[karajorma]

Fook, I can't get at the retail missions as I don't have the CD handy.

The HOTU Crack Pack on my FAQ's Downloads Page has a cracked 1.2 exe which should solve that issue.

[General Battuta]

Fook, I can't get at the retail missions as I don't have the CD handy.

The HOTU Crack Pack on my FAQ's Downloads Page has a cracked 1.2 exe which should solve that issue.

Thanks.

Difference confirmed on 'Surrender Belisarius', retail vs. build 5862 no MVPs. Friendly fighters fire Rockeyes on retail, but do not fire them on 5862. On both versions, enemy Hercs will fire some Rockeyes, but much more frequently on retail.

Something broke retail AI.

[Dilmah G]

[General Battuta]

Mantised. Hope I did it right.

[Sushi]

Fook, I can't get at the retail missions as I don't have the CD handy.

The HOTU Crack Pack on my FAQ's Downloads Page has a cracked 1.2 exe which should solve that issue.

Thanks.

Difference confirmed on 'Surrender Belisarius', retail vs. build 5862 no MVPs. Friendly fighters fire Rockeyes on retail, but do not fire them on 5862. On both versions, enemy Hercs will fire some Rockeyes, but much more frequently on retail.

Something broke retail AI.

Unfortunately, there are quite a few places where FSO breaks retail AI. QuantumDelta probably knows the most about the differences. I've investigated a few and have fixed them where I could find/reproduce them.

I'll look into this one ASAP. One important question (posted on Mantis as well) is when exactly this buggy behavior appeared.

[Reprobator]

I use to mantis it already six month ago and i used to find from which build exactly the bug appear.

It seems like everyone is playing on insane around here  

[chief1983]

Not me, I can't even beat the game on Hard.  Medium is pretty intense.  Course it could be the way the AI have behaved in recent years.

[General Battuta]

I only play on Insane on retail, and frankly I've been spending most of my time with Fury's AI recently (as have many others), so that might be why this problem has gone unnoticed for so long. Thank goodness people picked up on it.

[Sushi]

I use to mantis it already six month ago and i used to find from which build exactly the bug appear.

It seems like everyone is playing on insane around here  

You're right, you did. It's even assigned to me. 

[Reprobator]

It has been assigned automatically 

[Sushi]

Found the problem: it was a subtle type-casting error introduced when I was doing some AI updates.

The code before:

Code: [Select]

firing_range = MIN((swip->max_speed * swip->lifetime * (Game_skill_level + 1 + aip->ai_class/2)/NUM_SKILL_LEVELS), swip->weapon_range);

The (broken) code after:

Code: [Select]

float secondary_range_mult = (aip->ai_secondary_range_mult == FLT_MIN)
? (Game_skill_level + 1 + (3 * aip->ai_class/(Num_ai_classes - 1)))/NUM_SKILL_LEVELS
: aip->ai_secondary_range_mult;

firing_range = MIN((swip->max_speed * swip->lifetime * secondary_range_mult), swip->weapon_range);

The fixed code:

Code: [Select]

float secondary_range_mult = (aip->ai_secondary_range_mult == FLT_MIN)
? (float)(Game_skill_level + 1 + (3 * aip->ai_class/(Num_ai_classes - 1)))/NUM_SKILL_LEVELS
: aip->ai_secondary_range_mult;

firing_range = MIN((swip->max_speed * swip->lifetime * secondary_range_mult), swip->weapon_range);

Basically, the problem is that the expression involving game_skill_level (and other integer variables) was previously being implicitly cast to float before division because it was multiplied with another float (swip->lifetime). When I moved that expression, that implicit cast from int to float was lost.

This is fixed in trunk rev. 6077, along with a couple of other AI bugs.

[Galemp]

Uh... good! *hands Sushi a Kelp-flavored Cookie*