35 replies to this topic

[Leomar]

I want to let you know, that we have found an older bug, which is related to vanilla BG1, but with BGT the problem still exist.

The problem depends on quests with killing all enemies:

As example: Also with Weinrc, and the zombies down at the farm. I have played through several times, each time I kill 20 zombies and go back, and he keeps telling me I need to kill more. Is there a 21st one hidden somewhere, or am I trying to do this one too early, or is some Global value not setting.

It is due to the fact that opponents do not count as killed if one they "destroy" in the most true sense of the word. This means if you hit the opponent so violently that he leaves no corpse on the ground in the death, he is not evaluated as killed. There is apparently for these quests a counter which checks whether everybody was killed for the execution of the quests necessary opponent. Only if the opponent leaves no corpse, he is not counted. The fact that the opponent leaves no corpse, it happens always then when you hit too violently.

Normally we thought that this was a problem with the German version of BG, because we have not the Gore option. At first it was said, we must install the Gore option to solve this problem. But we have find out, that this is not the solution. It was checked again and we thought this will not happen in a BGT megamod. But now the bug-reports tells us it otherwise.

10thLich in-game solution is as follows at the moment:
There is a "Fix", one kills the suitable opponents simply with a little weaker party member, best of all with distant fighters. After I kill the opponents with my archers, everything has functioned perfectly.

Here one of the bug-reports.

Greetings Leomar

[Ascension64]

There is a fix for this. It has been implemented for Reevor's rats and Drizzt's gnolls.
Is the zombie farm the only other instance of the kill count quests that haven't been fixed?

[10th]

No, there's still that 'kill the Xvarts and save the cow in distress' quest in AR3300.

10th

[Hoppy]

I have not had the issue of counted kills on my BGT games with any of the aforementioned quests even with more tweaks like permanent corpses from BG1UB if that makes a difference. Even if I do destroy the zombies, gnolls or the like. It is a very strange difference that has to be accounted for somewhere.

I use the original CD formats if that matters.

Edited by Hoppy, 28 October 2008 - 09:06 PM.

[Leomar]

There is a fix for this. It has been implemented for Reevor's rats and Drizzt's gnolls.
Is the zombie farm the only other instance of the kill count quests that haven't been fixed?

Great, so we have a solution to the problem. How it was fixed? Is it for each quest differently? The same problem happens with a quest in SoBH which should be fixed, too. I'll crosslink with the SoBH forum.

Greetings Leomar

Edited by Leomar, 28 October 2008 - 10:54 PM.

[Ascension64]

Essentially, they no longer rely on a script that increment a global when the creatures die. Rather, they rely on the SPRITE_IS_DEAD<DV> incrementation, and checks it with NumDeadGT() used by BG1 Fixpack. Probably still not completely infallible, but works better so far.

[Leomar]

Thanks for your information.
Will you fix the 3 additional quests for the next BGT release?

Greetings Leomar

[Ascension64]

Thanks for your information.
Will you fix the 3 additional quests for the next BGT release?

Greetings Leomar

Yes I will.

[Leomar]

Thank you very much.


Greetings Leomar

[Autar]

The Paladin Bjornin in Beregost is not recognizing that I have killed the ogres he ask for. It may be another problem related to this issue.

Edited by Autar, 01 December 2008 - 09:44 AM.

[melkor_morgoth75]

The Paladin Bjornin in Beregost is not recognizing that I have killed the ogres he ask for. It may be another problem related to this issue.

He is ... what Ogre did u defeat?

mm75

[Autar]

That ones in BGT AR3300 (original BG1 AR4400). But I defeat them before meeting the Paladin thefirst time (if it matters).

[Leomar]

Right, the Paladin Bjornin quest is the next which should be fixed. I've heard it before and forgot to mentioned it.

And here happens another death counter bug. This time it is Silke.

There could be more and have to do with BG1 quests and their needed death counter.

Greetings Leomar

Edited by Leomar, 09 December 2008 - 11:10 PM.

[Taimon]

I'm still wondering why this is happening. The Die() trigger seems to work reliable in BG2, if no other script is blocking it. (Independent of chunking.)

[Ascension64]

From what I remember, people most had difficulties when killing multiple enemies at the same time (e.g. fireball), so it may be an issue with either too many Die()'s triggering at once, or too many IncrementGlobal()'s triggering at the same time. The latter may make more sense since the way to code an IncrementGlobal would be to:

1. Get current value
2. Add 1
3. Set the value

If a second IncrementGlobal() intervened between steps 1 and 3, then the same value would get set to x + 1, when it should be x + 2. This assumes that there is no lock on access to the global.

[Taimon]

Sounds logical, but unfortunately I've not been able to reproduce this.
And in Silke's case this even is a straight SetGlobal, so no potential race condition here.

I guess I'll file that under unsolved mysteries of the IE.

[Ascension64]

I've fixed all the ones here now, but still querying the problem with Silke and provided the method for whoever maintains SoBH in the relevant thread.

However, the following quests also use what we think is the flawed method of IncrementGlobal and Die(). I'm not sure if I should bother fixing them if they aren't broken, so I need to people to report if the following quests for some reason don't get finished properly:

1. Ardrouine and killing the worgs
2. Aldeth Sashenstar and killing druids
3. Sarhedra and killing ogres
4. Aldeth and Brandilar and discovering infiltrates of the Merchant's League
5. Scar and investigating the Seven Suns

Edited by Ascension64, 15 December 2008 - 03:29 PM.

[mouser]

From what I remember, people most had difficulties when killing multiple enemies at the same time (e.g. fireball), so it may be an issue with either too many Die()'s triggering at once, or too many IncrementGlobal()'s triggering at the same time. The latter may make more sense since the way to code an IncrementGlobal would be to:

1. Get current value
2. Add 1
3. Set the value

If a second IncrementGlobal() intervened between steps 1 and 3, then the same value would get set to x + 1, when it should be x + 2. This assumes that there is no lock on access to the global.

I can imagine all sorts of bugs like this coming out now. Multi-core processors weren't even a dream when the Infinity Engine was being developed (certainly not for PC's at any rate). Windows '95 looked like it was multi-tasking, but the processor was still capable of doing only one task at a single time - it just switched back and forth between tasks really fast. Now, you've got true multi-threading processors that can wreak havoc on code that didn't account for them.

If you think that's causing a bug, go to task manager, and right click BGMain under processes and set the affinity so it runs on only one core. If that fixes it, you've found your culprit.

Edit: and don't assume that since something appears to be a simple set value, that there can't be a threading issue. Consider:

float a=4.4,b=6.3,c=2.54

{
a=a*2.66
b=b/3.0
c=c+23.0
}

Seems simple enough. But all the compiler guarantees is that the three variables will have one set of values at the start of the block, and the changed values at the end. It does not guarantee in what order those lines will occur. In fact, what will probably happen is the compiler will go and change the low bits for all three variables, then go back and change the high bits for all three. So you have a few processing cycles where the actual value stored in those variables is completely wrong. If something else reads that variable at that time...

Edited by mouser, 15 December 2008 - 04:17 PM.

[erebusant]

There is a fix for this. It has been implemented for Reevor's rats and Drizzt's gnolls.
Is the zombie farm the only other instance of the kill count quests that haven't been fixed?

Great, so we have a solution to the problem. How it was fixed? Is it for each quest differently? The same problem happens with a quest in SoBH which should be fixed, too. I'll crosslink with the SoBH forum.

Greetings Leomar

The Bonehill Rat quest is not a kill count issue. It's an issue with a proximity trigger not firing properly.

Edited by erebusant, 15 December 2008 - 09:46 PM.

[erebusant]

I've fixed all the ones here now, but still querying the problem with Silke and provided the method for whoever maintains SoBH in the relevant thread.

However, the following quests also use what we think is the flawed method of IncrementGlobal and Die(). I'm not sure if I should bother fixing them if they aren't broken, so I need to people to report if the following quests for some reason don't get finished properly:

1. Ardrouine and killing the worgs
2. Aldeth Sashenstar and killing druids
3. Sarhedra and killing ogres
4. Aldeth and Brandilar and discovering infiltrates of the Merchant's League
5. Scar and investigating the Seven Suns

In my current install I did not get the correct count with the Xfarts at Hulrik's (took 3 tries to count correctly), with Bjornin's Half-Ogres in AR3100, and with Sarhedra's Ogres.