evil-forward-section-begin misbehaves

[Ambrevar]

This bug is linked with evil-mode, so I first opened an issue there:
emacs-evil/evil#844

foo

# header 0

foo

# header 1

foo

# header 2

foo

# header 3

foo

Pressing ]] on header 1 moves the point at the end of the body of header 2. Same with header 2 and 3.
Only the last 3 headers are concerned. The rest is fine.

[[ (evil-backward-section-begin) does not display the issue.

The Evil maintainer suggests that Markdown-mode might be to blame.

[jrblevin]

Let me describe the Markdown mode defun movement functions. That might clear things up?

If the point is in the whitespace between the second "foo" and "# header 1", then you are at the end of the "# header 0" defun. The docstring for end-of-defun says "Move forward to next end of defun." So, the next end of defun is the end of the "# header 1" defun.

Markdown mode's defun functions behave like those in c-mode and emacs-lisp-mode, for example. When the point is at whitespace after a function, it is considered end of defun and end-of-defun moves over the following function.

So, with your test document if you move to the beginning of the buffer and press C-M-e repeatedly, Markdown mode moves the point just after "foo" each time. So, it's consistent in markdown-mode by itself (not different behavior for the last three sections). It seems like perhaps an inconsistency in evil mode?

Edit: I meant to write C-M-e.

[Ambrevar]

Let me quote @wasamasa:

This is the result of another oddity with thingatpt.el. Movement for ]] is done in terms of evil-defun (which is like a regular defun except that movement signals whether it's been successful or not) and generally done with evil-forward-beginning. In the case you describe it first goes to the end of the evil-defun at point, then moves to the end of the next one with forward-thing, then back to the beginning of the next one with beginning-of-thing. The last step fails for your test data, to reproduce go to the last newline and evaluate M-: (bounds-of-thing-at-point 'evil-defun).

I'm not sure whether evil can do anything about this, other than trying to move back until it's possible to navigate by a defun. This might be the fault of markdown-mode if their functions for beginning-of-defun and end-of-defun are suboptimal. Can you reproduce this problem with other modes defining those?

The thing is, other modes are fine.

[jrblevin]

Thanks. I did read the original issue, but after re-reading it more carefully I now understand that the issue is when the point is at the end of the buffer and I can reproduce it (independently of evil-mode).

[jrblevin]

I think that should fix your issue, but please let me know if not.