82 lines
2.2 KiB
Plaintext
82 lines
2.2 KiB
Plaintext
Subject: dynamic loading on NeXTSTEP
|
|
From: horst@il.us.swissbank.com (Ted Horst)
|
|
To: python-list@cwi.nl
|
|
Date: Fri, 6 Sep 1996 15:02:26 GMT
|
|
X-Newsgroups: comp.lang.python
|
|
X-Organization: Swiss Bank Corporation CM&T Division
|
|
|
|
|
|
These are the steps that I took to get dynamic loading working under NeXTSTEP
|
|
(3.3 m68k, i486) for Python1.4beta3 (including the Numeric stuff).
|
|
Unfortunately, I don't know anything about the configure scripts, so I don't
|
|
know how to fix these things properly, but this works for me. I would be
|
|
interested in how other people have done this as well.
|
|
|
|
/configure --prefix=/home/horst/Unix --with-readline=/home/horst/Unix/lib
|
|
--with-thread --with-next-archs='m68k i486' --without-libm
|
|
|
|
edit Python/importdl.c
|
|
under:
|
|
#ifdef USE_RLD
|
|
add:
|
|
#define SHORT_EXT ".a"
|
|
#define LONG_EXT "module.a"
|
|
|
|
edit Modules/Setup
|
|
uncomment *shared*
|
|
(other editing as desired)
|
|
|
|
make (just to create the Makefiles)
|
|
|
|
edit Makefiles
|
|
change all AR= ar to AR = libtool
|
|
change all $(AR) cr ... to $(AR) -o ...
|
|
in Modules/Makefile change:
|
|
libm =
|
|
LDSHARED= cc -r -nostdlib $(CFLAGS)
|
|
CCSHARED= -r -nostdlib
|
|
LINKFORSHARED= -u libsys_s
|
|
|
|
make "OPT=-O3 -fschedule-insns2 -ObjC -arch m68k -arch i486"
|
|
|
|
|
|
Ted Horst
|
|
(not speaking for Swissbank)
|
|
|
|
|
|
Subject: Re: dynamic loading on NeXTSTEP
|
|
From: horst@il.us.swissbank.com (Ted Horst)
|
|
To: python-list@cwi.nl
|
|
Date: Fri, 6 Sep 1996 16:27:32 GMT
|
|
X-Newsgroups: comp.lang.python
|
|
X-Organization: Swiss Bank Corporation CM&T Division
|
|
|
|
Ted Horst writes
|
|
>
|
|
> These are the steps that I took to get dynamic loading working under NeXTSTEP
|
|
> (3.3 m68k, i486) for Python1.4beta3 (including the Numeric stuff).
|
|
|
|
<SNIP>
|
|
|
|
> make "OPT=-O3 -fschedule-insns2 -ObjC -arch m68k -arch i486"
|
|
>
|
|
> Ted Horst
|
|
> (not speaking for Swissbank)
|
|
|
|
Wow, I guess I was overzealous with the optimizations. Upon further
|
|
testing, I discovered that the Numeric stuff did not work properly on
|
|
Intel machines, but did work properly on m68k. I dropped the
|
|
optimization to -O2 (leaving out the -fschedule-insns2), and it seems
|
|
to behave properly on both architectures now.
|
|
|
|
So,
|
|
|
|
make "OPT=-O2 -ObjC -arch m68k -arch i486"
|
|
|
|
looks like a better choice.
|
|
|
|
Sorry for the misguidance,
|
|
|
|
Ted Horst
|
|
(still not speaking for SwissBank)
|