I found a case where the code in question is needed, which none of the
suggested patches handle correctly.

Steps to repeat:

    (setq ns-alternate-modifier nil)

    Press left CMD-ALT-9

    An unmodified Emacs replies "s-]" is not bound. (This assumes a Swedish
keyboard layout, other layouts would yield a different character, but the
principle is the same).

    With either of the two patches, Emacs respond with "s-9" is not bound,
which isn't correct.

Unfortunately, I don't know how to distinguish between the cases where we
need to strip away modifiers (C-s-a) and when we shouldn't, so I'm leaving
this open for now.

    -- Anders Lindgren


On Wed, Dec 30, 2015 at 9:50 AM, Anders Lindgren <andlind@gmail.com> wrote:

> Hi,
>
> I'm looking into a key binding bug on OS X reported multiple times (19977,
> 21330, 21551). Two different patches have been submitted.
>
> The original code looks like:
>
>       if (is_left_key)
>         {
>           emacs_event->modifiers |= parse_solitary_modifier
>             (ns_command_modifier);
>
>           /* if super (default), take input manager's word so things like
>              dvorak / qwerty layout work */
>           if (EQ (ns_command_modifier, Qsuper)
>               && !fnKeysym
>               && [[theEvent characters] length] != 0)
>             {
>               /* XXX: the code we get will be unshifted, so if we have
>                  a shift modifier, must convert ourselves */
>               if (!(flags & NSShiftKeyMask))
>                 code = [[theEvent characters] characterAtIndex: 0];
>
> One of the patches simply removes the `if (EQ(...))' statement. The other
> modifies the code to strip away modifiers.
>
> First question: What is the code in the `if (EQ(...))' supposed to do? In
> other words, what will stop working if it is removed?
>
> Second question: if it is needed for the LEFT command key, should the
> corresponding code be added for the RIGHT?
>
> I've tested removing the `if' and Emacs still seems to be working as
> intended, both with a normal key layout and when Dvorak is used. If no one
> comes up with a reason to keep the code, I will remove it.
>
> Third question: Does anybody know of a good way to automatically test
> things like this? What I'm looking for is a way to send keystrokes like
> Cmd-Alt-a to Emacs, that way it could be possible to write tests ensuring
> that things like this don't break in the future.
>
>     -- Anders Lindgren
>
>