On Sun, 19 Apr 2009, Roland McGrath wrote:
> is the goal for elfutils to eventually supplant binutils with a
> set of tools on ELF-only platforms or just to provide alternate
> implementations (or a baseline for alternate tooling) which would sit
> along-side binutils?
Those kinds of goals are pretty abstract given the realities of where we
are now. I don't find it real meaningful to say what "the goal" is in the
absence of any material plans, schedules, or resource commitments to get
Many of the tools that elfutils provides now are as probably good or better
than their binutils counterparts (addr2line, ar, nm, ranib, readelf, size,
strings, strip). Some of those are already used in preference to binutils
tools in some production build processes (eu-strip in rpm builds). In
these cases, we might think those programs are ready to "supplant" their
binutils counterparts (or might be soon with only moderate effort).
Other tools are already significant new features that binutils has never
had and probably never will (elfcmp, elflint, unstrip, with dwarfcmp and
dwarflint coming later).
elfutils in recent times have been in the DWARF-related areas that mostly
do not show up as new command line tools. Those goals are more or less
unrelated to providing the other command line tools binutils has that
elfutils does not have equivalently functional implementations of yet
(objcopy, objdump, as, ld) and/or making the already "more finished"
elfutils tools listed above be better/sufficient drop-in replacements for
their binutils counterparts (and verifying the world of uses, etc.).
The "if we could wave a wand" goals include perfectly featureful ld, as,
etc. so that elfutils tools would replace binutils tools for all uses we
think are not insane. (Various BFD features like support for non-ELF
formats or strange embedded targets we would do in different ways rather
than ways necessarily drop-in compatible, if at all.) Work toward those
goals happens sporadically, basically in hackers' personal spare time, and
is not formally planned or scheduled at all. (Perhaps this is what you
meant by "a baseline for alternate tooling", I'm not sure.)
As I indicated, all the proximate goals are library, API, and arcane kinds
of tool things, not the rest of what the average person probably thinks
elfutils and binutils are about. Both are in fact about more things than
the familiar collection of command-line tools.
so the dev community is interested primarily in the library work, and
the user community is often the dev community. Fine, fine. I don't
think I'm the right person to write documentation on the library.
OTOH, I'm happy to document anything in /bin or /usr/bin. Interest
sounded tepid, but I'll try and start somewhere...
Technical Account Manager, Red Hat, Inc
office: 919 754 4157, cell: 919 758 9234