Should gcc/mkmap-symver.awk really use "nm"?

[Christian Jönsson <c dot christian dot joensson at telia dot com>]

I'm trying to understand what the file gcc/mkmap-symver.awk does....

I have been running boostraps of gcc-3.1 joined with binutils-2.12 for
a while now and I configure, build and check it like this:

(../configure sparc-linux --enable-shared --enable-threads=posix --enable-symvers >& configure.log; make -j 2 bootstrap-lean >& build.log; make -j 2 -k check RUNTESTFLAGS="--target_board 'unix{-fPIC,-fpic,}'" >& check.log ) &

that is, with symvers enabled. However, looking at the generated
libgcc.map file I'm confused, moreover, I'm confused by the mentioned
file: gcc/mkmap-symver.awk

In there, we have these lines, for instance:

# We begin with nm input.  Collect the set of symbols that are present
# so that we can not emit them into the final version script -- Solaris
# complains at us if we do.

state == "nm" && /^%%/ {
  state = "ver";
  next;
}

state == "nm" && ($1 == "U" || $2 == "U") {
  next;
}


Now, are these lines referring to "nm" as the binutils nm? If so,
should it really use nm... I mean, what about NM_FOR_TARGET or
whatever...

I am curious if the in-the-bootstrap generated in-tree binutils-2.12's
nm-new shouldn be used. Or shouldn't it?

Cheers,

/ChJ