bug#12615: 24.2.50; Non-ignored case in insert-char

[Juri Linkov <juri@jurta.org>]

>>> Some observations: C-x 8 C-m is bound to insert-char, a C function.
>>> And insert-char calls read-char-by-name, which let-binds
>>> completion-ignore-case to t.  The clear intention is that unicode name
>>> searches should always be case insensitive, and this seems always to
>>> be the case if completion-ignore-case is not buffer local.
>>
>> Indeed, a buffer-local setting of completion-ignore-case can
>> bring surprises.  I think the patch below will fix this problem for this
>> particular case.
>
> I just stumbled upon the case where this fix causes the regression:
> typing `C-x 8 RET *acc TAB' results in "[No match]".
>
> I don't know why this case disobeys the let-binding of
> `completion-ignore-case' and whether a simpler fix is possible,
> but at least this patch fixes it for the emacs-24 branch:

Please ignore this patch.  Just search the source tree with
grep "completion-ignore-case t" and see the remaining 100 places
that have exactly the same problem.  Take for example the first grep
hit in bookmark.el.  Evaluate the following in the *scratch* buffer:
(make-local-variable 'completion-ignore-case)
Then type `C-x r b' (`bookmark-jump')
followed by a lower-case letter and TAB.
Completion is not case insensitive.

Maybe to fix all them at once, `read_minibuf' should make
a local variable `completion-ignore-case' in the minibuffer
and copy its value from the original buffer?