Re: Help with vger's rules, was Re: [ANNOUNCE] Git for Windows 2.10.2

From

Jeff King <peff@peff.net>

Date

Nov 3, 2016, 15:50 UTC

Message-ID

<20161103155054.24h3byh4fnpf3s6c@sigill.intra.peff.net>

In-Reply-To

<20161103153845.nhquuoko4utecit2@sigill.intra.peff.net>

On Thu, Nov 03, 2016 at 11:38:45AM -0400, Jeff King wrote:

> On Thu, Nov 03, 2016 at 11:34:53AM -0400, Jeff King wrote:
> 
> > This is missing a Content-Transfer-Encoding. I think the default is the
> > traditional 7-bit ascii encoding, but your body has characters with the
> > high-bit set (your UTF-8 bullet).
> > 
> > Try adding:
> > 
> >   Content-Transfer-Encoding: 8bit
> > 
> > I haven't seen this before, but I do recall that vger's MTA is very
> > picky about this and wants to rewrite transfer-encodings, so it seems
> > plausible.
> 
> Technically, I think you'd also need a:
> 
>   MIME-Version: 1.0
> 
> header. That being said, I just peeked at the send-email code and it
> looks like we try to add in these headers as necessary. It's possible
> there's a bug, though. What does "git send-email --dry-run" say is in
> the headers it sends out?

This probably helps:

diff --git a/git-send-email.perl b/git-send-email.perl
index da81be40c..784bb874b 100755
--- a/git-send-email.perl
+++ b/git-send-email.perl
@@ -1568,11 +1568,13 @@ foreach my $t (@files) {
 	push @cc, recipients_cmd("cc-cmd", "cc", $cc_cmd, $t)
 		if defined $cc_cmd && !$suppress_cc{'cccmd'};
 
-	if ($broken_encoding{$t} && !$has_content_type) {
+	if ($broken_encoding{$t}) {
 		$xfer_encoding = '8bit' if not defined $xfer_encoding;
-		$has_content_type = 1;
-		push @xh, "Content-Type: text/plain; charset=$auto_8bit_encoding";
-		$body_encoding = $auto_8bit_encoding;
+		if (!$has_content_type) {
+		    $has_content_type = 1;
+		    push @xh, "Content-Type: text/plain; charset=$auto_8bit_encoding";
+		    $body_encoding = $auto_8bit_encoding;
+		}
 	}
 
 	if ($broken_encoding{$t} && !is_rfc2047_quoted($subject)) {

but I think there is more cleanup to handle this case (e.g., it
auto-guesses utf-8, but it should pull the value from the existing
content-type header).

-Peff