Re: [Pan-users] References header posting error patch?

[Duncan]

Dominique Dumont posted on Sun, 10 Oct 2021 11:53:25 +0200 as excerpted:

> On mercredi 6 octobre 2021 19:51:50 CEST Duncan wrote:
>> Now that we're getting a somewhat workable pan again, the biggest
>> remaining issue I've been seeing is the posting error due to pan
>> messing up the references header (bad character conversion I believe).
> 
> I've tested followup post with pan from git and I got no problem. Here's
> the relevant output of pan --debug

Try replying to a message with a longer message-id.  See below...

> Could you test on your side ?

It's intermittent but I'm still seeing it, altho I've not updated in a 
few days.

I follow mailing lists, including this one, using pan via gmane.io.  So 
assuming this posts you can see the git commit I'm on in the user-agent 
header.

While I've not tried pan --debug yet (just went back to work so suddenly 
have less time again), I know some replies post just fine, while others 
the server refuses and I get an error in the log:

Posting of "Re: Whatever title" failed: 441 Invalid syntax encountered in 
References header field body (unexpected byte or empty content line)

The post does end up in pan's sent folder, however, and (after a pan 
restart so it shows as cached), I can look at it there.

Again on gmane.io, an example that fails a reply (still the newest thread 
as I post this):

From: Mark Knecht <markknecht@gmail.com>
Newsgroups: gmane.comp.kde.linux
Subject: Google Chrome - no title bar this morning
Date: Wed, 22 Sep 2021 06:51:56 -0700
Approved: news@gmane.org
Message-ID: <CAK2H+ed34A-
Kg0h81h1z7RhHQSg4iVhw5tReXa73bsDQhTjQxA@mail.gmail.com>

Note the length of that message-id (which I see pan wrapped above).

When I look at the reply I tried to send in the sent folder, it seems the 
problem is that pan tries to wrap the references line, leaving only the 
bare "References:" on one line, with the message-id tab-folded to the 
second line.

Now it has been awhile since I looked at the rfcs, but while tab-folding 
is (IIRC) allowed, based on the behavior and the error, a bare header 
line with only the label, *all* the content folded to the next line, 
appears to be illegal -- or at least gmane is saying it is.

Assuming that's the problem the error would only trigger (for the 
references header anyway but presumably it could happen to any header 
with a similarly long unspaced-string) on real long message-ids.  Shorter 
message-ids would be fine, thus explaining the intermittency.

As I mentioned in the OP, I thought I saw a patch somewhere fixing this, 
tho IIRC in the case they were seeing it was a different header, but at 
the time I couldn't get pan to build at all so I left the patch until I 
could get pan building and I've not found it since.  Unfortunately I 
don't remember whether it was on the list (dev or user) or on gitlab that 
I saw it.

Additionally, it /may/ be that patch fixed a different bug and you may 
have applied it already, because from that patch description I had 
interpreted the described bug to be a character-conversion bug while this 
one seems to be a header-folding bug.  But it was preventing some replies 
as this one is, so I thought I'd ask about that patch's status and if the 
author could point me to it again if it hadn't already been applied, thus 
this thread.

Well I only have about four hours to sleep before I should be getting up 
for work, so... will this message post?

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman