AC_LIBOBJ problem?

[Harlan Stenn <Harlan.Stenn@xxxxxxxx>]

I haven't looked at this hard.

tftp-hpa's configure.in has a line that looks like this:

 AC_SEARCH_LIBS(xmalloc, iberty, , LIBXTRA=true [AC_LIBOBJ(xmalloc)])

Under autoconf-2.53 this eventually produced:

 LIBXTRA=true LIBOBJS="$LIBOBJS xmalloc.$ac_objext"

while under autoconf-2.58 and -2.59 it produces:

 LIBXTRA=true case $LIBOBJS in
    "xmalloc.$ac_objext"   | \
  *" xmalloc.$ac_objext"   | \
    "xmalloc.$ac_objext "* | \
  *" xmalloc.$ac_objext "* ) ;;
  *) LIBOBJS="$LIBOBJS xmalloc.$ac_objext" ;;
 esac

and I see this on ome of my machines:

 ...
 tftp-hpa-0.28/configure: line 11526: syntax error near unexpected token `)'
 tftp-hpa-0.28/configure:
 line 11526: `  *" xmalloc.$ac_objext "* ) ;;'
 gmake: *** [foo] Error 2

the target machine is a redhat 7.3 box.

While I'm digging to find the root cause, I wondered if it wouldn't be
better in general if the test was something more like:

 LIBXTRA=true case " $LIBOBJS " in
  *" xmalloc.$ac_objext "* ) ;;
  *) LIBOBJS="$LIBOBJS xmalloc.$ac_objext" ;;
 esac

And I have vague recollections of several older shells hating something that
looked like that...

H (who may be too tired to think right now...)