Re: usage() in install(1)

On Fri, Feb 06, 2004 at 07:28:44PM +0100, Poul-Henning Kamp wrote:
> In message <20040206100605.GB27888_at_FreeBSD.org.ua>, Ruslan Ermilov writes:
> 
> >Poul-Henning,
> >
> >I was very puzzled today trying to understand what is this line
> >number that install(1) is telling me about, especially that I've
> >just run it without any options, to see its usage.
> >
> >Please revert the part of the revision 1.62 where it was added.
> >I have the following reasons why it shouldn't have been done:
> 
> I added it because the error messages from [x]install are singularly
> uninformative and at least the line numbers could serve as "guru
> meditation material".
> 
There are four cases where it displays the usage:

1.  When asked (``install -?'' or ``install'').

2.  Incorrect option or missing option argument: getopt(3) takes
    care of that.

3.  Incompatible options are specified (-d and -s): the usage is
    shown to let you decide what do you really want: -s or -d.
    This is how most bin/ utils do that.

4.  Wrong number or types of arguments: the usage is shown to
    let you know the correct syntax.

> The correct solution of course is to replace the totally uninformative
> usage message
> 
Come on now!  How is it totally uninformative?  It presents you
with the correct usage when you misuse it.  You'd better watch
the cpio(1) usage.  ;)

> which the program indiscriminatly spits out on any sort
> of trouble with intelligent informative error messages.
> 
Only 3 and 4 lack the verbose reason, but this is how traditional
BSD programs behave.  If it will make your life easier, feel free
to commit the attached patch.


Cheers,
-- 
Ruslan Ermilov
FreeBSD committer
ru_at_FreeBSD.org