Re: echo vs. printf regression (darwin8)

[Paolo Bonzini]

On 08/16/2010 03:54 PM, Eric Blake wrote:
> On 08/16/2010 07:49 AM, Eric Blake wrote:
> > On 08/15/2010 04:11 PM, Paul Eggert wrote:
> > > On 08/15/10 21:26, David Fang wrote:
> > > > I'm desperate enough to use sed to patch config.status after it's been
> > > > generated:
> > >
> > > If you're that desperate, you're desperate enough to use bash
> > > instead of the broken Darwin shell, no?
> > >
> > > To others:
> > >
> > > Would it be appropriate to patch Autoconf to generate a 'configure'
> > > that rejects that implementation of 'echo', if it discovers the bug?
> >
> > The bug is not in 'echo', but in the improper use of $(ECHO) within ``
> > inside the Makefile.  Autoconf has no business inspecting whether
> > Makefile.am has a use of $(ECHO) that might be broken due to libtool's
> > change of how $(ECHO) is defined, and whether that change will cause
> > breakage because $(ECHO) is being used incorrectly in the Makefile.
> >
> > I don't see how changing anything in autoconf could help in this situation.
>
> That said, it may still be possible to patch libtool to define $ECHO in
> such a way that will be more robust to use with `` in a Makefile.

Considering libtool does not do AC_SUBST([ECHO]), I don't think this is 
useful.  If at all doable: consider that $ECHO might even expand to 
func_fallback_echo on some systems.

Paolo