[squeak-dev] Re: vm crash

Andreas Raab andreas.raab at gmx.de
Tue Nov 24 16:54:25 UTC 2009


Gonzalo Romano wrote:
>     I have a production system running on top of squeak and the vm 
> crashes after about 6 or 7 ours of testing, I need some help 
> understanding the attached crash.dmp because i don't really know what's 
> going on besides of what you can read on the stack dump, apparently 
> theres is a problem with a BalloonEngine drawing primitive, i was 
> wandering if somebody could provide me a hint on what could be failing...

Sure. From the exception code (C00000FD) we can learn that the crash is 
a C stack overflow normally caused by custom plugins or other external 
libraries having some sort of stack leak/corruption. From the loaded 
plugins we can see that you're using the FFI to do some stuff which 
leads me to suspect the problem is caused by one of your FFI calls 
(there is little reason to assume that one of the other plugins would 
cause this since they're used heavily in many production environments 
and have never shown such problems).

The first thing you should do is to very carefully screen your FFI 
calls. Are they declared correctly? Is the calling convention 
appropriate? Are the argument types correct? The other thing you could 
do to track this is to have a function which can return ESP from a 
primitive. In general the value should be identical when you call that 
primitive, so when it starts go differently, you're on to something.

Cheers,
   - Andreas




More information about the Squeak-dev mailing list