[Duke] Sector objects on E2L2 are just a wee bit pooched

[Rachael]

To reproduce, simply type "levelwarp 2 2" in the console. The problem manifests itself immediately.

0.7.3_alpha-245-g977223287

[mjr4077au]

DN3D: LSD Edition 1969... Must have come in with all the array access changes Graf's been working on. Will try to backtrace

[Graf Zahl]

Don't worry, we'll find those things. It was to be expected that there's some problems.

BTW, I started with similar work on Blood. But that code is in a different league. Instead of spreading actors across two arrays, Blood has seven! This is one of the main reasons why the actor code looks so messy - it needs to play around wil lots of pointers to get things done, and to top it off, some of the added data is optional, meaning the code is also full of rather needless checks for the presence of these, too...
The positive is that it contains less spaghetti code - it's not free of that and the spaghetti that's present is even worse than Duke, but the amount is limited to a handful of functions, not the dominant style throughout.

[Graf Zahl]

This one's weird. From the looks of it it starts with this commit: 4c3435bacf69363f870b6d61c8dacc36fbed07f8, but I cannot find out what precisely. The results are totally inconsistent. If a build crashes it is reproducable, but some builds from the same revision work, others do not.

[mjr4077au]

Optimisation issue with release builds vs. debug builds?

[Graf Zahl]

No. It seems to be random. I was unable to find any immediate cause. I suspect some uninitialized data. But that's a very small commit and nothing in there seems to be related to the sector effector in question.

[mjr4077au]

Was this commit meant to fix the issue? I can't reproduce anymore but am conscientious you said you found it to be a random occurence.

[Graf Zahl]

Yes, that commit was supposed to fix it. There was an undefined case there with bad owner setup - it never handled the 'no owner' case properly.

[mjr4077au]

Legendary