OK - sorry, it took an extra day. Here is the new (third, and hopefully last) test case.

1 Steps to Reproduce:

1.1 Start Emacs clean

/Applications/Emacs.app/Contents/MacOS/Emacs -Q

1.1.1 Note: Emacs Version

M-x emacs-version <RET>
GNU Emacs 24.0.95.1 (x86_64-apple-darwin, NS apple-appkit-1038.36) of 2012-04-03 on bob.porkrind.org

1.2 Clone Git HEAD Org-Mode (to get the new commit)

1.3 Load Cloned version of org-mode

Evaluate the following Lisp in *scratch*

(require 'ido)
(ido-mode t)

(setq load-path (cons "~/tmp/new-new-new/org-mode/lisp" load-path))
(setq load-path (cons "~/tmp/new-new-new/org-mode/contrib/lisp" load-path))
(require 'org-install)
(setq org-completion-use-ido t)
(setq org-tag-alist (quote (
                            (:startgroup)
                            ("tag_a" . ?a)
                            ("tag_b" . ?b)
                            (:endgroup)
                            ("tag_c" . ?c)
                            )))

1.4 Open the following org-mode file

* Header 1
** Header 1.1
* Header 2                                                                      :Personal:

1.4.1 Note: Org-Version

M-x org-version <RET>
release_7.8.07-201-g72d9
Org-mode version 7.8.08 (release_7.8.07.201.g72d9)

1.5 Point should be on the first asterisk for “Header 1”

1.6 Type C-c C-q to bring up Tag selection

The expected result is the fast tag selection window-split; this works

1.7 Type <TAB> to enter free tag selection

The expected result is that the Minibuffer prompts with Tag: and the ido completion (this works)

1.8 Type t to begin to ido select one of the tags

The expected result is the ido completion narrowing the choices; the result

Error in post-command-hook (ido-exhibit): (wrong-type-argument sequencep 97)

Note that M-x toggle-debug-on-error doesn’t show anything different than the above.


On Apr 2, 2012, at 11:50 AM, Bastien wrote:

Hi Mike,

Mike McLean <mike.mclean@pobox.com> writes:

The dangers of minimal testing setups is rearing its ugly head.

You bet :)

I can
confirm that your latest patch does fix the minimal example that I
provided. It does not, sadly, fix a similar case where the
org-tag-alist variable includes the :startgroup and :endgroup tokens.
If you can my most recent steps to reproduce, step 1.3, and add those
tokens, the error is "Wrong type argument: stringp, :endgroup"

Can you resubmit the full test case?

Thank you for you patience :)

Thanks for your collaboration :)

--
Bastien