Invitation: Dwarf2

[Jim Dehnert]

Various efforts, many of them for IA-64, are raising questions about Dwarf.
While some are IA-64-specific, most are more general, and it seems
appropriate to revive a committee to deal with them.

I have sent this to the participants in the C and C++ ABI committees
for IA-64, along with a few other interested parties.  Please forward
it to the relevant people (if any) in your organization (or elsewhere),
and have the interested people send me name, email, phone, and FAX --
I'll construct a list.  (Please don't assume you're on it until you
let me know you're interested.)  After next week, I'll construct an
alias and we can use that instead.

I suggest that we construct an initial list of issues and participants,
and begin meeting in 2-3 weeks.  We can volunteer a location (more or
less central to Silicon Valley participants) and a call-in phone number
for remote participants.  I'll set this up, but as debugging isn't really
central to my activities, I'll be looking for a volunteer to drive
it.  Let me know if you want to (or just can) take this on...

Some of the issues I'm aware of are:

General:

  * C++ has a need to be able to remove duplicate copies of objects
    (functions, virtual tables, ...) at link time, and it is desirable
    to remove the associated Dwarf information along with them.  Define
    an approach for decomposing the Dwarf information in a relocatable
    object file so this can be achieved by just removing some sections.

  * The original Dwarf specification assumed 32-bit programs, and there
    is some disagreement about the appropriate extension to 64-bit data
    models.

  * Fortran-9x support.


IA-64:

  * Register bindings.

  * The IA-64 ABI defines stack traceback descriptors for such purposes
    as exception handling which (largely? entirely?) duplicate the
    .debug_frame info.  Can/should we just bind to it instead of
    duplicating the information in .debug_frame?

-	    Jim Dehnert		dehnert at sgi.com
				(650)933-4272