bug#21072: 24.5; inconsistent behaviour of `C-M-h (mark-defun)' in Emacs Lisp

[Marcin Borkowski <mbork@mbork.pl>]

Hi everyone,

it's me again.

I looked at the tests of `mark-defun' for Python, and they did shed some
light on the weird implementation.  In short, `mark-defun' first jumps
to the beginning and the to the end of the current defun, checks whether
the point is now before or after the initial position, and if before,
back up and jump to the end first and to the beginning next.  This might
actually happen, since e.g. in Python, when point is on a "def"
beginning a function definition, `beginning-of-defun' goes to the start
of the _preceding_ defun.  In fact, I think this is a bug in
beginning-of-defun (or more precisely, its Python version), and
consequently I think the weird code in `mark-defun' is a workaround for
a deeper bug.

I can now do one of a few things.  (Not ATM, I'll have to take my time.)

1. Fix the docs so that #21072 is no longer a bug (i.e., change the docs
so that there's no discrepancy between them and code).

2a. Just fix the problem reported in #21072, hoping that it doesn't break
anything else.

2b. Write a few tests for `mark-defun',_then_ fix the problem, and check
whether these tests pass.  Of course, all these tests would be for Elisp
(and maybe for C and/or JavaScript).

3. Try to come up with a better `mark-defun', still given the
limitations of `beginning-of-defun', e.g., make `mark-defun' use the
numeric prefix argument in a reasonable way (i.e., what Drew proposed).

4. Report a bug in `beginning-of-defun', and hope someone (probably not
me, I do not know enough about ppss now and do not have time to learn
that now), and then write a simpler `mark-defun' from scratch.

WDYT?

Best,
Marcin



On 2016-06-05, at 08:30, Marcin Borkowski <mbork@mbork.pl> wrote:

> On 2016-05-07, at 08:47, Eli Zaretskii <eliz@gnu.org> wrote:
>
>>> From: Marcin Borkowski <mbork@mbork.pl>
>>> Cc: rfflrccrd@gmail.com, 21072@debbugs.gnu.org
>>> Date: Sat, 07 May 2016 05:47:18 +0200
>>> 
>>> I'm not sure what to do next with this bug.  I tried with the
>>> abovementioned short Elisp file and failed.  Since no-one said that the
>>> current behavior is wrong, maybe I'll try to study the code of
>>> `mark-defun' and modify the docstring/manual accordingly.
>>
>> I'd suggest to change the doc string according to what I recommended a
>> few messages ago.  I think that would be good enough.
>
> Hi Eli, hi all,
>
> and sorry for replying after such a long time (again) - I'm afraid
> I cannot help this, my time has become extremely limited recently.
>
> I studied the code of mark-defun, and it seems that the reason for
> #21072 is quite simple.  I enclose a patch where the two (seemingly)
> offending lines are commented; if this is acceptable, of course I'll
> prepare a proper patch.  (And while at that, I'll propose replacing
> `(and transient-mark-mode mark-active)' with `(use-region-p)'.)
>
> The problem is, I'm not sure whether this change won't break anything.
> The only tests I found that deal with `mark-defun' are in
> python-tests.el, and my version passes all three of them.
>
>> Thanks.
>
> Best,


-- 
Marcin Borkowski
http://octd.wmi.amu.edu.pl/en/Marcin_Borkowski
Faculty of Mathematics and Computer Science
Adam Mickiewicz University