This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PATCH: add PGI (SGS Thomson) DWARF extensions to readelf


> Normally I would say that this patch does need a copyright
> assignment, since it is adding a new feature - the ability to decode
> PGI extensions to DWARF2 to the readelf tool.  I might consider it
> to be "obvious" if it could be described in a single, simple
> sentence, but without documentation of the extensions I do not think
> that this can be done.

Unfortunately I do not have any documentation which I'm allowed to
provide. (I'm not a PGI employee, but work on the TotalView debugger
at Etnus).

> I am sorry, but I agree with Doug on this one.  Without some
> non-proprietary documentation to verify the numbers and meanings of
> the PGI extensions we cannot really accept the patch.  If you can
> point us at such documentation, then I would be happy to reconsider
> the patch.  Alternatively if you would care to obtain an FSF
> copyright assignment, and then resubmit the patch, we will
> reconsider it then.

I've applied for a copyright assignment, so I expect that'll be done
in a few weeks (given that the forms have to cross the Atlantic
twice). 

Would a patch which just added the PGI enumerations to the headers be
acceptable in the meantime ? 

One of my main objectives in submitting the code in the first place is
to try to keep your dwarf2.h coherent with as many extensions to
DWARF2 as I know about.

Since there seems to be no other central repository of dwarf extension
enumerations I'm hoping that by having them in your dwarf2.h at least
_some_ compiler writers will look there before choosing the numbers they
use for their extensions and clashing with pre-existing extensions
used by other vendors. 

As someone who has to consume DWARF generated by many compilers, the
easier I can make it to avoid the situation in which one enumeration
value means more than one thing, the happier I am ! (A uniqueness
property which readelf also wants, I think ;-)

-- Jim 

James Cownie	<jcownie@etnus.com>
Etnus, LLC.     +44 117 9071438
http://www.etnus.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]