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

[Simon Stapleton]

On 27 Apr 2004, at 20:41, Richard Frith-Macdonald wrote:
> Most likely there is some system specific bug ... quite possibly 
> nothing to do with your code.
> If the variables are being cleared between the end of the +initialize 
> and the start of +processInfo you *really* need to single step through 
> to find where it's going wrong.  Perhaps another thread could be 
> clearing the ivars the moment the lock in the runtime library which 
> protects the +initialize is released.  However ... there is no code 
> which *should* ever clear the variables afaik.

OK, so I managed to get it running, but I still don't see how.  Debug 
builds still crash (of course), so I can't debug, but I think I'm 
getting there.

What I did was this.

In NSProcessInfo.m, function _gnu_process_arguments(), I got rid of the 
IF_NO_GC macros.  They shouldn't have been doing anything anyway.  And 
miraculously, it started working.

What this seems to imply (to me, at least) is that for some reason base 
thinks I want GC.  GC may not be installed, but I think it thinks it 
is, for some bizarre reason.  To further bear this out, here's some of 
the output of gsweb/Testing/DynamicElements when run and accessed 
through a browser (I hadn't bothered to actually access this before, as 
it started OK), which is dumping log messages that look distinctly 
GC-ish

> session 0x10183c10 _releaseAutoreleasePool START
> session 0x10183c10 _releaseAutoreleasePool START
> session 0x10183c10 _releaseAutoreleasePool after garbage
> session 0x10183c10 _releaseAutoreleasePool after garbagesession 
> 0x10183c10 _releaseAutoreleasePool STOP
> session 0x10183c10 _releaseAutoreleasePool STOP
> Response Sent
> Response Sent
> threadWillExit START
> threadWillExit START
> GC** GarbageCollector collectGarbages START
> GC** GarbageCollector collectGarbages STOP
> threadWillExit STOP
> threadWillExit STOP

This still doesn't explain why _my_ app wouldn't start, but 
DynamicElements and Hello would.  Or, for that matter, eogenerator, 
defaults, all the other apps I've been using.  If the root cause was 
(or is) GC being on but not working, _everything_ should fail at 
startup, no?  Or maybe it's just plain (bad) luck.

Heck, I don't know.  I'm really losing the plot here.  This is my third 
day of beating my head against this, and I can't see the wood for the 
trees any more.

As I have the app working for the moment with the non-debug version, I 
think I'm just going to get on with developing my actual code, and let 
this drop for a bit.

Simon
--
PGP Key Id : 0x50D0698D