From: ridg...@inls3.tmc.edu (Doug Ridgway)
Subject: Unixware binaries?
Date: 1996/08/05
Message-ID: <4u45mv$k24@news1.ucsd.edu>#1/1
X-Deja-AN: 172188757
references: <4u2snn$36t@nadine.teleport.com>
organization: Institute for NonLinear Science, UCSD
newsgroups: comp.os.linux.misc
Can one run Unixware binaries under Linux using iBCS? How is it supposed
to work? A friend sent me a couple of Hello world binaries compiled on
his Unixware box, and I've been unable to get them to run. The dynamically
linked one (http://wessex.ucsd.edu/alp/a.out) generates ``Command not found''
while the static linked one (http://wessex.ucsd.edu/alp/as.out) generates
a seg fault. The iBCS module appears to be loaded into my kernel. I'm
running 1.2.13, Red Hat 3.0.3.
Anybody have a clue what I'm doing wrong? Or is it not possible?
Thanks for the help,
doug.
dridg...@uscd.edu
(Please mail in addition to following up, UCSD's news feed is flaky...)
From: Mike Jagdis <ja...@purplet.demon.co.uk>
Subject: Unixware binaries?
Date: 1996/08/06
Message-ID: <1006.32090753@purplet.demon.co.uk>#1/1
X-Deja-AN: 172816491
x-mail2news-path: purplet.demon.co.uk
x-nntp-posting-host: purplet.demon.co.uk
sender: "newsout1.26" <ufg...@purplet.demon.co.uk>
organization: FidoNet node 2:252/305 - The Purple Tentacle, Reading
newsgroups: comp.os.linux.misc
* In a message originally to All, Doug Ridgway said:
DR> Message-ID: <4u45mv$...@news1.ucsd.edu>
DR> Can one run Unixware binaries under Linux using iBCS?
In theory, yes. It's a long time since I tried any Unixware stuff against it
and even then it was only Unixware 1.x stuff ('cos that's all I had).
DR> How is it supposed to work?
You just run them.
DR> A friend sent me a couple of Hello world binaries
DR> compiled on
DR> his Unixware box, and I've been unable to get them to run.
DR> The dynamically
DR> linked one (http://wessex.ucsd.edu/alp/a.out) generates
DR> ``Command not found''
The ELF loader can't find the dynamic loader required for the program
(/usr/lib/libc.so.1). Eric's substitute library
(tsx-11.mit.edu:/pub/linux/BETA/ibcs2/) works for many SVR4 things.
DR> while the static linked one
DR> (http://wessex.ucsd.edu/alp/as.out) generates
DR> a seg fault.
A static binary won't be recognised as non-native until it makes a non-Linux
system call. Because of this the ELF loader does not know that SVR4
behaviour is required and leaves page 0 unmapped for the process (where as
SVR4 maps it read only I believe - see linux/fs/binfmt_elf.c for true
enlightenment).
Three possibilities: 1. iBCS is broken for Unixware in some way (use the
iBCS trace program - see the iBCS docs for details). 2. The SVR4 library
code dereferences a null pointer when it shouldn't - and gets away with it
on a real SVR4 system. 3. Your friend's code is broken and dereferences a
null pointer - and gets away with it on a real SVR4 system.
Mike
From: r...@rmkhome.com (Rick Kelly)
Subject: Re: Unixware binaries?
Date: 1996/08/26
Message-ID: <4vrg3v$ceh@news-central.tiac.net>#1/1
X-Deja-AN: 176497514
references: <1006.32090753@purplet.demon.co.uk>
organization: The Man With Ten Cats
reply-to: r...@seahag.rmkhome.com
newsgroups: comp.os.linux.misc
Mike Jagdis (ja...@purplet.demon.co.uk) wrote:
: A static binary won't be recognised as non-native until it makes a non-Linux
: system call. Because of this the ELF loader does not know that SVR4
: behaviour is required and leaves page 0 unmapped for the process (where as
: SVR4 maps it read only I believe - see linux/fs/binfmt_elf.c for true
: enlightenment).
: Three possibilities: 1. iBCS is broken for Unixware in some way (use the
: iBCS trace program - see the iBCS docs for details). 2. The SVR4 library
: code dereferences a null pointer when it shouldn't - and gets away with it
: on a real SVR4 system. 3. Your friend's code is broken and dereferences a
: null pointer - and gets away with it on a real SVR4 system.
Just a matter of semantics but:
iBCS = SVR3 = COFF
iBCS2 = SVR4 = ELF
Unixware handles both COFF and ELF binaries.
--
Rick Kelly r...@seahag.rmkhome.com r...@rmkhome.com
http://tencats.rmkhome.com