Languguage OS 2/Languguage OS II Version 10-94 (Knowledge Media)(1994).ISO/gnu/objcissu.lha/bugs

Internet Message Format | 1993-03-01 | 6KB

Date: Wed, 11 Nov 1992 14:51:38 -0500
From: [e-mail] (John Hassey)
Message-Id: <9211111951.AA22923@matrix>
To: [e-mail]
Subject: problem with initialize_class() in core.c


  Running on m88k-dg-dgux

  There is no initialize method for class Object, so

    SEL           sel = sel_getUid ("initialize")

  results in sel having the value 0.

  Later,
    method
      = searchForMethodInList (class->class_pointer->methods, sel_getName (sel));

  gets an assert failure in record_get() since the selector is 0.

  The fix seems to be to add a check for (sel != 0).

  john.



Date: Tue, 5 Jan 1993 13:43:12 -0500
From: [e-mail] (John Hassey)
To: [e-mail]
Subject: Symbol __realloc

   On DG/ux the symbol _realloc in core.c is conflicting with
   a _realloc in libc.a.  Any other system have this problem ?

   I think we should change _realloc (and friends: _alloc, _dealloc ...)
   to something like: _objc_realloc.  To have a symbol with the same name
   as a libc routine, (or very close) is just asking for trouble.

   comments ?
   john.


John Hassey        [e-mail] (NeXT mail ok)
Data General Corp.    Research Triangle Park    NC, 27709


Date: Fri, 29 Jan 93 13:41:42 -0500
From: [e-mail] (Richard Stallman)
To: [e-mail], [e-mail]
Subject: [[e-mail]: symbol __objc_class_name_Object undefined]

Can one of you please see if the latest version has this bug,
and if so, get it fixed?

To: [e-mail]
Date: Thu, 28 Jan 1993 18:00:49 GMT
Organization: Oracle Corp., Redwood Shores CA
From: [e-mail] (Chris Oei)
Sender: [e-mail]
Subject: symbol __objc_class_name_Object undefined


When I try to use gcc-2.3.3 on AIX 3.2, I got the following:


amnesia> gcc -c m.m
Assembler:
/tmp/ccUOMDM2.s: line 175: 1252-023 The symbol __objc_class_name_Object is not d
efined.
/tmp/ccUOMDM2.s: line 175: 1252-040 The specified expression is not valid.
        Make sure that all symbols are defined.
        Check the rules on symbols used in an arithmetic expression
        concerning relocation.

The file in question is

#include <stdio.h>
#include <Object.h>

@interface A : Object {
        int x;
}

- show;
@end

@implementation A : Object {
        int x;
}

-show
{
        printf("%d\n",x);
}

@end

I also got the same error using gcc-2.3.1 on SunOS 4.1

Anyone have any ideas what's wrong?

If anyone has gotten the objective-c part of gcc to work properly on either
machine, I'd be interested in hearing about it.  At least I'd know there
exists some way of getting it to work.

Help!

(please?)

Thanks,
Chris
[e-mail]


Return-Path: <[e-mail]>
From: [e-mail] (Lele Gaifax)
Subject: Using GNU runtime on a NeXT
To: [e-mail]
Date: Mon, 22 Feb 1993 15:12:54 +0100 (MET)
X-Mailer: ELM [version 2.4 PL3]
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Length: 1662

Hi all,
since I just sent my "subscribe msg" for this list, please CC any answer to
[e-mail] too for a while...

So,
I started playing with GNU objc runtime, using gcc 2.3.3, on my NeXT.

Configured with "configure next" gcc does not produce anything usable on the
NeXT, since next.h state a #define NEXT_OBJC_RUNTIME and core.c does not check
for it: the program simply die with a assertion failed on "initialized",
because "objc_execClass" is not called at all.

I tried to modify the sources to get a gcc producing GNU standard objc runtime,
but with all combination I get an error from ld about section
(__OBJC, __message_refs): it seems to me that the NeXT ld tries to "simplify"
the object file merging all similar messages (same name) into one unique
symbol:
literal pointer section (__OBJC,__message_refs) does not have is exactly one relocation entry for each pointer
(BTW, what exactly this message mean? Is it a typo or my english knowledge is
fooling me?)

This is specific to the NeXT runtime: the native gcc produce a table
in that section with a label for each message. The GNU runtime instead wants
an array of simbols in that section.

Is it possible to get GNU Objc working with its own runtime on a NeXT too?
Ie, is it possible to tell the linker to disable or skip this step?

I'd like to use the gnu runtime. Since gcc doesn't implement ObjC as NeXT does,
I cannot use it with the appkit (it lacks @protocol, @class, ...), and so
without its runtime the compiler is of no use.

Thanx a lot,
lele.

--
Lele Gaifax - [e-mail]
...calling Emacs convenient, of course, is like calling oxygen useful
                    Rens Troost

Date: Mon, 1 Mar 93 11:01:34 -0500
From: [e-mail] (Vivek Khera)
To: [e-mail] (Geoffrey S Knauth)
In-Reply-To: [e-mail]'s message of 28 Feb 93 23:35:31 GMT
Newsgroups: comp.lang.objective-c
Subject: Re: gnu objc robustness
References: <[e-mail]>
Distribution:

In article <[e-mail]> [e-mail] (Geoffrey S Knauth) writes:

   After more testing, soon you will see a new runtime.


that sounds fantastic.  but... when i compiled and installed gcc 2.3.3
recently, i had it do the build for Objective C so i could play with
it some.  but after following the directions, i ended up without the
Object.h header file installed (even though the libobjc.a got
installed).  do i have to copy the objc/*.h files into the installed
location manually or did i miss a step somewhere?