-15 tiles.
AI will only shoot at heard enemy, and some random offset is added to exact noise location based on distance, so there's no cheating or bugs as far as I know.

Quote:
I get regular locks with r657&659 (prevous relases not tested) for specific path condiditions

It's well known problem, reported many times. There are some bugs in 1.13 path code which result in deadlock, for example if you place jumpable fence and then some impassable structure right behind it, the AI will think he can jump over, but when he tries to jump he realises he cannot, and hangs. It's even reproducible with player mercs on first (landing) wildfire map.

Quote:
Stock 7609 exe will resolve the issue after deadlock_delay_time

Yes, there is code in stock 1.13 to break the AI after certain time, unfortunately it breaks AI when in fast forward mode (because time limit passes too fast and it often breaks normal AI behavior), that's the reason that I disabled this code in Ja2+AI.
Quote:
r657&659 left me no other option than ALT+X(waited 5+ minutes)

You don't have to wait 5 minutes, just press ESC if you see endless clock.

Quote:
The door in the NW wall blocks access to tile 11924(aka direction NW) for a soldier who opened it on tile 12088 and stepped on tile 12087, and "he" appears uable to recogize that he has to close the door in order to proceed into the direction of my attackers.
This happened with stock 7609 exe and r657/r659, and reliably locked r657&659(previous untested),

I know of this door problem, but currently it is more a map maker's job to fix such issues.

Quote:
while old 7609 exe manages to escape the loop(or whatever)

Yes, and breaking AI in situation when fast forwarding enabled or player presses [-] on enemy turn.

Quote:
Pls let me know if you would want me to upld any of those saves.

Sorry, I don't plan currently to debug pathfinding code.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352386 is a reply to message #352374] Mon, 12 February 2018 19:11 Go to previous messageGo to next message
edmortimer is currently offline edmortimer

 
Messages:1533
Registered:January 2015
Location: Home Free
Quote:
E.g. D15 map in AV 1.10:
The door in the NW wall blocks access to tile 11924(aka direction NW) for a soldier who opened it on tile 12088 and stepped on tile 12087,
and "he" appears uable to recogize that he has to close the door in order to proceed into the direction of my attackers.


I fixed this map issue - it'll be in the next update at the end of the month.

Report message to a moderator

Sergeant Major
Re: Experimental Project 7[message #352395 is a reply to message #352386] Tue, 13 February 2018 15:37 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Fixed critical bug in r666, please update.


Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352546 is a reply to message #352395] Sat, 24 February 2018 21:26 Go to previous messageGo to next message
Taleman is currently offline Taleman

 
Messages:68
Registered:October 2010
Location: Finland
Runtime error, playing under Wine in Debian GNU/Linux 8, ja2_7609en+AI_r675.exe

Tried to upload a file, seems 0 bytes is max size for the attacment.

Now trying to insert an image I uploaded to web: http://taleman.fi/Kuvakaappaus-Jagged_Alliance_2-1.png

Report message to a moderator

Corporal
Re: Experimental Project 7[message #352548 is a reply to message #352546] Sat, 24 February 2018 21:53 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
I need a save just before the bug to try to reproduce and fix it. Posting a screenshot with some random bug doesn't help.


Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352549 is a reply to message #352548] Sat, 24 February 2018 21:58 Go to previous messageGo to next message
Taleman is currently offline Taleman

 
Messages:68
Registered:October 2010
Location: Finland
It did not happen again, when I started from end turn save.

Report message to a moderator

Corporal
Re: Experimental Project 7[message #352550 is a reply to message #352549] Sat, 24 February 2018 22:00 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
How often does this bug happen? I need something that can be reproduced to fix it.
You can also try to disable tactical panel and see if it helps.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352567 is a reply to message #352550] Mon, 26 February 2018 11:02 Go to previous messageGo to next message
Taleman is currently offline Taleman

 
Messages:68
Registered:October 2010
Location: Finland
Crash, with another error message. I believe this same error message about invalid UTF-8 character A did happen previously, but I did not take screenshot then. I have had these kinds of crashes about 5 times in past two weeks, before that I played a year without crashes.

This time screenshot, stack trace and end turn save file are available at http://taleman.fi/JA2/. Playing under wine in Debian GNU/Linux 8 Jessie and ja2_7609en+AI_r675.exe. Unfortunately the crash does not happen when continuing game from that save. Same thing has been every time with these types of crashes.

Report message to a moderator

Corporal
Re: Experimental Project 7[message #352568 is a reply to message #352567] Mon, 26 February 2018 11:12 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Sometimes this UTF-8 error happens when using SSA feature with localized game. Try disabling it if you use it.
What mod are you using? I cannot check the save if I don't know with what mod it was made.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352569 is a reply to message #352568] Mon, 26 February 2018 13:02 Go to previous messageGo to next message
Taleman is currently offline Taleman

 
Messages:68
Registered:October 2010
Location: Finland
I have no mods. I have only played the usual JA2 v 1.13.

What is SSA?

Report message to a moderator

Corporal
Re: Experimental Project 7[message #352570 is a reply to message #352569] Mon, 26 February 2018 13:12 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Taleman wrote on Mon, 26 February 2018 16:02
What is SSA?

Sector steady ambients feature:
Quote:
[Extended Options]
ENABLE_SSA = TRUE
DEBUG_SSA = FALSE
VOLUME_SSA = 25



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352582 is a reply to message #352570] Wed, 28 February 2018 20:31 Go to previous messageGo to next message
Taleman is currently offline Taleman

 
Messages:68
Registered:October 2010
Location: Finland
Different runtime error. I have now turned off SSA.

Maybe there is a typo in file name? I have a file named S_K_SWAT.sti, but the screenshot says it is missing S_R_SWAT.sti.
Quote:
taleman@yosemitesam:~$ locate -i swat.sti
/opt/ja2/ja2-datadir/Data/Anims/S_MERC/S_K_SWAT.sti
/opt/ja2/ja2-datadir/Data-UB/Tilesets/50/G-SWAT.STI
/opt/ja2/ja2-win/Data/Anims/S_MERC/S_K_SWAT.sti
taleman@yosemitesam:~$


Stacktrace, end turn save file and screenshot available from http://taleman.fi/JA2-2018-02-28/

No mods used. Playing under wine with wine ja2_7609en+AI_r675.exe. Debian GNU/Linux 8 Jessie. Unfortunately runtime error does not happen when continuing game from that Save00.sav. Which is kind of strance, why did it not need the file now? I tried to move MERC to the same position where the runtime error occurred, but no.

Report message to a moderator

Corporal
Re: Experimental Project 7[message #352583 is a reply to message #352582] Wed, 28 February 2018 21:09 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
S_R_SWAT.sti is in ANIMS.SLF
The problem is definitely not in normal code execution.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352587 is a reply to message #352583] Thu, 01 March 2018 12:56 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
I got a non resolveable and 100% reproduceable freeze with r675 & r666 in AV1.10,
happens as soon as i hit D(one).
It does not occur with stock 7609 exe.


Freeze also appears not to be related to changes done in .xml&.ini files
as it also happens with values set to the defaults of AV mod.
Same is valid if i outcomment all options for your exe which i could find in this thread
and which were added to \data-av\ja2Options.ini(i love to tinker;)


In case you want to take a look: sabercathost[dot]com/dxYn/SavedGames.rar

Since i saved every turn in this battle,
(game still runs extremely sluggy with AI exes, perhaps path calculation issues based on frequency of deadlocks,
e.g. 3/turn is not unusual, for battles with many enemies i watch videos on other notebook
because the enemy turns take sooo long, and because of that often miss info from enemy turns;)
the archive also contains the save from turn before, as it appears that already there the freeze in next turn is inevitable to prevent
by performing different (reasonable, didnt check a sarifice;) actions of PCs which would change the "seed" for next turn.

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352603 is a reply to message #352587] Sat, 03 March 2018 15:14 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
@townltu
Thank you for bug report and save, r679 should work better.

Quote:
game still runs extremely sluggy with AI exes, perhaps path calculation issues based on frequency of deadlocks,
e.g. 3/turn is not unusual

It's a problem of 1.13 path AI having problems with some maps, usually if there's a fence and someting directly behind it.
As for your save, the turns are very quick, with a whole turn less than half a minute maybe.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352605 is a reply to message #352603] Sat, 03 March 2018 16:06 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Thank you!

I salivate in anticipation to use new exe for imminent attack in Estoni ;)

Quote:
As for your save, the turns are very quick, with a whole turn less than half a minute maybe.

Well it differs widely here, sometimes i have turns that take > 5 minutes,
other times < 1min, estimated average is 2 - 3 minutes.

Also recently i got up to 24 "aborting ai deadlock for [n]"/turn within only seconds wait time for each message,
although i did not use [-] or [esc] keys to speed things up,
this may last for several turns or until i do something to draw attention.
If this is an alarming sign for you,
i should still have the saves where this occured in high numbers over many turns,
so in case let me know.


Keep in mind i am running the game on old hardware with T9400/9600 CPU,
using minimum energy consumption profiles and cpus locked @min multiplier
to prevent thermal stress on gpu from shared cooler design flaw which leads to flaws in bga connections.
(one machine already goes bsod after ~ 30min under full power:( )

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352607 is a reply to message #352605] Sat, 03 March 2018 16:59 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
townltu wrote on Sat, 03 March 2018 19:06
Also recently i got up to 24 "aborting ai deadlock for [n]"/turn within only seconds wait time for each message,
although i did not use [-] or [esc] keys to speed things up,
this may last for several turns or until i do something to draw attention.

The game will break AI for soldier if DEAD_LOCK_DELAY seconds of ingame time have passed (which may be much faster if fast enemy AI option is selected) and the player pressed ESC or 10 seconds of real time have passed (to avoid breaking the AI by highspeed game timer as in regular 1.13).
To fix deadlocks, someone should fix maps that cause them, or fix the path AI in source code (jump over fence part).

Quote:
Keep in mind i am running the game on old hardware with T9400/9600 CPU,
using minimum energy consumption profiles and cpus locked @min multiplier
to prevent thermal stress on gpu from shared cooler design flaw which leads to flaws in bga connections.
(one machine already goes bsod after ~ 30min under full power:( )

I have 3Gz core 2 duo and 4gb of memory, and can play game without any problems at decent speed. I don't play epic 64 vs 64 battles though.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352671 is a reply to message #352607] Sun, 11 March 2018 10:59 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Can you teach soldiers more effective use of ammo,
i.e. consolidate mags instead of dropping partially emptied ones?
Or let them pick up the mags they dropped and consolidate then if thats easier to code.


long story
The wild_reply_fire made me think whether i could use it against them
to let them run out of ammo with e.g. athlets repeatedly shoot&run/go_full cover etc,
which appeares to work as they stopped firing at some point and many charged then.
btw imo much deserved after the rain of lead, so dont give them more ammo;)

But the victory was diminished
as in sector inventory i discovered several mags which had a couple of bullets left,
probably as their ar/br/smg users had reloaded for suppression fire,
so if you can teach them to consolitate mags without too much effort, pls do it.


Btw, can you also teach soldiers to handle only_headshots_work/kill_permanently zombies in the "dying" state?
While they cant target body zone of prone victim, they keep fire at dying zombie until running out, then charge zombie.
Better step close&ignore until up, then pop head.

[Updated on: Sun, 11 March 2018 11:00]

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352674 is a reply to message #352671] Sun, 11 March 2018 15:21 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
townltu wrote on Sun, 11 March 2018 13:59
Can you teach soldiers more effective use of ammo,
i.e. consolidate mags instead of dropping partially emptied ones?
Or let them pick up the mags they dropped and consolidate then if thats easier to code.

long story
The wild_reply_fire made me think whether i could use it against them
to let them run out of ammo with e.g. athlets repeatedly shoot&run/go_full cover etc,
which appeares to work as they stopped firing at some point and many charged then.
btw imo much deserved after the rain of lead, so dont give them more ammo;)

But the victory was diminished
as in sector inventory i discovered several mags which had a couple of bullets left,
probably as their ar/br/smg users had reloaded for suppression fire,
so if you can teach them to consolitate mags without too much effort, pls do it.

AI soldiers try to reload before using suppression fire, that's the reason they have pertially filled mags. But they don't throw away those mags but keep them in inventory and use later. They would only need to merge them to be able to use those mags for suppression fire, but I doubt that the effect will worth the effort needed to code that.
Also, there is a code for AI to not use suppression fire when low on ammo, and even in black state (attacking seen enemy) they will try to use short bursts and then single shots. I don't know the exact situation in your battle but it should work this way for a long time already.

Quote:
Btw, can you also teach soldiers to handle only_headshots_work/kill_permanently zombies in the "dying" state?
While they cant target body zone of prone victim, they keep fire at dying zombie until running out, then charge zombie.
Better step close&ignore until up, then pop head.

In r684, AI soldiers will prefer to shoot at head if ZOMBIE_ONLY_HEADSHOTS_WORK option is set, also the hit to the other body parts will also have effect though much lowered (damage divided by 4) to avoid possible problems with AI not being able to kill lying zombie.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352676 is a reply to message #352674] Sun, 11 March 2018 17:12 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Thank you, sounds good if breath&dmg resistance in .ini still come into effect,
else i think nerf has too much impact if you want to have some tough zombies.
Besides i usually dont like to have zombies interfere in battle with army,
they tend to make battle easier unless there are many civilians to protect,
only forgot to deactivate zombie option and realized enemies issues to kill the dying zombie,


"dropping" refered to what dropped loot i found after the battle,
i didnt think they threw their mags away,
and it should be also "let them pick up the mags which other soldiers dropped when they died",
sorry for unprecise phrasing.

However i totally agree that if its an effort to teach them merging mags,
the effect of some more autofire from that is not worth it,
as they appear to have enough ammo for the usual battle.

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352698 is a reply to message #352676] Fri, 16 March 2018 14:30 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Some imo questionable stuff happening with forced Turn Mode,
keep in mind i am literally forced to use it
because after some secs in realtime the program appears to give enemies 99.99% resources
at least execution of my inputs for simple commands like [esc] to stop,[l]ook, etc are delayed up to 10sec,
respectively after merc death anim ;)
(it gets better if i allow CPU to run @2.8 ghz, but 3 sec delay is still deadly)


Enemies always get the 1st turn when they attack(=enter the sector),
even if my mercs see them through scope while still out of their sight range,
effectively making up to ~35+ tiles from all potential entry map borders an unsafe zone,
so in worst case leaving only ~70x70 tiles "safe" space to position mercs in center of vanilla size maps.
Its even more valid for disguised merc as suspect lvl skyrockets with alert,
causing ok disguise to drop and spy to die in 1st turn before we act at all.
Would prefer "attacker only gets first turn if no merc/soldier spots a member of other faction, else roll initiative"


Another one: mercs start with base ap in 1st turn, while enemies appearently come with bonus from non existing previous turn.
(confirmed with soldier tooltip detail lvl = debug, and if breath = energy they also do not lose correct amount of energy from traveling;)



Besides, had a weird forced turn mode related glitch which fortunately did not occur again:
Dropped team from Heli in Tonys sector to sell items and found troops as expected,
combat started while heli in air and ended instantly before any merc had touched the ground,
game played "battle lost music" but i missed to recognize the possible consequences,
fought the battle and found out that ~1200 items worth >250k had beeen stolen in ~2 sec by 20 soldiers :D
Was 100% reproduceable with the save before touchdown, independant of ingame option auto fast fwd ai turns.
All attempts to reprodue that glitch on similar occasion fortunately failed, so far.

[Updated on: Fri, 16 March 2018 14:32]

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352700 is a reply to message #352698] Fri, 16 March 2018 21:07 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
townltu wrote on Fri, 16 March 2018 17:30
Some imo questionable stuff happening with forced Turn Mode,
keep in mind i am literally forced to use it
because after some secs in realtime the program appears to give enemies 99.99% resources
at least execution of my inputs for simple commands like [esc] to stop,[l]ook, etc are delayed up to 10sec,
respectively after merc death anim ;)
(it gets better if i allow CPU to run @2.8 ghz, but 3 sec delay is still deadly)

Strange as it seems you are the only one with this problem, other players reported about lags in realtime sometimes, but the game is playable even on old pc.

Quote:
Enemies always get the 1st turn when they attack(=enter the sector),
even if my mercs see them through scope while still out of their sight range,
effectively making up to ~35+ tiles from all potential entry map borders an unsafe zone,
so in worst case leaving only ~70x70 tiles "safe" space to position mercs in center of vanilla size maps.
Its even more valid for disguised merc as suspect lvl skyrockets with alert,
causing ok disguise to drop and spy to die in 1st turn before we act at all.
Would prefer "attacker only gets first turn if no merc/soldier spots a member of other faction, else roll initiative"

It's intended to prevent cheating when player places mercs/militia close to edge of the map and kills most of the attackers with grenades on the first turn. So attacker always gets initiative. As a defender, you can build fortifications or hide in buildings.
Suspicion level depends on the number of enemies that see spy, and you should not keep spy close to opponents or friends known to opponent when in combat. Do you have suggestions how to change suspicion counter for better balance without making it overpowered? I think it should be mostly used in peaceful time or for recon in combat when there are not much enemies around (searching for last enemy etc)

Quote:
Another one: mercs start with base ap in 1st turn, while enemies appearently come with bonus from non existing previous turn.
(confirmed with soldier tooltip detail lvl = debug, and if breath = energy they also do not lose correct amount of energy from traveling;)

All soldiers should start battle with normal amount of AP, without carryover from last turn.
Enemy not losing BP from travelling is not correct, but I don't think it would be easy to fix as they don't exist before you load tactical map.

Quote:
Besides, had a weird forced turn mode related glitch which fortunately did not occur again:
Dropped team from Heli in Tonys sector to sell items and found troops as expected,
combat started while heli in air and ended instantly before any merc had touched the ground,
game played "battle lost music" but i missed to recognize the possible consequences,
fought the battle and found out that ~1200 items worth >250k had beeen stolen in ~2 sec by 20 soldiers big grin
Was 100% reproduceable with the save before touchdown, independant of ingame option auto fast fwd ai turns.
All attempts to reprodue that glitch on similar occasion fortunately failed, so far.

Item related bugs happen sometimes, I applied all possible fixes from the main trunk, don't think I can do more.
Stealing of items happens when enemy fully controls sector (so no mercs/militia in sector) and with a small amount every hour or so.

[Updated on: Fri, 16 March 2018 21:11]




Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352715 is a reply to message #352700] Tue, 20 March 2018 15:32 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
Well i only have old notebooks(with worn out keyboards as texts may indicate;),
encounterd to much lag related bad stuff in realtime on all,
so i go with turn based, perhaps switch to real time if there are only few enemies left.
I have some patience, and think playing with improved ai is certainly worth the higher time effort,
so no real issues with that here.


Placing mercs at border is not my style, let me explain:
previous battle ended and the sector instantly popped up with no time to crouch/prone mercs
in their standard defense positions at ~ 25+ tiles from border.
(as usual i spent time to get a grip on view conditions and determine the best defense spots)
Spy was at least 20 tiles away from other mercs when enemies entered and spotted unprepared mercs, opened fire etc,
enough to drop diguise and get poor spy killed.
Solved it by reloading save while in previous battle, move order for unprepared mercs&cancel to get away from the border,
nevertheless a little annoying.


Regarding suspicion counter:
In general works well for me after some time to get familiar with it,
still limited use(and i think it should be) for disguised spy on max alert in town/much cover sector
e.g. with extreme caution and much time effort, so balancing appears ok.

However i also got some weird behaviour of suspicion counter where it appeared to go erratic:
A) disguised merc in building near wall, enemy outside on adjacent tile,
during the following turns suspicion went up until disguise dropped,
only to automatically come up again next turn with pure yellow spyname.
B) similar for NW sam site of AR 1.10 but there the spy was moving prone in the southern building
next to&behind the outer wall with the windows facing SW, while enemies were outside the fenced area, ~15 tiles away.
I also switched to realtime and moved the merc repeatedly forth and back,
when passing behind the windows he became more and more suspicious and then redisguised a couple of times.


Soldier still standing at map border had 114 AP
when checking a save from 1st turn with alt+e in AR 1.10 @insane(= +5AP) and no deputy/squadleader in enemy team.


p.s.
20 AP for looking through scope appears too high since scoped rifle takes less,
no additional 20 AP for other direction(unless prone)
and last not least reveals entities in vision filed instantly instead of next turn,
which btw leads to paradox that spotter may get killed by "invisible" person in front of him.
Perhaps split spotter job and merely looking through scope for info about battlefield in 2 different things
and either have the latter take less AP for each look or allow to keep looking when we change direction.

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352720 is a reply to message #352715] Tue, 20 March 2018 21:38 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
townltu wrote on Tue, 20 March 2018 18:32
Placing mercs at border is not my style, let me explain:
previous battle ended and the sector instantly popped up with no time to crouch/prone mercs
in their standard defense positions at ~ 25+ tiles from border.
(as usual i spent time to get a grip on view conditions and determine the best defense spots)
Spy was at least 20 tiles away from other mercs when enemies entered and spotted unprepared mercs, opened fire etc,
enough to drop diguise and get poor spy killed.
Solved it by reloading save while in previous battle, move order for unprepared mercs&cancel to get away from the border,
nevertheless a little annoying.

So, what do you suggest? Giving first turn to defenders means great disadvantage for attacking soldiers and allows using exploits for easy wins. And attacker should have initiative in my opinion.

Quote:
However i also got some weird behaviour of suspicion counter where it appeared to go erratic:
A) disguised merc in building near wall, enemy outside on adjacent tile,
during the following turns suspicion went up until disguise dropped,
only to automatically come up again next turn with pure yellow spyname.
cool similar for NW sam site of AR 1.10 but there the spy was moving prone in the southern building
next to&behind the outer wall with the windows facing SW, while enemies were outside the fenced area, ~15 tiles away.
I also switched to realtime and moved the merc repeatedly forth and back,
when passing behind the windows he became more and more suspicious and then redisguised a couple of times.

At the end of each turn, a small amount of suspicion is added if some opponents have seen or heard spy this turn.
So even if you stay invisible but enemy heard you this turn, the suspicion will raise.
Also, suspicion raises if you killed some enemy this turn and there are soldiers who saw killed soldier and saw/heard spy this turn.
As for automatic redisguise, I probably need a save to understand what's happening.

Quote:
Soldier still standing at map border had 114 AP
when checking a save from 1st turn with alt+e in AR 1.10 @insane(= +5AP) and no deputy/squadleader in enemy team.

Don't know why it happens.

Quote:
20 AP for looking through scope appears too high since scoped rifle takes less,

Sounds like a mod balance problem if raising rifle with high power scope costs less than 20 AP.

Quote:
no additional 20 AP for other direction(unless prone)

Yes, but you can use binocs for spotting. I can probably allow spending APs when watching but not moving (reset watching status on entering new tile).

Quote:
and last not least reveals entities in vision filed instantly instead of next turn,
which btw leads to paradox that spotter may get killed by "invisible" person in front of him.

It never happened to me. Even if you get full sight bonus only at next turn, you don't lose your basic sight when you use binocs, so there should not be anybody 'invisible' in front of you. Sounds like already mentioned problem with extremely high cover settings in this mod. You will not meet invisible opponents when playing a balanced mod like SDO, for example.

Quote:
Perhaps split spotter job and merely looking through scope for info about battlefield in 2 different things
and either have the latter take less AP for each look or allow to keep looking when we change direction.

I would prefer to completely remove day vision bonus from both scopes and binocs as it will make playing on small maps much more interesting and balanced.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352724 is a reply to message #352720] Wed, 21 March 2018 00:52 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
I agree that attacker should get initiative,
works fine on big maps where the ground which attacker makes in the 1st turn
does not reduce the "safe" area for defender as much as on small ones,
especially for attacks from multiple sectors.
Perhaps you can reduce attackers AP in 1st turn for some amount like 20-30%,
still enough to allow killing instantly spotted mercs/militia at map border;)


Sound raising suspicion level explains it,
perhaps also why redisguise happened as enemy stopped hearing spy when disguise dropped, so counter was reset.
Anyway will check if a save from D2 is still present.



Well "in front of him" was quite exaggerated, sorry, merc looking through binocs was shot by soldier
in <30 tiles distance who was not invisible due to camo settings, which i adjusted to imo balanced results in game.
long story:
In standard dev releases you get instant vision cone of scope/binoc according to shift+v overlay,
but vision is only updated if we spend some ap, no matter how, e.g. change stance, use canteen, etc.
There is the exploit to go strategic map and back as it updates view, always works.
With AI exe the same is also possible and it shows that we dont see enemies on 1st look through scope
as soon as they are 1 tile out of sight rage for unarmed eye,
camo obviously cant be too high, else we wouldnt spot them at all even when forcing vision update with switch to other map.
Regarding mods low ready cost, take extreme 50 AP to raise the gun and add 7x6AP,
still a full 360 degree vision control in one turn is possible,
for ai scope that takes 8x20 + 7x6 AP, without the 8 turns for delayed vision updates.
If player does not shoot unseen enemies, spotters make less sense, better use a gun to watch and be ready to shoot,
but for shooting unseen targets player and all players in early game with no scopes it makes a difference.


I dont think vision bonus should be completely removed, but like in most mods things become op over time,
just another cool item that is a tiny bit better than the last one, which already was beyond balance.

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352785 is a reply to message #352724] Mon, 26 March 2018 18:53 Go to previous messageGo to next message
crackwise is currently offline crackwise

 
Messages:113
Registered:April 2013
Hi Sevenfm, thanks a lot again for the awesome AI mod! However I have a serious problem: I wanted to download the latest builds but my Antivirus software keeps detecting Trojans in the ja2_7609+AI.exe files which are located on your google drive.

Could it be possible that somehow those files on the google drive got infected? Alternatively my antivirus software (Microsoft Security Essentials) could be giving a false error but I think it functions correctly.

Thanks!

Report message to a moderator

Sergeant
Re: Experimental Project 7[message #352787 is a reply to message #352785] Mon, 26 March 2018 20:09 Go to previous messageGo to next message
townltu

 
Messages:384
Registered:December 2017
Location: here
crackwise wrote on Mon, 26 March 2018 18:53
...
However I have a serious problem: I wanted to download the latest builds but my Antivirus software keeps detecting Trojans in the ja2_7609+AI.exe files which are located on your google drive.

Could it be possible that somehow those files on the google drive got infected? Alternatively my antivirus software (Microsoft Security Essentials) could be giving a false error but I think it functions correctly.


According to all scanners on virustotal.com, ja2_7609en+AI_r685 is ok (link to scan result)
Check your file on virustotal, if infected your system may be compromised,
i.e. imo time for full system scan with multiple virus scanners.

Report message to a moderator

Master Sergeant
Re: Experimental Project 7[message #352790 is a reply to message #352787] Mon, 26 March 2018 23:28 Go to previous messageGo to next message
Flugente

 
Messages:3509
Registered:April 2009
Location: Germany
Could it be the Tools/ExShell/Exshell.exe? If yes, that is a tool to start the game with some graphic modifiers. It also causes any antivirus scanner to throw a fit, which is why it is best to delete it if you won't use it.


I know now that it could never work between us, as much as we wanted to, it could never be! Not because you're a rabbit, but because you're black.

If you want, you can donate to me. This will not affect how and what I code, and I will not code specific features in return. I will be thankful though.

Report message to a moderator

Captain

Re: Experimental Project 7[message #352798 is a reply to message #352785] Tue, 27 March 2018 18:00 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
crackwise wrote on Mon, 26 March 2018 20:53
Hi Sevenfm, thanks a lot again for the awesome AI mod! However I have a serious problem: I wanted to download the latest builds but my Antivirus software keeps detecting Trojans in the ja2_7609+AI.exe files which are located on your google drive.

Could it be possible that somehow those files on the google drive got infected? Alternatively my antivirus software (Microsoft Security Essentials) could be giving a false error but I think it functions correctly.

Thanks!

Some players already reported that, most likely the reason is paranoid antivirus software, as online scans do not find any viruses in exe. I have avg installed on my pc, it also doesn't show anything.

Quote:
Could it be the Tools/ExShell/Exshell.exe? If yes, that is a tool to start the game with some graphic modifiers. It also causes any antivirus scanner to throw a fit, which is why it is best to delete it if you won't use it.

There is no Exshell.exe tool on my google drive.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352800 is a reply to message #352798] Tue, 27 March 2018 20:12 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Ja2+AI r686

- NCTH: apply IRON_SIGHT_PERFORMANCE_BONUS at full day vision range
- NCTH: increase movement penalty
- breaking AI deadlock: show action/data/movement mode
- DetermineMovementMode: don't allow crawling near fences to avoid pathing problems
- show message when enemy raises red alert
- in Red/Black AI, if soldier is alerted (sees/hears enemy), always raise sector alert before doing anything
- Red AI: use sniper/suppression fire/mortar fire/skills only at the start of the turn
- use suppression fire more often
- lowered CTH check for sniper fire
- increased min bullets for suppression fire to 5
- allow using smoke to cover movement more often



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352828 is a reply to message #352800] Thu, 29 March 2018 15:45 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Ja2+AI r688

- update vision immediately after starting to watch/spot (no need to manually update vision by switching to strategic and back)
- switch off radio if we cannot use it anymore (take off radio, remove battery), when soldier spends AP/BP or at the start of the new turn (no more active radio operator with dropped radio)
- CheckForChangingOrders: change orders if soldier is not in town, not in underground sector and not in a room, also change orders if soldier is under fire (more active AI garrison when not defending town)



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352874 is a reply to message #352828] Sat, 31 March 2018 21:48 Go to previous messageGo to next message
szczurszymon is currently offline szczurszymon

 
Messages:6
Registered:January 2018
Hi Sevenfm!

According to WINDOWS DEFENDER Trojan:Win32/Azden.B!cl is located in ja2_7609en+AI_r688.exe. I've checked my computer with Malwarebytes and look up all the registry file and did not locate any trace of malware on my computer. Neveertheless Windows Defender does not let me use your newest file. There were no issues with earlier files. Is it a paranoid behavior of Windows Defender or your newest file is truly infected with this trojan? Please recheck r888 for this trojan particulary.

[Updated on: Sat, 31 March 2018 21:49]

Report message to a moderator

Private
Re: Experimental Project 7[message #352876 is a reply to message #352874] Sat, 31 March 2018 22:27 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Just checked the whole google drive and my pc and there are no viruses there.
You can also check any file with online antivirus scans.
And the file may become infected on the way from google drive to your hard drive.
Maybe a false alarm by windows defender, I don't know.



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #352881 is a reply to message #352876] Sun, 01 April 2018 12:29 Go to previous messageGo to next message
szczurszymon is currently offline szczurszymon

 
Messages:6
Registered:January 2018
Hi Sevenfm,

I've just reported to Microsoft false positive Windows Defender report made on your r688 file and other execs from your ai improvement (I found that issue has been reported by WD since r684). I've double checked all the files and my system (FRST, register, Malwarebytes, WD check, online scanners) and found out no issues on my System or files itself. I hope that Microsoft will fix the isssue and allow me and othe W10 users to use files properly. On my Linux instalation it works perfectly fine.
Sorry for inconvenience.

Report message to a moderator

Private
Re: Experimental Project 7[message #353085 is a reply to message #352881] Thu, 12 April 2018 19:26 Go to previous messageGo to next message
Jean Luc Picard is currently offline Jean Luc Picard

 
Messages:24
Registered:June 2014
Location: Germany
Hello Sevenfm,

because i've started a new Ja2 Run, i looked at the progress of this experimental project. Referencing to your Youtube - demonstration for the new ambient sounds, i looked at your other demonstrations too (destroying samsite as a spy with mouse for example). My Question is where i can get the new music - soundtrack that plays in background. It gives a much better fighting atmosphere. Is it part of your experimental project? Looked at kermi ftp but there they are all different.

Report message to a moderator

Private 1st Class
Re: Experimental Project 7[message #353086 is a reply to message #353085] Thu, 12 April 2018 19:47 Go to previous messageGo to next message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
Jean Luc Picard wrote on Thu, 12 April 2018 21:26
Hello Sevenfm,

because i've started a new Ja2 Run, i looked at the progress of this experimental project. Referencing to your Youtube - demonstration for the new ambient sounds, i looked at your other demonstrations too (destroying samsite as a spy with mouse for example). My Question is where i can get the new music - soundtrack that plays in background. It gives a much better fighting atmosphere. Is it part of your experimental project? Looked at kermi ftp but there they are all different.

Hi!
I use many music tracks in my demo videos, can't remember all of them.
Some of the music I use can be found here:
https://yadi.sk/d/Fj_B0WEKNjRQJ/Mods/Music



Left this community.

Report message to a moderator

Lieutenant

Re: Experimental Project 7[message #353089 is a reply to message #353086] Thu, 12 April 2018 20:31 Go to previous messageGo to next message
Jean Luc Picard is currently offline Jean Luc Picard

 
Messages:24
Registered:June 2014
Location: Germany
Thanks! There are the tracks i was looking for.

Report message to a moderator

Private 1st Class
Re: Experimental Project 7[message #353195 is a reply to message #353089] Tue, 17 April 2018 23:21 Go to previous messageGo to next message
Jean Luc Picard is currently offline Jean Luc Picard

 
Messages:24
Registered:June 2014
Location: Germany
While playing with this executable revision 688, i noticed suspicious lines in console when the enemy does his turn. This is not every turn. It happens irregularly. I made a screenshot. Maybe it is not a bug but suspicious, so i report it here:


https://www.dropbox.com/s/p2dyasdn9mje5e9/Unbenannt.jpg?dl=0

[Updated on: Tue, 17 April 2018 23:24]

Report message to a moderator

Private 1st Class
Re: Experimental Project 7[message #353196 is a reply to message #353195] Tue, 17 April 2018 23:28 Go to previous messageGo to previous message
Deleted.

 
Messages:2663
Registered:December 2012
Location: Russian Federation
It's a bug in AI that I need to fix. Nothing serious, but AI behaves less effective with it.
I wasn't sure if it still happens and how often so I decided to put a screen line.



Left this community.

Report message to a moderator

Lieutenant

Previous Topic: Trunk AI and improvements
Next Topic: New Attachment System Beta
Goto Forum:
  


Current Time: Thu Mar 28 10:24:41 GMT+2 2024

Total time taken to generate the page: 0.03766 seconds