Re: [PATCH] i386-netbsd* configury cleanup

[Jason R Thorpe <thorpej@wasabisystems.com>]

On Mon, Sep 02, 2002 at 11:44:40PM +0200, Mark Kettenis wrote:

 > There's a slight problem here: it seems to be impossible to
 > distinguish an OpenBSD binary from a NetBSD a.out binary.  That was
 > the main reason why I made i386-*-opensbd* identical to i386-netbsd.
 > As long as you keep the support for older NetBSD's this should work
 > fine.  Unfortunately...

Basically nothing that is done for i386-netbsdelf is pertinent to
i386-openbsd and I as the NetBSD target maintainer don't want to be
stuck in the position of having to worry about a breaking support for
a system, which *happens* to be based on an ancient version of NetBSD,
which I don't use.

In the signal trampoline example, suppose OpenBSD changed the VM layout
such that the singal trampoline start/end was different from NetBSD's.
If you can't distinguish the NetBSD and OpenBSD a.out binaries, then you're
stuck anyway.

What really has to happen is for there to be new OSABI numbers assigned
for OpenBSD, the default OSABI to be set based on the configured target,
and for the OpenBSD configs to not depend on NetBSD-specific code.

Sure, OpenBSD has made the bad decision of sticking with an inflexible
object format like a.out (no ABI ID tags), and made the bad decision
years ago to not change their a.out magic number (it's not like those
OpenBSD a.out binaries will run on NetBSD, even though they appear to
be NetBSD binaries).  However, I don't think it's particularly fair to
ask me to deal with the consequences of those bad decisions.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>