bug#33225: [debbugs.el] Don't send control message immediately

[Noam Postavsky]

Michael Albinus <michael.albinus@gmx.de> writes:

> Thanks for the patch. In general, it works fine, so I guess you could
> push it. We (I)'ll play then for some days, and if there's no serious
> complaint, I'll make a new debbugs release.
>
> But I still have some comments :-)

I've updated according to your comments, I'll wait another couple of
days before pushing in case we can catch any more silly mistakes.

Regarding making a release, I have in mind next to bring in my commands
which produce control messages from git commits (both for attaching
proposed patches from an unpushed commit, and closing from a pushed
one).  Perhaps you want to wait for that before making a new release?
(or perhaps just the opposite, not sure what your release policy is)

> Please rename
>
> debbugs-control-make-message → debbugs-gnu-make-control-message
> debbugs-control-message-keywords → debbugs-gnu-control-message-keywords
> debbugs-control-message-commands-regexp → 
> debbugs-gnu-control-message-commands-regexp
> debbugs-control-message-end-regexp → debbugs-gnu-control-message-end-regexp
>
> Please add (the keybindings of) debbugs-gnu-make-control-message to the
> Commentary sections of debbugs-gnu.el and debbugs-org.el and to the nodes
> "Tabulated Lists" and "TODO Items" of debbugs-ug.texi.

Done and done.

> Compiling debbugs-gnu.el results in
>
> debbugs-gnu.el:1534:1:Warning: Unused lexical variable ‘addr’
> debbugs-gnu.el:1552:1:Warning: Unused lexical variable ‘micro’
> debbugs-gnu.el:1552:1:Warning: Unused lexical variable ‘subject’
>
> Maybe you can fix it?

Oops, yes.

>> +(defconst debbugs-control-message-keywords

>> +    "usertag" "user"
>> +    "documentation" ;; usertag:emacs.documentation
>> +    ))
>
> I suppose "user" is needed in case of debbugs-gnu-make-control-message
> only. Could we separate this, and offer completion only for that
> function?

Actually, thinking about it, there's really no need for a "user" control
message.  I've just removed it instead (it only needs to go in
debbugs-gnu-control-message-commands-regexp).

>> +        ((member message '("merge" "forcemerge"))
>> +         (format "%s %d %s\n" message bugid
>> +                 (read-string "Merge with bug #: ")))
>
> You have removed the bug number completion via
> debbugs-gnu-expand-bug-number-list. Why?

>> +        ((member message '("block" "unblock"))

> dito

Ah, I think what happened is that I started from a very old version
debbugs-gnu-send-control-message which didn't use
debbugs-gnu-expand-bug-number-list.  Fixed now (I also added back the
version arg to "close").

>> +@item user
>> +"user @var{username}"
>> +
>> +The username, read interactively, is either a package name or an email
>> +address.
>
> Please mention, that this is used in order to avoid giving the user name
> again, for further commands.

As mentioned above, I decided to simply remove this instead.  And I also
removed the duplicate avoidance feature of "documentation".  Making the
message shorter isn't that much of a benefit, and having it adds more
possible failure modes (e.g., if there is both "user emacs" and "user
something-else", things get tricky).

v3-0001-New-command-debbugs-control-make-message-Bug-3322.patch
Description: patch