glibc 2.0 targets

[Hans-Peter Nilsson]

On Wed, 18 Jun 2003, Dara Hazeghi wrote:
  Jim Wilson wrote:
> > There is no special
> > configuration name for glibc2.0 targets.  And I see
> > that the install
> > documentation already says that glibc-2.2.5 or later
> > is required for
> > libstdc++ to work.
>
> Good point. In that case, my only request is that
> somebody have a look at the patch included with the
> bug report (4068), since the other issues are
> documented. Thanks,

(As mentioned in private conversation; always use the lists.)
That patch isn't good enough, IMHO:

It's not appropriate, partly because it wouldn't apply
cleanly; there's an #ifndef USE_GNULIBC_1 there now.  You
should be able to put the patch code inside that ifndef.

Another reason why it's inappropriate is that there's no point
for the #else clause with the #includes since
MD_FALLBACK_FRAME_STATE_FOR still wouldn't use them; the #else
clause should be removed.

A third reason is that some part of gcj (I don't know where)
will fail silently due to lack of MD_FALLBACK_FRAME_STATE_FOR,
and that should be told to the user somewhere (no, not just in a
#warning).  Some configure check somewhere.  Or maybe the
documented requirement of glibc 2.2.5 is good enough.

If the first two issues are fixed, the patch stands a chance.
(I'm willing to test a patch to check that it doesn't break
libc1 with --enable-languages=c.)

brgds, H-P