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

[Simon Stapleton]

On 27 Apr 2004, at 12:31, Richard Frith-Macdonald wrote:
> > It occurs to me that for some reason, base has built with garbage 
> > collection, although I neither want nor need it (as the offending 
> > _gnu_arguments and friends are disappearing, never to be seen again, 
> > between +initialize and +processInfo - if gc was not being expected, 
> > they should be being retained).
> >
> > Does this make sense?
>
> No ... in the first place, garbage collection should not have this 
> effect.

Well, it would if garbage collection were not working (as it doesn't 
under my setup) but the system were configured for gc - RELEASE(), 
RETAIN, AUTORELEASE et al would then be defined to be effectively 
no-ops - any calls to them which one might expect to do something to 
the retain count of an object would (silently) have no effect 
whatsoever.

By the way, what's with all the IF_NO_GC(RELEASE(x)) stuff in 
NSProcessInfo.m, anyway?  The RELEASE/RETAIN etc macros should deal 
with GC transparently anyway, no?  Seems like a bunch of irrelevancies 
to me, although I'm willing to admit I may well be wrong on this ;-)

> in the second, you would have to explicitly turn on garbage collection 
> when you configured .... pretty much impossible to do by accident.

OK.  That makes sense.  I was wondering if configure might have done 
something behind my back, but it seems not.

> The values of _gnu_noobjc_argv and _gnu_noobjc_env should be zero 
> after +initialize is called, while _gnu_processName, _gnu_arguments 
> and _gnu_environment should be non-nil.

Yep, that's what I'm seeing (at least at the end of +initialize).  It 
seems that they are being released later, although why, I can't say.   
I _can_ say it's not me (explicitly) doing it, though.  They might be 
being trampled, but I can't see any reason why that should happen, 
either.

> I suggest you run under gdb (version 6 or later) and examine 
> _gnu_processName, _gnu_arguments, and _gnu_environment at the end of 
> +initialize and after stepping through each instruction until 
> +processInfo.
> That way you can find where these variables are being cleared.

Currently running with a watch on _gnu_environment to see what's going 
on, but of course that takes an age on my (prehistoric) machine.  
Still, I'll leave it chundering away, it seems like a reasonable course 
of action.

> If you can't run under gdb, add a function to NSProcessInfo.m to print 
> them out along with a label, and call that function after each line of 
> your code.

I suppose that might work, but I might be better off hooking into the 
Objc runtime to do that.  I have a _LOT_ of code...  And for some 
reason, other apps are working fine - the 'Hello' example that comes 
with gsweb, for example.  So it's got to be something to so with my 
code, or its organisation, or a relatively obscure bug that doesn't get 
exercised by that code.

more later.  thanks for the indulgence.

Simon

--
PGP Key Id : 0x50D0698D
--
Eagles may soar, but weasels don't get sucked into jet engines.