Closed bretthoerner closed 10 years ago
Brett Hoerner notifications@github.com writes:
This must be a recent change, because it was fine a few weeks ago. I verified by removing autopair and everything works. Details:
I can't reproduce this. What emacs version are you using? Can you reproduce by starting emacs with the -Q switch and enabling just autopair and nothing else?
You're right, if I use -Q and just load autopair I can't reproduce. But if I just drop autopair from my config it also fixes the issue. Any tips on how to debug further? I feel like it must be autopair + something else?
Hmmm, must be something in isearch-mode-map, or maybe autopair is kicking in in isearch-mode which it shouldn't. You can bisect your config, commenting parts of it, but keeping autopair, to find out the other culprit. Send me your config, or paste it to a github gist, and I'll have a quick look. Also grep your config for "isearch" or "autopair". Another option, since you say it broke some weeks ago, is to look up changes to your config that you made at the time.
Meanwhile, closing the issue, but feel free to comment further.
This is my config: https://github.com/bretthoerner/dotfiles-emacs
Nothing very weird with isearch, it must be something in MELPA...
I'll try the bisect thing soon, there's a ton of stuff, though. :(
I have the same problem and I suspect it is some change in Emacs trunk which triggers this bug.
I'm using Emacs 24.3.50.4 checked out yesterday from git master.
My recipe:
I'm using el-get for package management and am at 7e19b0499ce17d2b60553ebe81c2bac94c7fcd55 currently.
Probably a dynamic binding problem? I don't know how autopair-fallback works when it should pass the key on to the "original" binding...
When in isearch "mode"(?), DEL should run the `isearch-delete-char' function but somehow it doesn't.
Furthermore, in older Emacs, DEL ran "the command isearch-other-control-char, which is an
alias for isearch-other-meta-char' in
isearch.el'."
Looking through Emacs' git log lisp/isearch.el, I guess http://git.savannah.gnu.org/cgit/emacs.git/commit/?id=aa04ac2c6fb020706c95d335f9a556b7db97d28b might be a first suspicious change causing the current behavior.
Great call, I also run Emacs trunk.
I've now reproduced this using an emacs build of 27/10.
I seem to have solved it by binding exactly the same keys as isearch-mode-map
does for the backspace and return events. I never seem to understand the synonyms for these keys, but apparently the [backspace]
cannot be found in isearch-mode-map
while \177
can. Normal operation of autopair-backspace
and autopair-newline
seems to be unchanged, but this should be tested also on older emacsen.
This must be a recent change, because it was fine a few weeks ago. I verified by removing autopair and everything works. Details:
I'm using autopair 20131009.1902 from MELPA, and now keys like
backspace
andenter
(which are bound toautopair
keys) inisearch-mode-map
break expected behavior.For example,
C-s hellp
. Oops.backspace
(so I can typeo
) now actually deletes the character under my point (not in theisearch
minibuffer). Enter is the same, it actually inserts a return in my buffer instead of leavingisearch
like it usually does.