I have tracked the crash to a bug in J602abeta in
x{"r y where x is machine-word integers 0s and
0=(($#y)-r){y . It erroneously returns a
corrupt result when it should signal index error.
Under certain conditions the corrupt result can
cause a crash. e.g.
|: (2-2){0 2 3$a:
The bug was previously reported and fixed in
http://www.jsoftware.com/pipermail/beta/2007-September/002386.html
and will not be in the next beta.
----- Original Message -----
From: Roger Hui <[EMAIL PROTECTED]>
Date: Wednesday, October 31, 2007 20:36
Subject: Re: [Jbeta] Crash in beta
To: Beta forum <[email protected]>
> It does crash in the J602a beta but does not crash in my latest DLL.
>
>
>
> ----- Original Message -----
> From: Henry Rich <[EMAIL PROTECTED]>
> Date: Wednesday, October 31, 2007 18:31
> Subject: [Jbeta] Crash in beta
> To: 'Beta forum' <[email protected]>
>
> > 9!:14 ''
> > j602/beta/2007-09-19/23:00
> >
> > The following sequence crashes Jbeta when executed with -jprofile:
> >
> >
> > uy_vx =: 2 : 'v~u'
> > keyskld =: {"1"_
> > getkl_defu_colsv =: 2 : '(<@{.n)&keyskld
> > uy_vx(i.!.0)((<@{:n)&[EMAIL PROTECTED] ::(u"_)"0 _)]'
> > getkl1d =: getkl_defu_colsv(0;1)
> > getklu1d =: 1 : '[:>u getkl1d'
> > (<'xxx') (<'') getklu1d 0 2$a:
> >
> >
> > I can't cut it down, I'm afraid. The failure shows up
> > differentlydepending on what sentences have been executed
> before
> > this sequence.
> > If I combine all the verbs into one sentence, I get strange
> > results: sometimes it works, other times I get varying results
> > when I execute the sentence repeatedly. It looks like
> something> is accessing memory incorrectly.
> >
> > If there's any way to get a dll with a fix for this I'd be obliged.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
