Posts from HoMM3 HD bug reports.

It all started with changing the behavior of flies in the Hive. Why was it even necessary to change this?

It all started with changing the behavior of flies in the Hive... Why was it even necessary to change that?

Actually, it's not quite like that.
It started with changing the mega-cheating quick combat to a more moderate, but still cheating version - like not using magic on a computer when attacking a human. This is especially noticeable with Conflux mages. And it seems that something has been tweaked in the AI, or am I imagining things - the computer behaves differently, it doesn't take nearby empty towns, but runs away as fast as it can to the far reaches :)

I now almost always agree to accept losses in the auto-battle against an AI mage, because if I don't, I often end up taking much more damage from ranged attacks. And in the auto-battle, he sometimes doesn't even use magic and loses.

I just haven't gotten around to it. My launcher doesn't start. More precisely, it crashes with a report when I try to launch it. The HD_Launcher.exe process is visible in the Windows Task Manager. I'm using all the latest versions. I launch Heroes by editing the sod.ini file. Only version 3.7 starts. Usually, the faulty module when launching the launcher for the first time is kernelbase32.dll, and in subsequent attempts, it's ole32.dll. There might be a problem with Softbed, but I'm not sure. Only the launcher doesn't start.

Regarding the two-hex creatures that approach a wall of fire, I once saw this happen and looked into the code to understand why. If the AI is attacking entire stacks or some powerful single units, this might be normal behavior for the original game. The AI can definitely see creatures right next to the wall of fire. But when it checks whether it wants to attack a stack, it has a check (if you're not playing on "peasant" difficulty): If the total value of the attack (taking into account damage from the wall of fire, but not considering that it damages before the attack is made) is less than 0, and if the attack will result in a bad "burn" for the AI or the enemy will retaliate and damage everyone around it significantly, then the AI will not attack. This is implemented as follows: when checking the "burn" and retaliation, the danger from these things is added to the danger of the hex that the creature wants to move to for the attack (the hex where the danger from the wall of fire is already stored). Then, the resulting danger of the hex is compared to the original danger, before these values were added. However, there is a bug here: if the creature is a 2-hex creature, then the hex that the creature wants to move to for the attack is the one of its 2 hexes that is more dangerous (the one that will be on the wall of fire). And the original danger is always taken from the rear hex of the creature – so if the "more dangerous" hex is the one where the creature's head will be, then the dangers of different hexes are being compared. Thus, if you place the wall right next to it, the 2-hex creature will see that the danger of the hex it wants to move to is high, but its original danger (which it took for its rear hex, which is not on the wall of fire) is low – and it will decide that the attack will result in a dangerous "burn" or retaliation against everyone around it, and it will not attack. At the same time, as I wrote above, not only the danger of the hexes is checked, but also the total value of the attack. If there is not a single unit behind the wall, but an entire stack (or maybe a powerful single unit), then the AI may decide that, even despite the damage from the wall of fire, it is more beneficial to attack the stack (especially since it does not know that the wall affects before the attack, not after). Then, the bug with the dangers of the hexes does not matter – and the AI will go to attack. However, if the AI is attacking single units like sprites, etc., then this is unlikely to be explained by the text above. --- Regarding magic in auto-combat, it is easy to verify that the AI can use it in principle – set up an AI hero in the editor with 1 Golden Dragon and Armageddon with 99 skill points, and attack it with a large army that will be decimated by Armageddon – in auto-combat, the AI will win, and this is only possible if it casts Armageddon. But I have also seen situations that are difficult to explain other than the fact that the AI is not casting spells in auto-combat. Perhaps it can only cast spells in some cases. Or auto-combat fights in a tricky way, forcing the AI to cast different spells or not cast spells at all.

In short, there is no definitive proof, but there is an opinion that a certain structure of neutral unit combat or a part of the Wall of Fire spell has been changed, because now two-hex neutral units periodically attack the Wall of Fire placed very close to them, and not all of them, but only a certain part... This can be seen in a stream; I first noticed this personally in a tournament about 4-5 months ago, and I thought it was a rare bug, but it turns out it's not... It started to repeat, and not just with neutral units, but also in Conquest mode, where nothing like this has ever happened before! And old-school players confirm that they don't remember anything like this.

To obtain clear evidence, the same battle needs to be conducted in three versions: without HD, with HD, and with HD+. After that, it will become clear whether this is due to HD or something else.

And Ivan firetosky claims that during a similar episode, a Bone Dragon went berserk against him! But there is no video evidence even for this; please investigate the situation.

Well, this is just ridiculous. Apparently, Georgy Petrovich put him under hypnosis. :smile16:

HD mod, latest version
The game crashed during an auto-battle with neutral creatures. I attacked the neutrals, and immediately after that, the game crashed.

Crash when passing from blue to red.

Regarding magic in auto-combat, it's easy to verify that the AI can use it in principle.

But the magic-bots have been using arrows for a long time now.

the total value of the attack (taking into account the damage from the wall of fire)

Sav, could you speak a little more simply, as if you were talking to accountants? The total value of the attack is how many units of X will the AI kill based on their Fight Value? And "damage from the wall of fire" is how many units from the attacking stack will the wall kill, based on their Fight Value?

Added 7 minutes later

when checking for burn-through and retaliation, the danger of these things is added to the danger of the hex for everyone around

Danger is how much Fight Value will be lost for the AI?
[off]These are probably elementary things, but I don't want to rely on guesswork in an unfamiliar area.[/off]

Value and danger can be considered relative concepts. They are actually calculated based on how much Fight Value is killed or risks being killed by a certain amount of damage, but many factors are taken into account, and describing how these values are calculated would be difficult.
The danger of a hex includes a wall of fire/moat; the presence of other creatures nearby in case they can be used as a shield against a burn, fireball, or area-of-effect attack by stepping onto the hex; and enemy stacks that a creature exposes itself to by stepping onto the hex.
The value of an attack includes the attack itself, the retaliation, the removal of retaliation, the burn, and the area-of-effect attack in both directions, the blocking of archers, and the danger of the hex that the creature steps onto.

To obtain clear evidence, the same battle needs to be conducted in three variations: without HD, with HD, and with HD+. After that, it will become clear whether this is due to HD or something else.

The fact that the AI attacks a wall of fire placed close to its own units with two-hex units has been known for some time and has been demonstrated repeatedly, including on my stream. This occurred in the following situations: 1) If there are not just single units behind the wall, but many units, some of which will die after the attack, especially if there are ranged units there. 2) If the single unit is not placed in the center of the right edge of the field, but at the top or bottom. In this case, part of the two-hex units attack, and part do not, in this way. It is difficult to say what this is related to, but I have noticed that this is true when one of the parts of the wall (top or bottom) is adjacent to the edge of the field or another obstacle.

After the update, when launching the game, this message appears: http://cs633128.vk.me/v633128690/13aba/__XP5gtnzTA.jpg
but it seems like there are no errors.

Most likely, something went wrong somewhere, but I still wanted to ask: I played against an opponent with a protected save file. The next day, after loading the save, the game didn't ask for a password. However, after the game, the opponent couldn't figure out my password (he checked several times), and I was also able to load the game without confirmation. The text file, strangely enough, wasn't updated on the day of loading, although before that, I played against another opponent, and he couldn't remember his password for a long time. The text file wasn't updated in either case. What could be the reason? I'll upload the save file and password just in case.

Feel free to use it for now:cool:
You can dismiss the last unit in the hero window by using Alt+Shift.

baratorch, I have a problem with the player Alin. According to him, the game crashes whenever he attacks any monster. We can't play a tournament game; please help. He sent you his save file and the crash log in a private message.