(j3.2006) Get_Command_Argument

[Van Snyder]

On Thu, 2011-12-15 at 17:54 -0800, Malcolm Cohen wrote:
> 
> >1.  Cross my fingers and hope C_Char is the same as default kind.
> 
> Will work for the time being, but...

Exactly.  I've had to rework 30-year-old code.  Until the idiots here
decided to re-write the spacecraft navigation suite in C++, the Fortran
version had legacy going back to 1956.  (They've invested 246 work years
in it, and it still doesn't work as well as the Fortran code it
"replaced" -- but it did win NASA's "software of the year" award.)

When somebody, in 2032, ports some of my codes that make the assumption
that c_char == default, which today's processors silently accept, and it
stops working, coping with that is an expense that could have been
avoided.

> >2.  Copy the string from the command line to the one that gets sent
> to
> >the C function.
> 
> You probably want to do that anyway to put the null char on the end.
> That would 
> seem to be the appropriate time to do whatever conversions are
> necessary.

I was doing
  l = len_trim(arg)+1
  arg(l:l) = char(0,c_char)

If c_char isn't default or ASCII or ISO_10646, I don't know how to do
the translation from default to it.  That's at least part of the reason
to want to get c_char characters from get_command_argument.

> >Which leads back to the largely-ignored original question:
> 
> Not ignored, you just didn't like the answer!

I said "largely."  And the answer was "it probably works almost
everywhere for the time being."  If "almost everywhere for the time
being" is good enough, why bother with a standard instead of an informal
agreement?

Then the discussion drifted off into whether the NUMBER argument has to
be default integer, and discussions of ISO 10646 file names, and ....