bug#59314: 29.0.50; EUDC and message-mode header completion

[Thomas Fitzsimmons <fitzsim@fitzsim.org>]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

[...]

>> Once I have that, I'll be able to extract all the configuration
>> resources (like .mailrc, ecompleterc, etc.) into a tarball.  Then we can
>> use that tarball to debug/redesign out-of-the-box completion scenarios,
>> starting from "emacs -Q + tarball", "emacs -Q + EBDB + tarball",
>> "emacs -Q + BBDB + tarball", etc.
>>
>> Without starting from "emacs -Q", it's impossible for me to know what I
>> might break in other people's configurations, when making changes to
>> message-mode's default completion code.
>
> I've already said this, but I don't think we should be pushing things in
> a direction that relies any external package or its particular behavior.
> The capf machinery should be enough, and that's very clearly defined.

I haven't figured out how the defaults are supposed to work for EUDC,
despite a bunch of attempts.  I'll keep at it, but wanted to post
status.

Here's a recipe to replicate what I'm trying:

cd test/lisp/net
/usr/sbin/slapd -h "ldap://127.0.0.1:3899" -d 0 -4 \
    -f eudc-resources/slapd.conf &

mkdir test-home
cat >test-home/.emacs <<EOF
(setq ldap-host-parameters-alist '(("ldap://localhost:3899"
                                    base "dc=gnu,dc=org" auth simple)))
(setq eudc-server-hotlist '(("ldap://localhost:3899" . ldap)))
(setq eudc-ignore-options-file t)
EOF
HOME=test-home ../../../src/emacs

C-x m emacs-ertTAB

There seem to have been many attempts in the code to get that to work by
default, but currently all the various approaches fail.  The result is
that three spaces are added to the buffer after "emacs-ert".

If instead of pressing TAB, one invokes M-x eudc-expand-try-all RET, the
EUDC multiple results UI is presented, and the completion one selects is
correctly inserted.  (This proves that message machinery is failing even
though EUDC and LDAP work.)

I've been hacking around the various EUDC code paths in message.el
(including the (setq message-expand-name-standard-ui t) one) trying to
get something to work by default but I haven't succeeded yet.

Thomas