So yesterday our shakedown run of 1900 mod Skyrim ran into a snag when we started to encounter mysterious crashes in the vicinity of Riften. Initially we weren't getting crash logs at all, but after very carefully approaching the site we were able to get some crash logs suggesting an animation or AI-related crash that seemed to be linked with combat events. At the end of yesterday's session, I regenerated the LOD as well as the animation register, and this morning I tried loading into the area in question... and I crashed without a log.

 

Now one additional thing I had done was to purge the grass cache, and when I went looking for the crash log I also noticed that it included an empty grass folder.

 

 

Now the grass caching is done by No Grass In Objects, which is a Net Script Framework plugin. So first things first I checked the main NSF operating log and as you can see below there don't seem to be any problems at that level.

 

 

Still, something seems a little funny here. As a test, I next went into xEdit and created a patch that reverted all the landscape data for this region to vanilla. This will probably look strange in the area where the guild house gets added, but it'll help us rule out the possibility that there's something wrong with the landscape or the grass in this area. 

 

This test didn't solve the problem, though that's just as well - I hate fixing other peoples' landscape issues because it stovepipes you into using the CK in a way that it's not well designed for. The CK is great for authoring or adjusting content (though in many cases xEdit is better) but it's very poorly suited for comparing two different mods side by side and integrating their changes. I did observe more clues about what seems to be going on here. First and most importantly I was getting severe script lag. Functions that should be completing before I entered line of sight were happening in front of me (Devious Striker substitutions for example) and other functions like Dynamic Animation Replacer were not executing in a timely manner. You can see several T-Posed actors (including the player character) below.

 

 

So what conclusion are we supposed to draw from this? Well, we can probably rule out the landscape from being the problem and we can mostly rule out the grass caching function. If it was involved at all, it was just in the context of some sort of script pileup causing issues. 

 

Now you might recall that some of the crashes were happening in the context of combat, and I thought it was interesting how in the Papyrus log, the last thing to show up was a weapon effect firing on the player. This doesn't mean that Devious Strike was the specific issue, but it does seem like combat has something to do with the issue. On the other hand, it's probably not the sole cause since we have plenty of cases where (the "no log" crashes) where the game is closing out during a teleport or on game load. 

 

 

So did we get a crash log this time? We did! Here's what we're seeing:

 

 

That's the NPC who shot me with an arrow right before the game crashed. Not visible here, but we're still crashing at AB387F. That's the same address as the last two crashes so we probably have a single common cause of these failures.

 

 

SkyBirds is still present in the log.

 

 

Similar probably call stack including a bunch of havok calls. All three logs have cell refs in the stack, though not always the same one. These are probably red herrings or at least are secondary to the underlying problem.

 

 

What I wonder now is if the association with the Rift hunting lodge is actually a red herring in and of itself and if the underlying issue will trigger instead in other areas where combat is happening. Let's turn on god mode and take a stroll through a different area.

 

 

Stepping out of the Resting Pilgrim, things are superficially running fine. SkyBirds spawns are functioning right and while there are the usual stutters and UI scramble that comes with a game during the startup process, that's mostly a consequence of not sitting in the start cell while they all sort themselves out and they should self-correct.

 

But then it crashed again. And yes, you aren't reading it wrong, it crashed while executing a Devious Strike script. In many ways, this is a different crash than what we've been getting out by Riften, but it speaks to a broader stability issue and - I dare say - it smells like script issues are still involved. 

 

 

A look at the Papyrus log is showing errors in Devious Strike scripts

 

 

...and even more errors involving that mod.

 

 

Now I know from previous testing that this mod does in fact work, but it seems like it is having some major issues here. I also note that this mod has been updated since I installed it and it appears to have major changes. When we continue the investigation, we'll update Devious Strike and try to drill further into the script lag issue we've been encountering. I have some suspicions about other mods that might be causing more harm than good and that might be getting the old heave-ho. Specifically, I'm looking at...

 

 

and

 

 

Both mods are great at face value, but if they don't work and instead gum up the game... they'll have to go.