Re: [PATCH] ruby: fix ruby 3.1 warnings

[Tomi Ollila <tomi.ollila@iki.fi>]

On Mon, Apr 19 2021, David Bremner wrote:

> Felipe Contreras <felipe.contreras@gmail.com> writes:
>
>> Hello.
>>
>> On Sat, Apr 17, 2021 at 7:20 AM David Bremner <david@tethera.net> wrote:
>>
>>> Thanks for the patch. I have a couple of questions / comments
>>>
>>> - It doesn't apply against current master (5248f55d5f1). Can you rebase
>>>   it?
>>
>> Huh? That's precisely the commit I used as a base. It applies cleanly.
>>
>
> Ah right, you hit an annoying mailman bug related to encodings and line
> endings. I have unmangled the patch and applied it to master.

...which is not actually a bug, just features of git and mailman
that does not work together...

IIRC what happens there:

- smtp server receives message encoded as 8bit, possibly only LF 
  (line-feed 0x10) as line endings (I saw this when sent to myself
   message routed back to me via smtp.iki.fi)

- if only LF as line ending, something in pipeline changes that to CRLF
  (carriage-return line-feed 0x13 0x10) (allowed according to specs)

- now (if not before) mailman gets the message, notices it has 8bit
  encoding, converts that message to have base64 encoding

- mailman sends the email now base64-encoded to recipients

- user saves message and runs git-am to the message

- since content now has CRLF line endings and the content it is comparing
  to has only NL, the content just does not match. git-am cannot know that
  (w/o some magical heuristics) that the CRs are actually extra characters
  should not be there (having CRLF may be as relevant as only LF)

In another machine I still have David's email to mailman in one firefox
tab; I recall trying to write something there to explain why mailman
is buggy there, but eventually could not be convincing enough ;/)

git-am could have an option to strip CR's from messages just to overcome
the possibility shown above 

(otoh?) latest git-send-email has this option: --8bit-encoding
-- and if not given (and not in git config), according to namual page,
will prompt for an encoding (IIRC I've seen this w/ git 2.30)

HTH :D

Tomi