Hi JCL,

could you please take a look at this thread:

http://www.opserver.de/ubb7/ubbthreads.php?ubb=showflat&Number=457609#Post457609

Maybe you could give me a hint where to take a deeper look at.

Also: Is there any chance to get a SED debugger that works with large projects? Currently the debugger allocates too much memory for a 32bit application so that the SED just crashes. Maybe you could get rid of the internally used xml structure?

Thanks.

Regards,
Sascha.

Crashes when loading a particular level B from a particular level A are normally either caused by a function that continues to run after level change, but accesses a now-removed object of the old level. Or it's an entity action of the new level.

Put a diag() call at the begin and after the wait() of all entity functions. This way you can see which function does not survive the first wait(). Similarly you can also find out which functions continue to run after level change.

Hi JCL,

thanks for your reply.

I've already removed all entity actions ... no change at all.

I've also logged out all currently running functions right before the level_load and removed them. The crash still occured.

Then I logged out all running functions right after the level_load and before the wait(1) ... the list doesn't differ from the one taken before the level_load.

After that I resaved all models used in both levels and removed the sky texture. Still no change.

Everything hints to a memory issue somewhere, but I wasn't able to find anything yet.

Is there any possibility to get more information about what the engine is currently processing?

Only what is printed in the log file. If it's not a function, next guess would be a damaged entity file - not necessarily a model. Maybe even the WMB file. Resaving is not guaranteed to fix it, but you could replace the entities with dummy models as a next step. When the crash is reproducible, you can find it by excluding.

The problem is that the crash disappears once I change anything in WED:

- I delete a model ... crash is gone
- I add a model ... crash is gone
- I replace a model ... crash is gone
- even if I delete a model and add it again ... crash is gone

, what do you mean with your 4th bullet/point; do you mean delete it in WED and create it in your script?

Have you tried to recreate the map through script (as in first load a new empty map than use ent_create etc. to put every used model/sprite/texture in it)?

As a last resort, have you tried just adding a dummy model (bullet/point 2) and does it give signs of memory/random errors?

Wait, if you delete an object and add it again the crash is gone? I would guess the wmb is't compiled right. Perhaps an unfortunate combination of models. Or an error in the saved wmp file wich got resolved, when you changed the level.

Why do you bite your teeth into this, if you can resolve the problem. I know you want to know what causes the error. But at this point I would delete and readd a model resave the Level and be happy with it.

Bullet point 4 seems to be related to changing the order of the models inside the level.
If I delete a model (not the last one listed in WED), then click on "undo" (the model appears now at the end of the list in WED) and then build the crash is gone. So the crash only appears if the models are in exact this order.

The reason why I'm not happy with this solution is that it just shifts the crash to another point of the game.

About 2 years ago I had exactly the same crash. The only difference was that the first level was a different one (but the second, the one that crashes, was the same). Then I just changed the order of the models inside the WED and was happy that it solved the problem. At least I thought so.
Now I noticed that the crash is still there, only the first level is different.

So editing the level just seems to move the crash, but doesn't fix it.

Did you try a full publish on the project? Does it crash in a different way or somewhere else? Maybe this gives a new idea where to further investigate.
Even if it does not crash - does the log of a published run show a proper engine shutdown or does it just stop logging (meaning it is not shutting down in a proper way)?

One of our beta testers noticed the same crash in the published version some months ago. It just took me some time to reproduce it.
The logfile of the beta tester looks the same as mine. No error message at all.

But I can give it another try.