On Sun, Oct 10, 1999 at 11:09:19PM -0600, Lou GLASSY wrote:
> 
> (borken is an intentional misspelling of "broken" :-)
> 
> dear all,
> 
> Has anyone tried out the NetBSD 1.4 Objective-C part of gcc lately?
> 
> I tried it this evening, and lo, it barfeth mightily.
> 
> The source code for a simple program I used to test gnu objc 
> is located at URL
> 
> 	http://www.cs.montana.edu/~glassy/software/obc.tar
> 
> 
> A script session of "What Went Wrong" follows:
> 
> ---cut here---
> 
> Script started on Sun Oct 10 22:28:35 1999
> $ uname -a 
> NetBSD caesar 1.4 NetBSD 1.4 (CARBON) #0: Mon Aug 16 12:28:57 PDT 1999     root@caesar:/usr/src/sys/arch/i386/compile/CARBON i386
> $ gcc --version
> egcs-1.1.1
> $ make 
> gcc -c main.m
> gcc -c ColorType-implementation.m
> gcc -o main ColorType-implementation.o main.o -lobjc
> ld: ColorType-implementation.o: RRS text relocation at 0x180c for "___objc_class_name_Object"
> $ ./main
> objc runtime: cannot find class Object
> Abort trap - core dumped
> $ exit
> 
> Script done on Sun Oct 10 22:29:09 1999
> ---cut here---
> 
> 
> I tried the same code under gcc-2.95.1 under both DUX 4.0e and Debian
> Linux 2.1 (2.036 kernel), and it ran correctly, so I'm guessing the 
> code itself is all right, and that my objc installation just needs
> some extra tickling to get it to work too.  The program is little
> more than a "hello world" kind of program, just enough to test out 
> making objects and calling methods.
> 
> If this kind of objc bug has already been seen & fixed in 
> 1.4.1 or -current, tell me & I'll go upgrade... :-)

I see this with -current, too, on NetBSD/m68k.
It goes away with COPTS=-static, so it has something to do with relocatable
code generation.

The old problem is still marked as open. I'll look at it, but can't promise
to understand it... I'm no toolchain wizard, normally.

Regards,
	-is