Subject: Re: COMPAT_AOUT removal
To: Robert Elz <kre@munnari.OZ.AU>
From: Andrew Gillham <gillham@vaultron.com>
List: current-users
Date: 08/07/2000 00:01:48
Robert Elz writes:
> Date: Sun, 6 Aug 2000 23:23:47 -0400 (EDT)
> From: Todd Vierling <tv@wasabisystems.com>
> Message-ID: <Pine.NEB.4.21.0008062322230.7553-100000@server.int.duh.org>
>
> | I've yet to hear compelling arguments as to why it's detrimental.
>
> And that "ls /usr" using an a.out version of ls (like perhaps an old
> colorls binary) doesn't really list /usr is not detrimental ?
Yeah, figuring out that "cd /usr/blah" ends up in /emul/aout/usr/blah while
"cd / and cd usr/blah" ends up in /usr/blah took me a bit on a recent
upgrade. :)
> | After all, binaries for other emulations use this scheme.
>
> It would be nice to change that as well. Perhaps only the very first
> pathname lookup might use the magic redirection for some emulations
> (just enough to find a suitable ld.so - for those emulations where the
> source of the applicable ld.so can be obtained and modified to look for
> its libraries in some other place than /usr/lib (or /lib))
I think if this behavior was configurable some how it would be nice.
Or even use the (FreeBSD?) method of having a "linux" (or whatever) binary
that does the right thing. e.g. 'linux netscape' to launch netscape with
the option of searching /emul/linux or not. Adding the jail() syscall
might be useful for this also. e.g. In some cases you want an emulated
binary to _only_ have access to /emul/whatever.
-Andrew