Re: Mixing X and GNUstep

[David Ayers]

Raffael Herzog wrote:

> On 2003-07-12 20:02:36 +0200 David Ayers <d.ayers@inode.at> wrote:
>
> > This is rather strange.  It did finish the NSLog(@"NOT document 
> > based"); after which it merely has to return NO; yet it takes 
> > relatively long before it logs the event.
>
> I had the impression that it logs immediately, but if you say so... :) 
> I'll use printf();flush() in the future. 

Using NSLogs is fine!  What I meant, is that the last output of 
NSLog(@"NOT document based"); successfully finished it's output.

> > Does line 136 of NSDocumentController.m correspond to the return NO; 
> > or is it the beginning of another iteration before it logs the 
> > bundle? (There seem to be mutliple calls to isDocumentBasedApplicaion).
>
>
> It corresponds to the line
>
>  NSDictionary *dict = [bundle infoDictionary];

Ok, so you are in a new invokation of isDocumentBasedApplication.  But 
this time, the output of NSLog([NSString stringWithFormat: @"Got bundle: 
0x%08lx", bundle]); did make it before the ... oops, my bad.  I though 
the "Received an untrapped event 94" was already part fo the error message.
/me cleans his glasses.

> According to the log, bundle is not nil, but the method call segfaults 
> at line #177 of sendmsg.c, in objc_msg_lookup. That's the line:
>
>      result = sarray_get_safe (receiver->class_pointer->dtable, 
>                 (sidx)op->sel_id);
>
> Transscript from gdb after the segfault:
>
>  (gdb) frame
>  #0  0x4064bd92 in objc_msg_lookup (receiver=0x818d1d0, op=0x403c0768) 
> at sendmsg.c:177
>  177           result = sarray_get_safe (receiver->class_pointer->dtable,
>  (gdb) print receiver
>  $2 = 0x818d1d0
>  (gdb) print receiver->class_pointer
>  $3 = (struct objc_class *) 0xdeadface
>  (gdb) print "LOL!"
>  $4 = "LOL!"
>  (gdb) print receiver->class_pointer->dtable
>  Cannot access memory at address 0xdeadfaee   <==== AHA! 

Yes, check NSObject's dealloc implementation.  IIRC we put 0xDeadFace 
into the class pointer to detect memory manegment errors.  It seems to 
be a retain/release of the bundle's infoDictionary issue.

> > I think maybe you should use another compiler.
Nope, sorry, my oversight.
/me checks bundle infoDictionary code....

Cheers,
Dave