Building unique object files and binaries

Building unique object files and binaries

[Ian S. Nelson]

I'm sure this has been discussed before but is there a patch or any 
desire to modify gcc to add timestamping and some sort of unique 
identifier to each object file output (or maybe this is a binutils 
thing) so that they are all different and unique for the purpose of 
identification?

I didn't see anything in the docs that makes me think it exists already.

thanks and have a happy Thanksgiving,
Ian

-- 
Ian S. Nelson <nelcomp@attglobal.net>        303-666-0315
Nelson Computing of Boulder Colorado
Networking/Contracting/Custom Software/Linux Fast and Personal service

Re: Building unique object files and binaries

[Geoff Keating]

"Ian S. Nelson" <nelcomp@attglobal.net> writes:

> I'm sure this has been discussed before but is there a patch or any 
> desire to modify gcc to add timestamping and some sort of unique 
> identifier to each object file output (or maybe this is a binutils 
> thing) so that they are all different and unique for the purpose of 
> identification?

There is a strong desire to _remove_ timestamps and unique identifiers
from object files, so that they can be compared.  A very useful part
of GCC's testing relies on object files being able to be compared.

-- 
- Geoffrey Keating <geoffk@geoffk.org> <geoffk@redhat.com>

Re: Building unique object files and binaries

[Florian Weimer]

"Ian S. Nelson" <nelcomp@attglobal.net> writes:

> I'm sure this has been discussed before but is there a patch or any
> desire to modify gcc to add timestamping and some sort of unique
> identifier to each object file output (or maybe this is a binutils
> thing)

It's probably a binutils thing.  I think you can add such sections
using the ld '-r' option.

Re: Building unique object files and binaries

[Adrian von Bidder]

"Ian S. Nelson" wrote:
> 
> I'm sure this has been discussed before but is there a patch or any
> desire to modify gcc to add timestamping and some sort of unique
> identifier to each object file output (or maybe this is a binutils
> thing) so that they are all different and unique for the purpose of
> identification?
> 
> I didn't see anything in the docs that makes me think it exists already.

Perhaps it would make sense to add the optional timestamp in the
.comment section (where GCC writes the '.GCC (GNU) 3.0.2'). I agree with
Geoff that this is not a desirable feature in most cases, but it should
be trivial to add -ftimestamp option to gcc.

greets from Zürich
-- vbi