Re: Object-file formats

patteeuw@etcv01.eld.ford.com (Jack Patteeuw)
Wed, 15 Feb 1995 15:55:23 GMT

          From comp.compilers

Related articles
Object-file formats Paul_Long@ortel.org (1995-02-12)
Re: Object-file formats patteeuw@etcv01.eld.ford.com (1995-02-15)
| List of all articles for this month |

Newsgroups: comp.compilers
From: patteeuw@etcv01.eld.ford.com (Jack Patteeuw)
Keywords: linker
Organization: Ford Motor Co., Electronics Div., Dearborn, MI
References: 95-02-107
Date: Wed, 15 Feb 1995 15:55:23 GMT

Paul_Long@ortel.org (Paul Long) writes:
>How widely used is the Elf/DWARF object-file format? I have been told that
>while COFF is still useable and not obsolete, it is a rather old format.
>IEEE-695 has come along since, followed most recently by Elf/DWARF. I had
>planned on using COFF for a new development environment but am wondering
>whether this is wise. I have heard that Elf/DWARF is more expressive but
>also more complicated to implement (I've ordered a spec sheet from Intel).
>Any observations or experience with Elf/DWARF? Disadvantages/advantages of
>the different formats?


As a member of the PowerPC Embedded Application Binary Interface standard
committee, we faced the same questions you are asking. We ultimately choose
ELF (Executable and Linking Format according to SVR4 ABI, 3rd edition) along
with DWARF V1.1 (debugging With Arbitrary Record Format) for several reasons.


First, COFF, and ELF for that matter, have no "standardized" way of include
debugging information. There are various extensions to COFF (typically called
XCOFF) but they are not compatible across vendors (ie. XCOFF != XCOFF).
Theorically, you could combine DWARF with COFF, but I know of no one who has
done this.


Second, IEEE-695 seems to be "stabilizing". That is, not many new, especially
32 bit, compiler vendors seem to be adopting this standard. To make matters
worse, Microtec Research Inc and HP "extended" the standard a few years ago,
yet these extensions have never been formally sent back to IEEE for review/
voting/inclusion to the standard.


Finally, with Sun being the largest vendor of 32 bit UN*X workstations (ie.
most units in the field), a large population is already familiar with ELF. The
documentation is accurate and complete (make certain to get the 3rd revision).


The biggest problem we are currently facing is the lack of a "home" for the
proposed DWARF V2.0 standard. This version contains many important extension
to V1.1 for C++. It also solves problems that were left open in the previous
version. Unfortunately, when AT&T sold UNIX to Novell, Unix System Labs was
shut down (so I've heard). The work on DWARF V2.0 was being coordinated
through that organization. The last version of the proposed DWARF V2.0
standard that I have is dated July 27, 1993, but I know of no schedule to move
this from "proposed" to "standardized" or who will maintain the standard.


Perhaps someone else has more information on DWARF V2.0 ?
Jack
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.