Re: echo vs. printf regression (darwin8)

[Paolo Bonzini]

On 08/16/2010 04:27 PM, Peter O'Gorman wrote:
> On 08/16/2010 07:37 AM, Paolo Bonzini wrote:
> > On 08/16/2010 12:54 AM, David Fang wrote:
> > > Actually, 'pogma' pointed out that this comes from the autoconf macros
> > > in libtool.m4, near AC_MSG_CHECKING([how to print strings]).
> > > I'm using libtool-2.2.10 right now.
> >
> > I don't understand, is it breaking in Makefiles that you wrote? I don't
> > know of anything in {autoconf,automake,libtool} that would use $ECHO
> > this way.
>
> I think it is a namespace problem. What business does libtool have
> setting vars like ECHO, MV, etc.

That's correct.  However...

> If libtool had not unconditionally set ECHO to something other than
> 'echo', David's Makefiles would have continued to work when he moved
> from libtool-1.5.x to libtool-2.x.

... this doesn't seem to be the case.  It looks like David's makefiles 
were relying on the definition of ECHO that libtool provided, and this 
is clearly an issue of using an undocumented interface.  He, or some 
package he used, had to explicitly do AC_SUBST([ECHO]), and this is a 
very strong point against this usage, even if the variable name doesn't 
show that.

Paolo