Re: proposed improvement of diagnostics for missing/extra operands

[Jim Meyering]

Paul Eggert <address@hidden> wrote:
>> I have mixed feelings about it.
>
> OK, some good points there, and I've composed a revised patch (below)
> which I hope addresses almost all the problems you mentioned.  The
> only exception is the last problem (when getopt reorders arguments),
> which I'll discuss below.

Thanks for adjusting those.

>> Going a little overboard, it can get a ugly if getopt
>> ends up reordering arguments:
>>
>>   new$ ./mv a --force
>>   ./mv: missing operand after `a'
>
> Here I think the new behavior is OK, since if you add the missing
> operand after the `a', things will work; and that's the most natural
> place to add the missing operand.

This seems fine.  No need to go overboard :)

> Here's the revised patch:
>
> 2004-06-20  Paul Eggert  <address@hidden>
>
>       * src/basename.c (main):
...

Applied.

I've made only one significant change so far:

2004-06-21  Jim Meyering  <address@hidden>

        * src/mknod.c (main): Don't segfault when calculating the
        expected number of operands for `mknod NAME'.

Index: mknod.c
===================================================================
RCS file: /fetish/cu/src/mknod.c,v
retrieving revision 1.80
diff -u -p -r1.80 mknod.c
--- mknod.c     21 Jun 2004 15:03:35 -0000      1.80
+++ mknod.c     21 Jun 2004 15:27:17 -0000
@@ -132,7 +132,12 @@ main (int argc, char **argv)
       newmode = mode_adjust (newmode, change);
     }
 
-  expected_operands = (argv[optind + 1][0] == 'p' ? 2 : 4);
+  /* If the number of arguments is 0 or 1,
+     or (if it's 2 or more and the second one starts with `p'), then there
+     must be exactly two operands.  Otherwise, there must be four.  */
+  expected_operands = (argc <= optind
+                      || (optind + 1 < argc && argv[optind + 1][0] == 'p')
+                      ? 2 : 4);
 
   if (argc - optind < expected_operands)
     {