refresh: always breaking '.' register

[Valloric]

I see the comment on line 626 about setting refresh: 'always' and then breaking the dot register. I saw that there was an attempt at fixing this, but the problem still remains even with Vim 7.3.475.

Any clue on when this will be fixed? Is there some sort of workaround?

[Shougo]

Any clue on when this will be fixed? Is there some sort of workaround?

Yes. In my environment, it still not fixed. Can you check your environment?

[Valloric]

For me it's not fixed as well. I'm using Vim 7.3.496.

On Tue, May 1, 2012 at 5:01 PM, Shougo <
reply@reply.github.com

wrote:

Any clue on when this will be fixed? Is there some sort of workaround?

Yes. In my environment, it still not fixed. Can you check your environment?

---

Reply to this email directly or view it on GitHub:
#185 (comment)

[Shougo]

Oh...
I think the bug must be fixed. But I have not much time and technique...

[mattn]

@Valloric please let us known the way to reproduce.

[Valloric]

I've noticed this problem while developing a separate Vim plugin, but it should be reproducible with neo as well. One needs to have Vim >= 7.3.418 and g:neocomplcache_enable_cursor_hold_i = 1. Actually, it might be easier to reproduce if you just comment out the conditions and always return dict.refresh = 'always' (just for testing).

Bottom line, if the completefunc returns a dict with refresh = 'always', then neo will be called on every keystroke. If you then just type through the completions without accepting anything, the '.' register holds bad history. For me, when I type a word that's present in neo's database but ignore the offered completion, the insert history in the '.' register is broken and does not have the whole word stored.

[Valloric]

Here's an example:

1. if "neocomplcache" is stored in neo's database
2. I have g:neocomplcache_enable_cursor_hold_i = 1 and a recent Vim
3. I enter insert mode and type "neocomplcache testing" (NOT accepting the completion offered, just typing)
4. I leave insert mode.
5. press '.' to insert all that text again
6. I get nonsense text instead of "neocomplache testing".

This is not a bug in neo but a bug in Vim; I get the same problem with a different custom completefunc.

[Shougo]

Thanks for your bug report!

[Valloric]

I'll be thanking you if you provide a fix for this in Vim :). I'm new to the Vim codebase and tracking down the root cause of this seems non-trivial. I saw that @mattn tried to fix this once (Vim patch 7.3.404) so maybe with this new data, it will be easier to track down and fix for good.

[mattn]

Sorry for delay. I'll look into it later.

[mattn]

I tested following.

1. added let g:neocomplcache_enable_at_startup = 1 and let g:neocomplcache_enable_cursor_hold_i = 1 into my vimrc
2. started vim and typed neocomplcache testing on insert mode.
3. typed <esc>:reg<cr>

I can see ". neocomplcache testing.

[Valloric]

Like I said, what you type in needs to be already in neo's database. You need to be offered a completion suggestion; the suggestion window needs to appear.

Here's a screencast of the problem: http://dl.dropbox.com/u/2076969/vim_broken_dot_register.mov

[Valloric]

@mattn Do you know when the patch you posted on the vim-dev thread is going to be officially upstreamed to Vim? I've been using Vim with your patch since you posted it and have had zero issues since. It fully fixes the problem I had.

[mattn]

Bram knows the answer. Probably it's in his todo list. He often pick up from the todo list in his priority. If you warry about, try to post 'ping'. Then, the post may change the priority. :)

[Valloric]

I've just pinged the thread on vim-dev. Thanks for writing the patch
and fixing the problem!

On Tue, May 15, 2012 at 6:19 PM, mattn
reply@reply.github.com
wrote:

Bram knows the answer. Probably it's in his todo list. He often pick up from the todo list in his priority. If you warry about, try to post 'ping'. Then, the post may change the priority. :)

---

Reply to this email directly or view it on GitHub:
#185 (comment)

[Shougo]

This problem was fixed in 7.3.561.
Thanks @mattn!