This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: Binutuls is broken now.
- To: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- Subject: Re: Binutuls is broken now.
- From: "H . J . Lu" <hjl at lucon dot org>
- Date: Mon, 18 Jun 2001 18:46:55 -0700
- Cc: binutils at sourceware dot cygnus dot com, nickc at cambridge dot redhat dot com
- References: <20010618182216.A3400@lucon.org> <200106190134.DAA14429@ignucius.axis.se>
On Tue, Jun 19, 2001 at 03:34:11AM +0200, Hans-Peter Nilsson wrote:
> > Please state clearly how you
> > want to fix the problem.
>
> I want to fix it by not allowing -r except in cases we know
> work, as does the presented patch. Clear enough?
Let me tell you one thing. It is a very hard problem. It is almost
the same as using objcopy on relocatable files. You may never know if
it works or not until the resulting binaries don't work as expected. I
worked on fixing objcopy from i386/PE to i386/ELF for a while and Ian
explained to me why it was hard. That is why I checked in
http://sources.redhat.com/ml/binutils/2000-06/msg00271.html
>
> Allowing -r for files in formats that *can have* but don't
> happen to have relocations in a specific file is a tad
> confusing; things would fail for arches that happened to have
> relocations in a particular object-file.
>
> > I really don't care how you fix it as long
> > as the binary input is allowed again.
>
> Pity. You should care how "fixes" are done.
I only have limted time. I have to pick and choose. I can care less
about this one :-). In any case, please fix it now.
Thanks.
H.J.