bug#9087: Crash reading from minibuffer with icomplete-mode

[Eli Zaretskii]

> Date: Sat, 07 Jan 2012 18:45:28 +0100
> From: martin rudalics <rudalics@gmx.at>
> CC: monnier@iro.umontreal.ca, jasonr@gnu.org, claudio.bley@gmail.com, 
>  lekktu@gmail.com, 9087@debbugs.gnu.org
> 
>  > Details, please.  What exactly are you suggesting that we do instead
>  > of calling signal_user_input?
> 
> When immediate_quit is set proceed in signal_user_input as if we
> processed C-g in post_character_message.

That means we will call signal_quit that interrupts system calls (not
relevant to this case) and "forcibly complete any deferred messages"
(whatever that means) by calling cancel_all_deferred_msgs, which again
is not relevant.  By contrast, Vquit_flag will _not_ be set to
Vthrow_on_input, which will break throw-on-input and while-no-input.

IOW, all we do when we see C-g is interrupt any system calls and
deliver the C-g character to the main thread.  We do nothing special
besides that, AFAICS.  I don't see how this could help the case in
point.