fold issue

[datanoise]

When switching to a previous buffer with CtrlPBuffer, this buffer is displayed with all folds closed. This is quite annoying since I lose track where I was editing previously.

[kien]

Does it happen if you comment out line 731 and 735 in the autoload/ctrlp.vim file? Also try with line 1362, separately.

[datanoise]

Nope. Commenting lines 731 and 735 doesn't help, as well as line 1362.

It seems like switching buffers make Vim to reevaluate modeline for some reason.

[kien]

Yep, I also get this just using :bnext, :bprev or <c-^> with files that have folds defined in modeline. So it's not a problem with ctrlp. It's just Vim.

[datanoise]

I don't see this issue when using :bn, :bp, <c-^> or any other standard buffer switching commands. No issue with using BufExplorer or Command-T either. I would file a bug report to Bram if that was a standard behavior a long time ago :)

[kien]

Weird, I definitely get it with this modeline for my plugin files:

" vim:fen:fdm=marker:fmr={{{,}}}:fdl=0:fdc=1:ts=2:sw=2:sts=2

And this in vimrc:

se foldmethod=marker
se foldlevel=1
se foldnestmax=10
se nofoldenable

If I'm outside a fold, <c-^> away and back will close all folds.

[datanoise]

Hm, strange. I've just tested your modeline and I don't see that behavior when switching back and forth with <c-^>. Must be something to do with your setup. I do see that when I'm using CtrlPBuffer though.

[kien]

OK, I tried again with --noplugin, <c-^> definitely closes all the folds, even if the cursor's in a fold when switching away:

vim --noplugin -u .vimrc

.vimrc

se nocp
"runtime! plugin/ctrlp.vim

With 2 buffers: buffertag.vim and dir.vim from autoload/ctrlp/.

Vim 7.3.401

[kien]

Alright, I searched around a bit to make sure I wasn't crazy or missed something. Someone had the same question on StackOverflow with using NERDTree: all folds close when switching to a different file and back

http://stackoverflow.com/questions/7221318/vim-nerdtree-folding-can-it-remember-the-state-of-folds

So if your tests are correct, cmd-t and bufexplorer probably do something to override this default behaviour of Vim.

[datanoise]

This is really weird. Actually, just tried that with Command-T and in some cases I see the same issue, but I don't see it when I switch with <c-^> or BufExplorer.

And I'm using the same revision number 7.3.401. That must be indeed a Vim's bug. I will try to debug it if the time permits.

[datanoise]

I think I have found the issue. CtrlPBuffer uses :edit command to switch to a buffer, which makes Vim to reevaluate modeline. Please consider this patch:

diff --git a/autoload/ctrlp.vim b/autoload/ctrlp.vim
index a95df44..a942e69 100644
--- a/autoload/ctrlp.vim
+++ b/autoload/ctrlp.vim
@@ -734,8 +734,10 @@ fu! ctrlp#acceptfile(mode, matchstr, ...)
        el
                " Determine the command to use
                let cmd = md == 't' || s:splitwin == 1 ? 'tabe'
-                       \ : md == 'h' || s:splitwin == 2 ? 'new'
-                       \ : md == 'v' || s:splitwin == 3 ? 'vne' : ctrlp#normcmd('e')
+                       \ : md == 'h' || s:splitwin == 2 ? (s:itemtype == 1 ? 'sb' : 'new')
+                       \ : md == 'v' || s:splitwin == 3 ? (s:itemtype == 1 ? 'vert sb' : 'vne')
+                       \ : md == 'e' && s:itemtype == 1 ? 'b'
+                       \ : ctrlp#normcmd('e')
                " Open new window/buffer
                cal call('s:openfile', a:0 ? [cmd, filpath, ' +'.a:1] : [cmd, filpath])
        en

That would work only if hidden option is set, otherwise this patch has no effect.

[kien]

I don't see it when I switch with <c-^>

And as you said later:

uses :edit command to switch to a buffer, which makes Vim to reevaluate modeline

But from the description in Vim's help:

CTRL-^         edit Nth alternate file (equivalent to ":e #N")

Seems like there's something else in your config that changes the behavior of <c-^> or :edit #N.

But set hidden does help when using the :buffer commands. So there's our use case. However, this patch doesn't account for commands after : (fail to open the buf), so I'll need to work it into the code and will commit later along with the others.

Thanks for the help!

[datanoise]

Ah, you are right. I do have <c-^> redefined. I completely forgot about it.

Thanks for looking into that issue anyway.