[1.610] List of known issues

[jojorabbit]

How many times have we heard that a Cap can't be placed on the AIs ability to spam units.

Who said that? You can place a limit to any unit for AI and AI won't build more units then limit, only few(3-6) units are limited in most of AIs. That spam issue is something that is a little different and has problem with demand, long story short if AI sees he can have a good unit in 1 min he will always wait for it that is why AI spams -> so it makes bottleneck that has to be resolved.
I will fix it asap.

Since the new developer came on board a lot of AI myths have been busted.

Thanks am trying to explain what i know almost every day i learn a new stuff .

The problem dear JoJo is not in our stars but in ourselves. Built into the LUA programming language (of which SCAR is a subset) is the flexibility to choose what ever style of I/O one chooses. Boolean Neverbuild_FCN if return=false  is not the same as FCN return= -1, never build; is it?

DEMAND_NeverBuild is constant with value -9999 which does not mean that we can be 100% sure AI won't build that unit, if AI gets demand table with all units with -9999 demand he will probably build one of them depending on lowest ETA, random factor and other factors that i will explain in other post. I will create some diagram or story to resolve that AI myth about demand.

Problem is someone (probably relic) made that bottleneck cuz all call-in units have demand_offset around 500-1000 that is why they are always on top in demand table and there is no other unit with similar demand and standard deviation(variance) has no effect. IMO all demand_offsets should be at max 100. Also some of them should depend on map control and other more human-like factors.

[Pac-Fish]

Who said that? You can place a limit to any unit for AI and AI won't build more units then limit, only few(3-6) units are limited in most of AIs. That spam issue is something that is a little different and has problem with demand, long story short if AI sees he can have a good unit in 1 min he will always wait for it that is why AI spams -> so it makes bottleneck that has to be resolved.
I will fix it asap.

I think I love you for saying this

[Otto Halfhand]

DEMAND_NeverBuild is constant with value -9999 which does not mean that we can be 100% sure AI won't build that unit, if AI gets demand table with all units with -9999 demand he will probably build one of them depending on lowest ETA, random factor and other factors...Problem is someone (probably relic) made that bottleneck cuz all call-in units have demand_offset around 500-1000 that is why they are always on top in demand table and there is no other unit with similar demand and standard deviation(variance) has no effect. IMO all demand_offsets should be at max 100. Also some of them should depend on map control and other more human-like factors.

I think demand_offsets should be maximum = 100 - variance(in integer format) - Demand(in integer format) -that is being offset.

Edit:DEMAND_NeverBuild(-9999)
Never build (for the rest of the game) > FCN(-1) means never.
Can the infinitesimal chance(-9999) be over-ridden by the Boolean (-1)

[jojorabbit]

I think demand_offsets should be maximum = 100 - variance(in integer format) - Demand(in integer format) -that is being offset.

IIRC variance multiplies demand but am not sure need to check it out, will try to explain demand myth asap just to finish this crazy ai error that show on start vs soviet AI.

Edit:DEMAND_NeverBuild(-9999)

This is kinda function call.
It is defined like this : DEMAND_NeverBuild = -9999
So when you see in code return DEMAND_NeverBuild it is same as return -9999.

Never build (for the rest of the game) > FCN(-1) means never.
Can the infinitesimal chance(-9999) be over-ridden by the Boolean (-1)

Am not sure what do you mean with this, but Boolean can only be true or false nothing else -1 or similar.
Can you explain this a little more .

[Otto Halfhand]

Am not sure what do you mean with this, but Boolean can only be true or false nothing else -1 or similar.
Can you explain this a little more .

Some of the LUA variables have a value(null). value(-1) is similar usage. Since Lua allows you to define Functions pretty much however you wish to programmer A will define a FCN_Q one way Programmer B will define the identical function as FCN_R or maybe even FCN_Q, Programmer A returns Boolean(null) Programmer B is uncomfortable using nulls --me for one and defines null=-1. I have seen examples both ways in the limited time I've spent studying LUA. I think you agree it is very confusing.

[Blackbishop]

Boolean values can be true(represented by with numeric value 1) or false(represented by with numeric value 0), AFAIK it shouldn't be able to accept such value(null/-1). You can assign a string with null or an integer with -1. As booleans cannot be null, they are false by default, unlike strings or integers than start as null/-1... at least on my experience XD.

EDIT:
When dealing with booleans you don't use 0 or 1 but true or false, although you know what is their numeric value if you need to convert something .

[Otto Halfhand]

I understand what you are saying. When I started seeing -1 in the LUA scripting it puzzled me too. Apparently Lua accepts the normal Boolean true,false paradigm; the old C language 1,0 paradigm and accepts a condition where nul and -1 are predefined non-false conditions, THere also seems to be a boolean form with 0,1,2 values. Unfortunately all three forms appear in the ai programming for CoH/EF. weirder and weirder.

An example of where the -1 value would be used in EF programming involves unique units like IS3. only one is allowed per faction per game. In evaluating whether to build this unit the boolean true, false is used. Once the unit is produced the boolean value should return nul or -1 IE do not build for the rest of the game. Dollars to doughnuts the coding for unique units does not always conform to this rule though.

Edit: BTW Cornell U. was involved with early C language development; probably because IBM had several facilities within an hour's drive. Cornell's engineering school has a tradition that disallows engineering students, (comp sci at the time), form studying logic as a mathematical discipline. binary notation was used for booleans. I am Cornell trained. Not surprisingly I think in binary rather than logical notation.
 

[jojorabbit]

I think I love you for saying this   

LOL, Thanks.

Apparently Lua accepts the normal Boolean true,false paradigm; the old C language 1,0 paradigm and accepts a condition where nul and -1 are predefined non-false conditions, THere also seems to be a boolean form with 0,1,2 values. Unfortunately all three forms appear in the ai programming for CoH/EF. weirder and weirder.

AFAIK lua is written in C. Don`t mix 0,1,2 with boolean data type. From what i have learned i school there are 3 values for states, 0, 1 and undefined or high resistance aka "don't care" value marked X, see K-tables for logic functions in wikipedia if you are interested. But that is digital electronics.
When i started with programming for me it was easiest to compare true/false with flag or bulb, true -> flag is on, false flag is off .
What else is so weird? I did not saw any 0,1,2 boolean values in AI, where did you saw them?

Hope it helps.

[TheVolskinator]

Too much AI mumbo jumbo.

Brain hurts. 

Back to fiddling with corsix now that I got it to work.

[MBJrP36]

Do not know if it is a bug but every conscript squad I build cost 7 pop with a full 5 equalling 35 pop, I rarely use them as I love Propaganda Doc so I never noticed it before.
Here is the replay

[Blackbishop]

Conscripts generate their own population, so you only need the initial seven free pop.

[MBJrP36]

Conscripts generate their own population, so you only need the initial seven free pop.

Now I feel dumb, thanks bishop

[Blackbishop]

Don't worry, I'll try to fix that for the next patch, I mean, they will cost zero pop when building so you don't need to have 7 pop free but once they hit the field they will cost 7 pop but will generate their own pop, so it will basically cover this problem.

[Otto Halfhand]

once they hit the field they will cost 7 pop but will generate their own pop, so it will basically cover this problem.

A most clever workaround. You just have to remember as your conscripts take casualties your Pop Cap goes down. This can be confusing when your trying to eke out enough pop for "Tank Riders" or ISU152

[tintin2]

This was a post of the 1.52 patch thread.

i was just playin an AI match and noticed that when the AI hulls down their zis guns they dont seem to know how to undo it, They just sit there picking it up and setting it back down. this includes enemies who recrew a dug in gun.
« Last Edit: August 26, 2011, 01:00:54 AM by DrRockzo1986 »

It still occurs in todays patch.
Is it possible to modify something in the files to remove the dig in/out ability for the russians ATs?
Cheers