[kaffe] CVS kaffe (guilhem): Implemented stack overflow detection.

[Guilhem Lavaux]

On Thu, 2004-04-22 at 21:21, Stuart Ballard wrote:
> Kaffe CVS wrote:
> >         * configure.ac: Check for sigaltstack.
> 
> Note that this feature on Mono caused problems when a Mono binary built 
> on a 2.6 kernel (by my friendly neighborhood Debian packager) was 
> deployed on an older kernel (like mine). Mono added a 
> --enable-sigaltstack=no (IIRC) option that distributors are supposed to 
> use to cope with this.
> 

Hmmm, I have just tried to compile and run kaffe on 2.6.4 (modified
mandrake 9.2) and then run the binary on a 2.4.21 (mandrake 9.1). It
seems to work perfectly. I had problems with segmentation violation when
signals were still locked during a stack overflow but that has been
fixed using sigprocmask. How old was your kernel ?

Cheers,

Guilhem.


> (In Mono's case, the end result was that you got segfaults instead of 
> NullPointerExceptions, which is bad ;) )
> 
> Is this same thing going to happen to Kaffe, and if so, would a similar 
> solution be in order?
> 
> Thanks,
> Stuart.