Re: org-store-link and help for a key

[Ihor Radchenko]

> Such behavior exists for ages, at least since 8.2.10 till current master.

I can reproduce this.

The reason of error is simple: help buffers for key combinations where
not considered when writing help: link type.
Current code does not even store the link to such help buffers correctly
- when the key sequence has more than a single key only the first key
will be stored (tested in <f1> k <f1> k help buffer).
The code to open help: link type also does not know that help: link can
point to anything other than function or variable.

I can see two solutions:

1. Avoid storing links to help buffers describing key sequences. Show
   error when trying to store such link instead
2. Write proper handlers for storing and following links to help buffers
   other than just function/variable help (in addition to key sequence
   help buffers, org-store/follow-link also fails to work with
   describe-face buffers at least)
   To do this, one need to change org-link--open-help and org-store-link
   (see "(eq major-mode 'help-mode)" line).

Patches are welcome.

P.S. Would it be cleaner to have separate :store link properties for
standard link types instead of hard-coding them into org-store-link?

Best,
Ihor


Maxim Nikulin <manikulin@gmail.com> writes:

> Either I missed something obvious or org-store-link from a help page for 
> a keystroke works incorrectly
>
> Let's try to save a link to the *keystroke* help for e.g. org-insert-link:
>
> 1. Open help page with =C-h k C-c C-l=
> 2. Switch to the help buffer =C-x o=
> 3. Store link =M-x org-store-link RET=
> 4. Switch back to the org file =C-x o=
> 5. Insert link =C-c C-l RET RET=
>
> I get =[[help:C-c]]= and an error if I try to follow the link with =C-c C-o=
>
>      user-error: Unknown function or variable: C-c
>
> If I ask a help page for the *function*
>
> 1b, =C-h f org-insert-link RET=
>
> I get the working link =[[help:org-inser-link]]=. I expect the similar 
> result for the keystroke help link or maybe something special that leads 
> to the description of the current keystroke binding.
>
> Such behavior exists for ages, at least since 8.2.10 till current master.