ID:2259995
 
BYOND Version:511
Operating System:Windows 7 Pro
Web Browser:
Applies to:Dream Daemon
Status: Open

Issue hasn't been assigned a status value.
Descriptive Problem Summary:

DD crashes often, but at seemingly random times. Had no luck figuring out what part of the code causes this.
Same happens on other recent versions.

Event Viewer log:

P1: dreamdaemon.exe
P2: 5.0.511.1384
P3: 591d134a
P4: byondcore.dll
P5: 5.0.511.1384
P6: 591d12d9
P7: c0000005
P8: 00152160
P9: 1a30
P10: 01d2e92861327167
511.1384 is no longer the current version. Does this happen in 1385, and if so what does that crash log say?

Mainly I'm asking because one particular bug in previous versions was likely fixed in 511.1385. I don't want to chase an old bug.
My bad. But the crash still happens in 1385.

P1: dreamdaemon.exe
P2: 5.0.511.1385
P3: 594200b3
P4: byondcore.dll
P5: 5.0.511.1385
P6: 5942004b
P7: c0000005
P8: 00154c66
P9:
P10:
It still looks as if there's a problem with null appearances. Very odd. I've improved the sanity checks for the next build.
Mysterious crashes still happen a lot in 512.1389. Should I make a new thread in Beta Bugs?

dreamdaemon.exe
5.0.512.1389
59d3f187
byondcore.dll
5.0.512.1389
59d3f117
c0000005
000ecd38
1ec8
01d33e2c2c4de66c
Hard to say. I'll trace into that and see what I can find. That appears to be a very different code location than before.
On review this does appear to be related. The crash you have this time is when trying to read turf.icon. The icon is being read from a unique cell that somehow has a null appearance, which shouldn't be possible.
Could it be related to turfs or turf icons with alpha less than 255, or empty icon_state?

Also, when CTRL is accidently pressed during drawing a turf in DM, it is drawn above the existing turf instead of replacing it, and only if their icon_states are different.
Not sure if it's intended, maybe that's the case.

Anyway, more sanity checks would be nice if they're possible.
In response to Repo82
Repo82 wrote:
Could it be related to turfs or turf icons with alpha less than 255, or empty icon_state?

No. It's likely something happening as a result of the number of unique cells passing the old 64K limit. I just don't see where the logic in that could be breaking down.
That's confusing.
Is there any way to get the exact line that causes the crash on Windows (I saw someone getting detailed logs on Ubuntu)?
The problem isn't where the crash is happening; the problem is that some data isn't what it's supposed to be, and wherever that data is set incorrectly is the real issue.