Re: C-u prefix behavior of TAB broken

[Miles Bader <miles@gnu.org>]

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

>>> Can't indent-for-tab-command simply, if there's a prefix argument given,
>>> rigidly indent the following sexp by the change in indentation?  That
>>> doesn't seem hard at all, and matches the behavior I remember (perhaps
>>> there are corner cases where it would differ, but I don't think I know
>>> about them...).
>
>> For `tab-always-indent' t it might make sense to do `beginning-of-line'
>> before the `forward-sexp'.  Otherwise it should work wherever
>> `forward-sexp' behaves reasonably.  BTW, isn't
>
>>    ((memq indent-line-function '(indent-relative indent-relative-maybe))
>>     (funcall indent-line-function))
>>    ;; Indent the line.
>>    (t
>>     (indent-according-to-mode))
>
>> semantically equivalent to
>
>>    (t
>>     (funcall indent-line-function))
>
> No, it's not: check the definition of indent-according-to-mode.
>
> The difference is to distinguish between the case where the user hits
> TAB (and hence wants some kind of effect) as opposed to when
> line-indentation is done non-interactively (e.g. by skeleton).
>
>         Stefan

But indent-according-to-mode is defined like this:

   (defun indent-according-to-mode ()
      ....
     (if (memq indent-line-function
               '(indent-relative indent-relative-maybe))
               ... do some stuff ...
       ;; The normal case.
       (funcall indent-line-function)))

So this:

   (cond
    ...
    ((memq indent-line-function '(indent-relative indent-relative-maybe))
     (funcall indent-line-function))
    ;; Indent the line.
    (t
     (indent-according-to-mode))

Is equivalent to this:

   (cond
    ...
    ((memq indent-line-function '(indent-relative indent-relative-maybe))
     (funcall indent-line-function))
    ;; Indent the line.
    (t
     (if (memq indent-line-function
               '(indent-relative indent-relative-maybe))
               ... do some stuff ...
       ;; The normal case.
       (funcall indent-line-function)))

and since the first cond branch catches the same cases that the `if'
catches, it will always use the `else' part, meaning it's actually:

   (cond
    ...
    ((memq indent-line-function '(indent-relative indent-relative-maybe))
     (funcall indent-line-function))
    ;; Indent the line.
    (t
     (funcall indent-line-function)))

Which of course is the same as:

   (funcall indent-line-function)


-Miles

-- 
Occam's razor split hairs so well, I bought the whole argument!