Patch for afsutils and getarg

[Chris Wing]

Hi. I downloaded the Arla code from CVS today and started playing with
getting klog, tokens, and unlog working properly with getarg(). I fixed a
few minor problems, but ultimately discovered that I couldn't get them to
work right without some changes to getarg() itself.

This patch:

http://www.engin.umich.edu/caen/systems/Linux/code/patches/arla-afsutils_and_getarg.patch

contains fixes to klog, tokens, and unlog, which makes them all use the
getarg() function and still work as before.

It also contains the following changes to getarg():

- changes for better pretty printing of the options in arg_printusage():
	- display ... within [] for arg_strings
	- display [-long, -s] for an argument with long and short names
	- display arg_generic options without long or short names
	  (see below)

- allow the short name to also match for an argument which has a long
  name. This means that:

	tokens -v
	or
	tokens -verbose

  will both give a verbose output, and

	tokens -version

  will work properly too.

  (previously, 'tokens -v' would be discarded as ambiguous between
   -verbose and -version)

- implement arg_generic_string and arg_generic_strings, which match any
  element of argv which does not match any of the other arguments. This
  is needed so that I can do:

	klog wingc
	klog -tmp wingc
	or
	klog wingc -tmp

  (an arg_generic_string matches 'wingc' in any of the above locations)

  likewise, in

	unlog cell1.edu cell2.edu

  'cell1.edu cell2.edu' matches an arg_generic_strings.


  A problem that I came across is that the existing code which calls
  getarg assumes that arguments can be given without a flag.
  (i.e. 'pts examine wingc'). This breaks the use of arg_generic types,
  because, according to this behavior:

	klog wingc

  'wingc' will be interpreted as matching the first argument in the list
  of getargs no matter what.

  (in this case, the first argument is "principal", so
	klog wingc
   is somewhat meaningful, it parses as:
	klog wingc at your.local.cell

   but
	klog wingc at umich.edu
   parses as:
	klog wingc at umich.edu@your.local.cell

   which doesn't make any sense)


  The above patch changes the behavior of getarg so that arguments cannot
  match without being given a flag, _unless_ they are defined as
  arg_mandatory. (which seems to be the case for the existing arla code in
  pts.c and elsewhere, but I haven't done an absolutely thorough check)


  In summary: the existing behavior of getarg (allowing arguments to be
  given without the flag), will be preserved as long as the arguments are
  defined as arg_mandatory. Is this okay?



Thanks,
Chris Wing
wingc at engin.umich.edu