Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> Eric S Fraga <e.fraga@ucl.ac.uk> writes:
>>
>>> On Wednesday,  4 Mar 2015 at 17:28, Eric Abrahamsen wrote:
>>>
>>> [...]
>>>
>>>> I'm still seeing an issue where, if I start right off typing a big
>>>> paragraph of text at the top of the message (no salutation or anything),
>>>> all the lines *after* the first line are indented by one tab. Subsequent
>>>> paragraphs are unaffected.
>>>
>>> Hi Eric,
>>>
>>> I had this problem for a long time.  It disappeared a some time ago now
>>> and I have no idea why.  However, while I had the problem, I trained
>>> myself to always start an email (that was not a response like this one)
>>> with some form of salutation!  More polite as well as avoiding the bug
>>> :)
>>
>> Well, sure :) I guess I'll try being politer!
>>
>> I just poked around a little bit, edebugging
>> `org-adaptive-fill-function'. I looked at the call to
>> `fill-context-prefix' two-thirds of the way down. I tested this with the
>> last email I sent, and I see that calling `org-adaptive-fill-function'
>> on the first paragraph results in `fill-context-prefix' being called
>> with the arguments 1 (the post-affiliated arg), and 447 (the end
>> position of the first paragraph). The result of that call is a tab.
>>
>> If I move to the second paragraph and do the same thing, the
>> post-affiliated arg was 447, and the end position is 475. The result of
>> that call was nil, which is probably what I wanted.
>>
>> My value of adaptive-fill-regexp, in this case is:
>>
>> "\\(\\([ 	]*[_.[:word:]]+>+\\|[ 	]*[]>|]\\)+\\)[ 	]*\\|[
>> ]*\\([-–!|#%;>*·•‣⁃◦]+[ 	]*\\)*"
>>
>> I will poke further as time allows. I don't know much about filling (and
>> have never understood what "post-affiliated" actually means), but assume
>> I can eventually get to the bottom of it...
>>
>> E
>
> It looks like the problem was that all the message headers are parsed as
> though they were part of the first paragraph of message body text. Why
> that should result in a secondary TAB indent I don't know, but
> regardless, Org probably should only be looking at the message body, and
> nothing else.
>
> The attached patch is a hack that adds the `mail-header-separator'
> regexp to the `org-element-paragraph-separate' regexp. That means it
> will only work for paragraphs, so there might still be weirdness if a
> message body starts with a list or what have you.
>
> Perhaps a better solution would be to narrow to the body of the message
> before doing the fill prefix calculation.
>
> Eric

And just for the heck of it, here's another patch that does it with
narrowing.

E