GSWeb crash upon startup [Was: Re: Eeeagh! I'm at my wits end.]

[David Ayers]

Simon Stapleton wrote:
> On 27 Apr 2004, at 03:34, Adam Fedor wrote:
>
> > > From: Simon Stapleton [mailto:simon@tufty.co.uk]
> > >
> > > > Uncaught exception NSInternalInconsistencyException, reason:
> > > > GNUSTEP Internal Error:
> > > > The private GNUstep function to establish the argv and environment
> > > > variables was not called.
> > > > Please report the error to bug-gnustep@some.bogus.org
> >
> > Is there anything printed before this, like an Assert or just a Log 
> > message with the same info?
>
>
> Nope.  It comes in (only if I make with debug=yes) all by itself.  If I 
> make without debug, it's a simple sigsegv.
>
> > > problem again.  Now, it's obviously not a problem with
> > > GNUstep per se,
> > > as other tools seem to work OK - eogenerator, for example.
> > >
> > > I guess It's something to do with my makefiles and the frameworks I
> > > build all the components as, but it's got me stumped.  Anyone
> > > have any
> > > ideas on what (other than a problem at the gnustep-base
> > > level) might be
> > > calling this?
> > >
> > > +[NSProcessInfo load] is definitely being called before
> > > main(), and is
> > > getting the args quite happily
> >
> >
> > The only thing reasonable I can think of is that maybe a +load method 
> > in your app or some other lib is improperly using ObjC statements 
> > (which might trigger a call to NSProcessInfo before all the load 
> > methods are complete and NSProcessInfo can call +initialize).
>
>
> Nor that.  Not one of my classes has a +load.  About the only thing I 
> can think of is that it's a threading issue.
>
> I put some debug printfs in NSProcess|nfo, rebuilt and installed base 
> (sigh - 200Mhz machine), and watched it happily do +load and parse argv 
> and argc.  But when It got to +initialize, its local argc and argv were 
> mysteriously empty, which is suspicious.
>
> Maybe I'll build everything (again) with threading disabled and see what 
> happens.  Unless you have any better ideas.

I've changed the subject line as I dismissed your first post as spam ;-).

Well threading (per se) should not be the issue.  In fact I suspect you 
may have more trouble with threading disabled wrt GSWeb (not that you 
can't make it work single threaded).

Could you start your app an a recent gdb (6.0/6.1) and post a backtrace? 
 Could try using a watchpoint?  Does the trivial gsweb/Examples/hello 
or the Testing/DynamicElements show similar behavior?

Cheers,
David