Re: Can someone do a small fix in CVS ?

[Richard Frith-Macdonald]

On Friday, January 3, 2003, at 12:34  am, Nicola Pero wrote:

> works fine ... at least, from a makefile point of view - autogsdoc is
> called with all the proper arguments.
>
> The run of autogsdoc later aborts with
>
> ../Tools/obj/autogsdoc: Uncaught exception NSRangeException, reason: 
> Range location + length too great
>
> but it's not a problem with makefiles - maybe Richard wants to look at 
> it
> but it's a separate matter.

It doesn't happen on my system ... perhaps you are using an old 
autogsdoc.

> If you have more details on what exactly goes wrong with the autogsdoc
> invocation, it might help ... but unless a fatal typo slipped in
> somewhere, we have not touched gnustep-make support for documentation 
> in
> the last month, so I can't see why it would be gnustep-make's fault.

If I remember correctly, Chris is using a BSD variant, and I think 
(from past emails) it is buggy in some way in passing arguments (or the 
NSProcessInfo code to retrieve the process arguments is not reliable on 
BSD).  I think that this is triggered by the size/composition of the 
argument list (which I increased a day or two ago).  I don't expect 
that anyone other than Chris can realistically debug this if it's a 
problem specific to his system.

To re-iterate from the lat time (october) this was raised, I think the 
problem can be ...

1. make not passing correct arguments on his system (bug in make or in 
system libraries)
2. the operating system not forwarding arguments to the new process 
correctly on his system (operating system bug)
3. NSProcesssInfo not retrieving the arguments correctly on his system 
(bug in NSProcessInfo or system libraries)

1 and 2 can be tested by substituting a trivial dummy program for 
autogsdoc, and getting it to log what it receives.
3 needs either insertion of logging or running under gdb to see why 
NSProcessInfo is failing to get the arguments.