Re: master fd92023: Make checkdoc work with qualified methods

[Mauro Aranda <maurooaranda@gmail.com>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>>     (`(,(and
>>          (or 'defun 'defvar 'defcustom 'defmacro 'defconst 'defsubst
>>              'defadvice 'cl-defun 'cl-defgeneric 'cl-defmacro 'cl-defmethod)
>>          def)
>>        ,(pred symbolp)
>>        ;; Require an initializer, i.e. ignore single-argument `defvar'
>>        ;; forms, which never have a doc string.
>>        ,_ . ,_)
>
> I think you can reduce that to
>
>     (pred symbolp)
>
> and then check that (function-get def 'doc-string-elt) is non-nil.

I thought about that: it would "add support" to other definitions, and I
wasn't sure if it would be desired to check the docstring for all of them.
An el-search reports the following (without counting what checkdoc
already supports):

define-abbrev-table (*)
defface
defgroup
deftheme (*)
define-ibuffer-sorter
define-ibuffer-op
define-ibuffer-filter
defimage (*)
define-skeleton
defvar-local
lambda (**)
define-widget
defmath
define-overloadable-function
define-mode-local-override
define-obsolete-function-alias
define-obsolete-variable-alias
cl-iter-defun
cl-defstruct
cl-deftype
define-derived-mode
define-minor-mode
define-globalized-minor-mode
easy-menu-define
ert-deftest
iter-defun
define-generic-mode
define-inline
define-advice
pcase-lambda (**)
pcase-defmacro
define-erc-module
define-ccl-program
defun-mh
defmacro-mh
defgroup-mh
defcustom-mh
defface-mh
ediff-defvar-local
defun-cvs-mode (*)

* the pattern would need tweaking to detect the form
** wouldn't match, because of the second (pred symbolp)

Also, for some definitions (e.g., define-erc-module), the report might
be different because the file was/wasn't loaded.

To sum it up, I feared that just checking for a non-nil doc-string-elt
could make checkdoc somewhat annoying, but I'd like to know what other
people think.

>> Note that I need to do (forward-sexp 1) so the requirements of
>> cl--defmethod-doc-pos are fulfilled.  It may get messy if other defining
>> forms declare a doc-string-elt function that assumes a different point
>> position.
>
> The starting position of `doc-string-elt` is currently defined de-facto
> by the code in `lisp-string-in-doc-position-p`, so
> `cl--defmethod-doc-pos` had no choice in the matter ;-)

I didn't know that.  No worries then.

>> BTW, I've noticed that I forgot to add the Bug tag to my commit, I'm
>> sorry.  This commit was part of Bug#46918.
>
> I think this deserves a harsh punishment.

:-(