RE: amd64

Trying to close on this issue.

If I understand correctly the way current binutils
builds work, apparently for many platforms but
excepting ia32 (?), bintuils configure misunderstands
that glibc gettext is available, and then manages to
build an invalid assembler binary because a required
build flag (_GNU_SOURCE) for the not-using-glibc-gettext
case is missing from our build procedure.  The patch 
you supplied will fix that problem, but it's not clear
to me that we want to proceed using the wrong gettext....

Binutils configure gets the wrong impression
because it depends on a command which does not come
from glibc at all, but from the separate
gettext package.  We can't currently build the
package in time for binutils to be happy, because:

- gettext package can't be built static, so it
  can't happen in the bootstrap
- it won't configure-and-build dynamic so early
  in the lsbsi-intermediate phase (I don't know
  the exact failure)

Although I don't have the complete details yet,
from the initial comments I heard from Marvin this
morning, it sounds like other platforms are likely
now failing in the same manner.

How do we solve this?  Do we patch binutils'
configure script so it that assumes glibc's gettext
without actually making the test it currently fails?

And is the patch to define _GNU_SOURCE still going
to be needed even in the using-glibc-gettext case?

Thanks,

Mats