Re: [PATCH] On the nasty "ghost key" problem on NS

[Po Lu]

Kai Ma <justksqsf@gmail.com> writes:

>  On Nov 4, 2022, at 23:09, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>
>  I'm glad we found a way to make the code work, apparently, but
>  Here we need a comment explaining why we do this gymnastic of
>  `safe_call` + `inhibit_quit` + `waiting_for_input`.
>
>  It's very unusual to have to do that.
>
> Thanks for the pointer, this indeed turned out unnecessary (and dangerous).
>
> A tester informed me of a problem in the v3 patch:
>
>   safe_call (0, Qns_in_echo_area)
>
> is incorrect. The 0 should be 1, or nargs will be -1 for funcall_general.  
> This will cause an error to be signaled, which explains why 
> `waiting_for_input` has to be masked.
>
> [ I guess we could add an assertion that numargs >= 0 in funcall_general or
>   somewhere else? ]
>
> This patch should be correct even without the ugly `waiting_for_input` hack.
> I’ve been running patched Emacs for some time without issues.
>
> With the current understanding of the bug, I guess the comment line could be
>   /* Protect against throw-on-input. */

I will install a change using internal_catch_all instead, if Stefan is
fine with that.  By doing so, you get all of the benefits of safe_call
and inhibit_quit, and it also protects against faulty code that changes
the value of inhibit-quit inside the Lisp function.