Re: master 496bed5: Change command-completion-using-modes-p to defun

Re: master 496bed5: Change command-completion-using-modes-p to defun

[Stefan Monnier]

> -(define-inline command-completion-using-modes-p (symbol buffer)
> +(defun command-completion-using-modes-p (symbol buffer)

`define-inline` can be used to define inlinable functions, but it's not
like `defsubst`: the body needs to be changed to look a bit more like
what a macro would look like.


        Stefan

Re: master 496bed5: Change command-completion-using-modes-p to defun

[Lars Ingebrigtsen]

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

>> -(define-inline command-completion-using-modes-p (symbol buffer)
>> +(defun command-completion-using-modes-p (symbol buffer)
>
> `define-inline` can be used to define inlinable functions, but it's not
> like `defsubst`: the body needs to be changed to look a bit more like
> what a macro would look like.

Right, that's what I thought, but Emacs didn't seem to beep at me when I
used it, so I assumed that it had somehow been changed lately or
something.  I didn't notice that calling

(command-completion-using-modes-p 'eww-form-select (current-buffer))

led to this appearing in my *Messages* buffer:

Compiler-macro error for command-completion-using-modes-p: (wrong-type-argument bufferp (current-buffer))

By the way, changing it back to a defun doesn't seem to clear out the
define-inline-ness?  Hm...  Ah:

(symbol-plist 'command-completion-using-modes-p)
=> (edebug #<marker at 81894 in simple.el> compiler-macro command-completion-using-modes-p--inliner)

Perhaps defun should

(put 'command-completion-using-modes-p 'compiler-macro nil)

?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 496bed5: Change command-completion-using-modes-p to defun

[Stefan Monnier]

> By the way, changing it back to a defun doesn't seem to clear out the
> define-inline-ness?

In the current session, indeed.

That's a general problem with the use of `put`: the info is
associated with the symbol rather than with the actual definition.
Same holds for `gv-expand`, or obsoletion, and many others.

But not for `advertized-calling-convention` because this one uses
a hash-table indexed with the actual function object ;-)


        Stefan