Re: LT_AC_PROG_SED

[Robert Boehne]

Albert:

OK, I fixed this little problem and tested to prove the
changes.  Checking in to HEAD.

Robert

Albert Chin-A-Young wrote:
> 
> On Tue, Jan 29, 2002 at 10:26:54AM -0600, china wrote:
> > On Tue, Jan 29, 2002 at 10:21:29AM -0600, Robert Boehne wrote:
> > > Albert Chin-A-Young wrote:
> > > >
> > > > RE: http://savannah.gnu.org/patch/download.php?id=102
> > > >
> > > > What's the purpose of:
> > > >   if test -z "$SED"; then
> > > >   AC_CACHE_VAL(...)
> > > >   fi
> > > >   SED="$lt_cv_path_SED"
> > > >   ...
> > > >
> > > > Why not leave out the if/fi?
> > > >
> > > > --
> > > > albert chin (address@hidden)
> > >
> > > The intent was to allow the user to set SED
> > > explicitly, i.e.
> > > ./configure SED=/usr/local/gnu/bin/sed
> > >
> > > Hmmm, I thought I tested that, does it look
> > > like it would work that way?  It almost looks
> > > like when SED is set it will un-set it.
> > > Waddaya think?
> >
> > That's what I thought. But, it won't work because you force:
> >   SED="$lt_cv_path_SED"
> >
> > If this was *in* the if/fi, then the user could override $SED.
> >
> > You should take a look at AC_PATH_PROG for how it allows the program
> > to be overridden. They do:
> >   case "$SED" in
> >   /*)
> >     lt_cv_path_SED="$SED"
> >     ;;
> >   ?:/*)
> >     lt_cv_path_SED="$SED"
> >     ;;
> >   *)
> >     [what's in the if/fi block]
> >     ;;
> >   esac
> >
> > I say duplicate that to make it look like AC_PATH_PROG.
> 
> Oh, and they do this *inside* the AC_CACHE_VAL. Makes sense because
> you want to cache and the user can successfully override $SED inside
> AC_CACHE_VAL.
> 
> --
> albert chin (address@hidden)

-- 
Robert Boehne             Software Engineer
Ricardo Software   Chicago Technical Center
TEL: (630)789-0003 x. 238
FAX: (630)789-0127
email:  address@hidden