Re: [PATCH v3 5/6] trailer: allow non-trailers in trailer block

Jonathan Tan [off-list ref] writes:

quoted

    * rs/c-auto-resets-attributes:
      pretty: avoid adding reset for %C(auto) if output is empty

And neither of the two colon containing line remotely resembles how
a typical RFC-822 header is formatted.  So that may serve as a hint
to how we can tighten it without introducing false negative.

The only "offending" character is the space (according to RFC 822),
but that sounds like a good rule to have.

I suspect that we should be willing to deviate from the letter of
RFC to reject misidentification.  I saw things like

	Thanks to: Jonathan Tan [off-list ref]
	Signed-off-by: A U Thor [off-list ref]

in the wild (notice the SP between Thanks and to), for example.
Rejecting leading whitespace as a line that does *not* start the
header (hence its colon does not count) may be a good compromise.

I think that "Signed-off-by:" is not guaranteed to be
present.

But do we really care about that case where there is no S-o-b:?  I
personally do not think so.

Defining a trailer line as "a line starting with a token,
then optional whitespace, then separator", maybe the following rule:
- at least one trailer line generated by Git ("(cherry picked by" or
"Signed-off-by") or configured in the "trailer" section in gitconfig
OR
- at least 3/4 logical trailer lines (I'm wondering if this should be
100% trailer lines)

I'd strongly suggest turning that OR to AND.  We will not safely be
able to write a commit log message that describes how S-o-b lines
are handled in its last paragraph otherwise.

I do not care too deeply about 3/4, but I meant to allow 75% cruft
but no more than that, and the fact that the threashold is set at
way more than 50% is important.  IOW, if you have

	Ordinary log message here...

	S-o-b: A U Thor [off-list ref]
	[a short description that is typically a single liner
        in the real world use pattern we saw in the world, but
	could overflow to become multi line cruft]
	S-o-b: R E Layer [off-list ref]

"last paragraph" is 5 lines long, among which 60% are cruft that is
below the 75% threshold, and "am -s" can still add the S-o-b of the
committer at the end of that existing last paragraph.  Making it too
strict would raise the false negative ratio.