Re: objc handle_class_ref correction for powerpc -mrelocatable

[Rainer Orth <ro at TechFak dot Uni-Bielefeld dot DE>]

Geoffrey Keating <geoffk@thief.cygnus.com> writes:

> handle_class_ref was trying to emulate assemble_constant, but had
> the section hardcoded, and the section was hardcoded wrongly.

should this version of handle_class_ref go on the branch as well?  I've
just tried (as a quick hack) to take the trunk version of the function
(your patch alone didn't apply since the function had been changed before)
to the branch, rebuild cc1obj without my hack to disable the invokation of
handle_class_ref

	http://gcc.gnu.org/ml/gcc-patches/2001-03/msg00125.html

and was able to build a shared libobjc.so.1 on Solaris 8/SPARC, something
that had failed before without my hack.

Unfortunately, all libobjc execution tests on that platform still fail:

objc runtime: cannot find class Object
FAIL: objc/execute/_cmd.m execution,  -O 

but that happened before as well, while there had been a time not too long
ago when even this worked ;-(

So maybe this is just the necessary patch to finally have cc1obj properly
generate PIC code on all platforms?

	Rainer

-- 
-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University

Email: ro@TechFak.Uni-Bielefeld.DE