Re: `missing' needs patch for Ultrix4.4

[Akim Demaille]

>>>>> "Jim" == Jim Meyering <address@hidden> writes:

Jim> Hi Akim,

Jim> So consider a name containing some other shell meta-character.  

I don't understand what difference you make between the two approaches
wrt meta-characters.

Wer are talking about

( exec "$@" )

vs

prog=$1
shift
"$prog" "$@"

and I see no difference wrt meta-characters.

| /tmp % cat cat\ file                                      
| #! /bin/cat                                               
| /tmp % cat foo.sh                                         
| #! /bin/sh                                                
|                                                           
| # Jim.                                                    
| ( exec "$@" )                                             
|                                                           
| # Alexandre                                               
| prog=$1                                                   
| shift                                                     
| "$prog" ${1+"$@"}                                         
| /tmp % PATH=/tmp ./foo.sh 'cat file' 'cat file'           
| #! /bin/cat
| #! /bin/cat
| #! /bin/cat
| #! /bin/cat

Jim> It's not as if the change I suggested is *that* ugly, is it?

No definitely not.  In fact, it is more like choosing the right idiom,
and spread it.  depcomp and compile also use this trick a lot.  I'm
all for choosing one and keeping it.

In fact, my main grief is I don't understand the special `:' case.
Can't you exec :?

Also, the splitting $1 issue is not that simple, as it involves all
the Autotools.  Up to now, we never prohibited -flags given to $(CC).
The interface of AC_PATH_PROGS and the like also prohibits prognames
with spaces in it (so that for instance you can look for `bison -y'
and `yacc', that's the only place I know where this ``feature'' is
actually needed).

And precisely, let's speak about `bison -y'.  My opinion is that you
patch can make things go weird.  And my quoting $prog about might be a
bad thing.  Of course, it is the invocation of missing which should
take care of this, but...



I'm not saying I'm against it, I'm saying that a direction must be
chosen, and propagated :)