Forwarded from a discussion began on source-changes and continued
in gnats in the context of PR 11961...

What's the scoop on the LD_PRELOAD format change?

--jhawk

----- Forwarded message from Thomas Klausner <wiz@netbsd.org> -----

Date: Tue, 15 May 2001 10:13:34 +0200
From: Thomas Klausner <wiz@netbsd.org>
To: John Hawkinson <jhawk@MIT.EDU>
Cc: gnats-bugs@netbsd.org, Masao Uebayashi <uebayasi@jp.netbsd.org>
Subject: Re: kern/11961: ld.so(1) manual is misguiding about LD_PRELOAD
In-Reply-To: <200105150617.CAA04878@multics.mit.edu>; from jhawk@MIT.EDU on Tue, May 15, 2001 at 02:17:07AM -0400

Mutt made me believe that John Hawkinson wrote:
> 1. I believe the PR should be re-opened.

Actually, I think we should have a 'missing ld.elf_so' PR instead of
that PR and the one by you.

> 2. I'm confused:
> 
> | This man page only documents behaviour on a.out. ld.aout_so uses
> | a colon separated LD_PRELOAD (in contrast to ld.elf_so), so undo last
> | commit.
> 
> So the syntax of LD_PRELOAD changed from a.out to elf?

Yes.

> That seems rather...bogus. Is there a good reason for it?

I was told it's usual for ELF that LD_PRELOAD is space separated.
I'm only a documenter in this case.

> If so, is it documented somewhere, or mentioned in the CHANGES file
> or the release notes?
> 
> Is there a reason why we cannot support both syntaxes?
> 
> Is this discussion better had on tech-toolchain@netbsd.org?

Perhaps. Will you start it?

Bye,
 Thomas

-- 
Thomas Klausner - wiz@danbala.tuwien.ac.at

----- End forwarded message -----