Re: [PATCH V2] git-send-email.perl: Add --to-cmd

[Junio C Hamano <gitster@pobox.com>]

Joe Perches <joe@perches.com> writes:

> +	if (defined $to_cmd) {
> +		open(F, "$to_cmd \Q$t\E |")
> +			or die "(to-cmd) Could not execute '$to_cmd'";
> +		while(<F>) {
> +			my $t = $_;

"my $t" masks another $t in the outer scope; technically not a bug, but
questionable as a style.

> +			$t =~ s/^\s*//g;
> +			$t =~ s/\n$//g;
> +			next if ($t eq $sender and $suppress_from);
> +			push @to, parse_address_line($t)
> +			    if defined $t; # sanitized/validated later

This "if defined $t" makes my head hurt.  Why?

 * The "while (<F>)" loop wouldn't have given you an undef in $t in the
   first place;

 * You would have got "Use of uninitialized value" warning at these two
   s/// statements if $t were undef; and

 * Even if $t were undef, these two s/// statements would have made $t a
   defined, empty string.

> +			printf("(to-cmd) Adding To: %s from: '%s'\n",
> +				$t, $to_cmd) unless $quiet;
> +		}
> +		close F
> +			or die "(to-cmd) failed to close pipe to '$to_cmd'";
> +	}

In any case, this whole codeblock obviously is a copy-and-paste from
corresponding $cc_cmd codepath, and I wonder if you can refactor the
original into a common helper function first and then use it to make the
addition of this feature a smaller patch.

	if (defined $cc_cmd) {
        	push @cc, recipients_cmd($cc_cmd, 'cc');
	}
        if (defined $to_cmd) {
	        push @to, recipients_cmd($to_cmd, 'to');
	}

If you did so, the first patch that refactors to create a helper function
can address issues Ævar raised in the review to clean things up, no?

I notice that you use parse_address_line() while $cc_cmd codepath doesn't.
I haven't studied other codepaths deeply, but my gut feeling is that the
reason why the $cc_cmd codepath does not call parse_address_line() before
pushing the result to @cc is _not_ because strings on @cc shouldn't be
sanitized (the codepath to parse "Cc: " calls parse_address_line and
pushes the result to @cc), but because the code is simply sloppy.  So I
suspect that it would be Ok for recipients_cmd to call parse_address_line
unconditionally.

Hmm?