Re: Linux compat issue(s)

[ Available lists | Index of freebsd-hackers | Month of Oct 1996 | Week of 15 Oct 1996 | Raw email | View thread | Wrap long lines | Reply | Tag ]
From
<sos@FreeBSD.org>
Date
15 Oct 1996 14:39:37
Subject
Re: Linux compat issue(s)
Message-ID
199610152139.XAA00198@SandBox.CyberCity.dk


[ Hide this part ]
In reply to Terry Lambert who wrote:
>
> > How come that took so long Terry ??
>
> Heh. Instantly isn't fast enough for you? ;-).

Ah it took you several hours :)

> > So do we in the dynamically linked case, almost all ELF implemetations
> > on the x86 platform use different named/located interpreters.
> > It is only the statically linked binaries that is the problem.
> > Linux has the same problems we do, they have implemented another=20
> > hack than the one I suggest, just their method isn't very robust
> > but they're used to that, right :)
>
> ELF has a general problem with binary type recognition.

Exactly, that my point...

> One way would be to steal codes from CPU type and distinguish with
> magic number, or vice versa.

Hmm, well, yes but that *could* break on other archs then..

> It should also be noted that it's kind of silly to follow the SVR4
> EABI if you don't have the same trap entry points (ie: reallly follow
> it).

So right, why do we bother with ELF at all, we're no SVR4 (thank god)...


-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Sren Schmidt (sos@FreeBSD.org) FreeBSD Core Team
Even more code to hack -- will it ever end
..


Elapsed time: 0.068 seconds