Constant crashes

[Louvaine]

I don't know what causes it, but the game crashes on transition a lot more than it used to. It was expected to crash when transitioning from Balor to Greypeaks, but now it crashes between most locations on a casual loot run. During our standard route (Chaos in Cloakwood, Lich and Frost King in Cloudpeaks, Forest of Wyrms, Greypeaks), I get at least 3 crashes. More in bigger parties.

This is annoying, since spells are re-cast and some spells don't work with it (I noticed Haste often doesn't re-apply). Buffed summons most of the time don't appear and therefore lose their buffs (so do animal companions). Because of re-applying of buffs, blood magic deals damage all over. Not a big deal for Blood Magus, but Shaman outright dies. Oh, did I mention their 19th level Contingent Heal doesn't work? It's just gone after re-log (crash included). Skill Synergies seem to apply after equipment is removed and put back on, so with my UMD requiring synergy from Linguistics and Spellcraft, I log in without my armour and boots. Bonuses to Epic Spellcasting also don't apply properly, so epic 'spell slots' gained from Lore (Arcana) are also lost on my wizard. That henchman thing that doesn't work also constantly gets in the way (I can't command your summons any more, the feat just doesn't work for me at all). It should be hidden by default, same as the combat feats. Those are both quite niche, and I imagine most of the server population doesn't need them on by default.

All those details, however, are not what this post is about. It is about the crashing that is recently just taken for granted. All those small details just make it more annoying and frankly, after 5th crash I usually just call it a day.

[gedweyignasia]

It's very, very difficult to cause clientside crashes with the server. Most likely some files in your NWN2 installation are corrupt, there's a hardware issue with your computer, or a configuration problem with your client. If you contact Skywing, who created the client extension, he may be able to walk you through sending your crash dumps to him so he can identify the cause and add a fix to the client extension so nobody suffers from those crashes in the future. Alternatively, you can try completely removing NWN2 and all of its data including haks, overrides, etc, updating Windows, updating your graphics driver, and reinstalling NWN2, re-downloading the module, etc. Hopefully this helps to resolve your problem.

When there's a problem with maps or code on the server, generally a few things will happen:
- Scripts running on the server will stop running partway through. Players will experience this as spells, feats, or some other script failing to complete and appearing to do nothing. We'll see this in serverside logs and correct it if we can figure out what's crashing it.
- Scripts running on the server may hit an infinite loop. This will cause the script to stop running partway through, resulting in similar behavior to above, but may also create lag for all players by eating up all the server's CPU cycles.
- Resources which are improperly included or corrupted. This can result in missing or bizarre textures, crashes for all players every time they try to enter a map, or even the server failing to boot! This behavior will be very consistent, and should not appear random.
- SQL problems. These might result in things not being saved properly between resets, possibly including duplication or deletion of data. Fortunately, these are generally easy to see in the logs as well, so we can see the root of the issue!
- Problems in the core NWN2 server executable. This could result in anything, but should be consistent across all NWN2 servers. Most likely, the whole server crashes.
- Problems in an NWNX4 extension. Same as problems in the core NWN2 server executable, but specific to our server. This might show up when your character is saved, for instance, since we have a bespoke NWNX4 extension for that, courtesy of Dae.

When there's a problem with something running on the client, a few things can happen:
- Client is unable to join the server at all
- Corrupt or missing textures on models or terrain
- Certain maps don't load
- Random or intermittent crashes for one client only, with the rest of the server unaffected
- Lag for all players on the server

Hopefully this helps to solve your problem! Please share your troubleshooting steps and what eventually solves it so other players with this issue can address it for themselves as well!

[DaloLorn]

Some of this stuff does happen fairly consistently, AFAIK. It's a widely known thing that transitioning away from the Balor map will crash your game, for instance. There were a couple of others, IIRC, but I can't remember them offhand.

Endy also told me a story once about how spending enough time on the Corm Orp map (but no other map that I'm aware of) will cause the game to misbehave in increasingly bizarre ways, though I'm drawing a blank on the details. My point is, some of these issues are our fault.

[Louvaine]

I will look into troubleshooting on my end and post some results. I do have to say that I'm having hard time to believe it's my side when my whole party crashes.

[gedweyignasia]

When the whole party crashes it's probably not on your side, but the crash dump will have information that can help us figure out the problem, so taking that to Skywing will help us.

[gedweyignasia]

Some of this stuff does happen fairly consistently, AFAIK. It's a widely known thing that transitioning away from the Balor map will crash your game, for instance. There were a couple of others, IIRC, but I can't remember them offhand.

Endy also told me a story once about how spending enough time on the Corm Orp map (but no other map that I'm aware of) will cause the game to misbehave in increasingly bizarre ways, though I'm drawing a blank on the details. My point is, some of these issues are our fault.

These sound like some kind of corrupted resources, whether it's a bad walkmesh (probably not that simple), some object in the area that's messed up, or something less tangible. Hard to debug without completely redoing the area, but maybe worth looking for assets that aren't used elsewhere or transitioning to a new map (preferably with some DM support to explain the change and let players shape it).

[Louvaine]

When the whole party crashes it's probably not on your side, but the crash dump will have information that can help us figure out the problem, so taking that to Skywing will help us.

Could you point me to crush dumps or any other records that I could pass to Skywing?

[Rhifox]

As far as I can tell it's related to the number of buffs on characters all applying at the same time. The more people in the group, with the more buffs on them, the higher the chance that crashes will occur during a transition. Of course it can still happen when solo, but it's most common in groups. It's why I adjusted the DM teleport code awhile back to teleport players with small delays between each character.

I don't know if it's related to the buff vfx, the number of scripts being applied at once, or something else.

[Louvaine]

That makes sense. I can't do anything about it, though, I play a wizard.

[gedweyignasia]

As far as I can tell it's related to the number of buffs on characters all applying at the same time. The more people in the group, with the more buffs on them, the higher the chance that crashes will occur during a transition. Of course it can still happen when solo, but it's most common in groups. It's why I adjusted the DM teleport code awhile back to teleport players with small delays between each character.

I don't know if it's related to the buff vfx, the number of scripts being applied at once, or something else.

Does this occur with all transitions or only cross-server transitions?

[gedweyignasia]

When the whole party crashes it's probably not on your side, but the crash dump will have information that can help us figure out the problem, so taking that to Skywing will help us.

Could you point me to crush dumps or any other records that I could pass to Skywing?

Skywing will be able to walk you through the whole process. I cannot.

[Aspect of Sorrow]

As you can imagine there are a lot of points of failure where NWN2 is concerned, but you'll need the generated DMP from your %localappdata%\Temp\nwn2\dumps\memory.dmp inspected by something like WinDbg and stepped through with the "!analyze -v" command. While there aren't client side pdbs for us to reference by (OEI gave a few of us the ones used in nwserver, but not nwn2main for the same version) you can correlate events such as the 0x441310 as a SEF based overflow (Rhifox's description falls under one of these and spells like Stoneskin is known for creating), or HRESULTS such as 0x80070490 which is typically a binary corruption (either on disk or in memory) and exposed by Windows as it may have escaped the nwn2process and occured external of the stack (nt, kernel, etc) to which an example of an unhandled missing texture sent to the dx9 drawcall will appear from.

[gedweyignasia]

As you can imagine there are a lot of points of failure where NWN2 is concerned, but you'll need the generated DMP from your %localappdata%\Temp\nwn2\dumps\memory.dmp inspected by something like WinDbg and stepped through with the "!analyze -v" command. While there aren't client side pdbs for us to reference by (OEI gave a few of us the ones used in nwserver, but not nwn2main for the same version) you can correlate events such as the 0x441310 as a SEF based overflow (Rhifox's description falls under one of these and spells like Stoneskin is known for creating), or HRESULTS such as 0x80070490 which is typically a binary corruption (either on disk or in memory) and exposed by Windows as it may have escaped the nwn2process and occured external of the stack (nt, kernel, etc) to which an example of an unhandled missing texture sent to the dx9 drawcall will appear from.

Because I imagine this might have been entirely unhelpful, I'll try to translate:

If you open Explorer and paste "%localappdata%\Temp\nwn2\dumps" in, you should see a "memory.dmp" file, which is where Skywing's client extension stores the latest memory dump. You'll want to collect this immediately after a crash. The dump file keeps a copy of what was happening in the program's memory when it crashed.

Skywing, AoS, Dae, or someone else with debugging/reverse engineering knowledge can read through this dump file to try to figure out what caused the crash. When you create a program, you have information that makes it easy to go through these crashes to figure out what happened. Someone from Obsidian provided this information to Skywing, AoS, and Dae, maybe some others, but they only did so for the server. Even without those hints, it's possible to figure out (with a good deal of work) what's causing a crash. Depending on where the problem was, it might be possible to look at the messages programs/functions pass around to keep track of errors. (Not every error results in a crash; programs are ready to handle a lot of errors, if developers planned out how to deal with that error.) This won't give the full picture of why things crashed, but might help narrow it down.

A problem rendering the effect for Stoneskin, for example, might crash the game in two different ways; one of the ways might be a bug in the way the game handles the special effects leading to corruption in the game's memory, and the person debugging the dump would have to track down how that happened to fix it. Another might be that a texture to draw the Stoneskin effect on a character couldn't be found (for one reason or another), and when the game asked Windows to draw that texture, Windows sent NWN2 an error that NWN2 wasn't ready to deal with because the Obsidian devs didn't anticipate that texture might be missing.

Which is all to say, Skywing can walk you through making sure your crash dump will be useful, and Skywing is the person who has to handle this because no one else can write updates for the client extension to fix the problem.

[Aspect of Sorrow]

As you can imagine there are a lot of points of failure where NWN2 is concerned, but you'll need the generated DMP from your %localappdata%\Temp\nwn2\dumps\memory.dmp inspected by something like WinDbg and stepped through with the "!analyze -v" command. While there aren't client side pdbs for us to reference by (OEI gave a few of us the ones used in nwserver, but not nwn2main for the same version) you can correlate events such as the 0x441310 as a SEF based overflow (Rhifox's description falls under one of these and spells like Stoneskin is known for creating), or HRESULTS such as 0x80070490 which is typically a binary corruption (either on disk or in memory) and exposed by Windows as it may have escaped the nwn2process and occured external of the stack (nt, kernel, etc) to which an example of an unhandled missing texture sent to the dx9 drawcall will appear from.

Because I imagine this might have been entirely unhelpful, I'll try to translate:

If you open Explorer and paste "%localappdata%\Temp\nwn2\dumps" in, you should see a "memory.dmp" file, which is where Skywing's client extension stores the latest memory dump. You'll want to collect this immediately after a crash. The dump file keeps a copy of what was happening in the program's memory when it crashed.

Skywing, AoS, Dae, or someone else with debugging/reverse engineering knowledge can read through this dump file to try to figure out what caused the crash. When you create a program, you have information that makes it easy to go through these crashes to figure out what happened. Someone from Obsidian provided this information to Skywing, AoS, and Dae, maybe some others, but they only did so for the server. Even without those hints, it's possible to figure out (with a good deal of work) what's causing a crash. Depending on where the problem was, it might be possible to look at the messages programs/functions pass around to keep track of errors. (Not every error results in a crash; programs are ready to handle a lot of errors, if developers planned out how to deal with that error.) This won't give the full picture of why things crashed, but might help narrow it down.

A problem rendering the effect for Stoneskin, for example, might crash the game in two different ways; one of the ways might be a bug in the way the game handles the special effects leading to corruption in the game's memory, and the person debugging the dump would have to track down how that happened to fix it. Another might be that a texture to draw the Stoneskin effect on a character couldn't be found (for one reason or another), and when the game asked Windows to draw that texture, Windows sent NWN2 an error that NWN2 wasn't ready to deal with because the Obsidian devs didn't anticipate that texture might be missing.

Which is all to say, Skywing can walk you through making sure your crash dump will be useful, and Skywing is the person who has to handle this because no one else can write updates for the client extension to fix the problem.

Thank you Ged!

[Rhifox]

As far as I can tell it's related to the number of buffs on characters all applying at the same time. The more people in the group, with the more buffs on them, the higher the chance that crashes will occur during a transition. Of course it can still happen when solo, but it's most common in groups. It's why I adjusted the DM teleport code awhile back to teleport players with small delays between each character.

I don't know if it's related to the buff vfx, the number of scripts being applied at once, or something else.

Does this occur with all transitions or only cross-server transitions?

All. It's a vanilla bug I think, it's just particularly common during DM events (where it can cause crashes even without transitioning, simply on players casting spells) and adventuring zones due to the greater likelihood of multiple players and the higher use of spells in these areas.

I don't know the exact specifics of how it works, just that my experience with it over years on NWN2 is that it's related to number of spells/buffs being cast/active simultaneously, and that this does not seem to be a matter of hardware (as the issue still happens even as player hardware has improved over the years).

As an epic-level server, we are particularly vulnerable.