Re: [PATCH v4 5/6] send-email: --in-reply-to=<file> populate header fields

[Samuel GROOT <samuel.groot@grenoble-inp.org>]

On 06/08/2016 08:23 PM, Junio C Hamano wrote:
> Samuel GROOT <samuel.groot@grenoble-inp.org> writes:
>
>> +if ($initial_reply_to && -f $initial_reply_to) {
>> +	my $error = validate_patch($initial_reply_to);
>
> This call is wrong, isn't it?
>
> You are not going to send out the message you are responding to (the
> message may not even be a patch), and you do not want to die with an
> error message that says "patch contains an overlong line".

We used to handle email files like patch files (with Cc:, To:, From:, 
... fields, a patch is almost an email, with some trailers).

But that `validate_patch` subroutine call is indeed useless here, I will 
remove it.

>> +	die "fatal: $initial_reply_to: $error\nwarning: no patches were sent\n"
>> +		if $error;
>> +
>> +	open my $fh, "<", $initial_reply_to or die "can't open file $initial_reply_to";
>> +	my $mail = Git::parse_email($fh);
>> +	close $fh;
>
> 	my $header = Git::parse_email_header($fh);
>
> perhaps?

Git::parse_email will be renamed into Git::parse_email_header in v5.

>> +	my $initial_sender = $sender || $repoauthor || $repocommitter || '';
>> +
>> +	my $prefix_re = "";
>> +	my $subject_re = $mail->{"subject"}[0];
>> +	if ($subject_re =~ /^[^Re:]/) {
>> +		$prefix_re = "Re: ";
>> +	}
>> +	$initial_subject = $prefix_re . $subject_re;
>
> I am not sure what the significance of the fact that the subject
> happens to begin with a letter other than 'R', 'e', or ':'.
>
> Did you mean to do something like this instead?
>
> 	my $subject = $mail->{"subject"}[0];
> 	$subject =~ s/^(re:\s*)+//i; # strip "Re: Re: ..."
>         $initial_subject = "Re: $subject";
>
> instead?

That's way cleaner, thanks!

> By the way, this is a good example why your "unfold" implementation
> in 4/6 is unwieldy for the caller.  Imagine a rather long subject
> that is folded, i.e.
>
> 	To: Samuel
>         Subject: Help! I am having a trouble running git-send-email
>             correctly.
> 	Message-id: <...>

It's a good point. What you suggested in 4/6 reply will be used in v5.