From: Thomas Rast <trast@student.ethz.ch>
Date: Thu, 22 Mar 2012 08:19:26 GMT
Subject: Re: [PATCH 10/9 v4] difftool: fix regression in '--prompt' options
Message-ID: <871uoljbe9.fsf@thomas.inf.ethz.ch>
In-Reply-To: <7viphxz37j.fsf@alter.siamese.dyndns.org>

Junio C Hamano <gitster@pobox.com> writes:

> Tim Henigan <tim.henigan@gmail.com> writes:
>
>> +# the '--prompt' and '--no-prompt' options require special treatment
>> +# because they may be specified more than once...the last one "wins".
>> +for (@ARGV) {
>> +	if (($_ eq "-y") or ($_ eq "--no-prompt")) {
>> +		$prompt = 0;
>> +	} elsif ($_ eq "--prompt") {
>> +		$prompt = 1;
>> +	} else {
>> +		push(@diffargs, $_);
>> +	}
>> +}
>
> I really do not like the direction in which this series is going.  We do
> not have a similar --no-gui option to defeat --gui option that may appear
> earlier on the command line, but when we fix that bug (isn't it a bug?),
> we would have to teach this loop about that option, wouldn't we?
>
> In the end, won't you end up resurrecting the argument parsing loop that
> you got rid of with the first patch in your series?  Isn't this working
> around the problem introduced only because you are using Getopt::Long and
> hitting its limitations?

Limitations?  You can basically steal code from git-send-email.  As an
example:

---- 8< ----
#!/usr/bin/perl

use strict;
use warnings;
use Getopt::Long;

my $foo = 0;
my $rc = GetOptions("foo!" => \$foo,
		    "n" => sub { $foo = 0; },
		    "y" => \$foo);

print "$foo\n";
---- 8< ----

$ ./getopt-test.perl -n
0
$ ./getopt-test.perl -y
1
$ ./getopt-test.perl --foo
1
$ ./getopt-test.perl --no-foo
0
$ ./getopt-test.perl --foo --no-foo
0
$ ./getopt-test.perl -y -n
0
$ ./getopt-test.perl --foo -n
0

--
Thomas Rast
trast@{inf,student}.ethz.ch