There's an update in svn that could fix crashes. I had some logging code
that looks like
static unsigned long long useclog[LOGSIZE];
static int ulogidx = (unsigned int)-1;
# define logusecs(usecs) do { sqLowLevelMFence(); \
if (logClock) useclog[++ulogidx % LOGSIZE] =
(usecs); \
} while (0)
which blows up when ulogidx wraps around. It the sign of -ve % +ve is
undefined in C.
The fix is simple; make the index unsigned:
static unsigned long long useclog[LOGSIZE];
static unsigned int ulogidx = (unsigned int)-1;
# define logusecs(usecs) do { sqLowLevelMFence(); \
if (logClock) useclog[++ulogidx % LOGSIZE] =
(usecs); \
} while (0)
I doubt this is your problem since it should take many hours for the clock
log to wrap. But updating and rebuilding is still probably worthwhile.
HTH
Eliot
On Mon, Jul 26, 2010 at 4:32 AM, Lukas Renggli <[email protected]> wrote:
> >>> The image I got is based on the sources from 27.06.2010.
> >>
> >> You may be missing some image-side changes. The CogVM was open sourced
> on
> >> 20.06.2010.
> >
> > Actually, where should I get the complete image-side changes from? I have
> > those prepared by Lukas a while ago.
>
> Yeah, these are the complete ones from the repository adapted to Pharo.
>
> Maybe these same socket related crashes that randomly appear with
> Seaside images?
>
> nanosleep: Invalid argument
> Exited with exit code: 1
>
> Lukas
>
> --
> Lukas Renggli
> www.lukas-renggli.ch
>
> _______________________________________________
> Pharo-project mailing list
> [email protected]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>
_______________________________________________
Pharo-project mailing list
[email protected]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
