Re: [PATCH] ui/cocoa: Do not raise keys before QEMU resigns active

From: Akihiko Odaki <akihiko.odaki@gmail.com>
To: Gerd Hoffmann <kraxel@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>,
	qemu Developers <qemu-devel@nongnu.org>
Subject: Re: [PATCH] ui/cocoa: Do not raise keys before QEMU resigns active
Date: Mon, 15 Mar 2021 19:15:26 +0900	[thread overview]
Message-ID: <CAMVc7JWu-Hp11X48Fe71A5dYcHkn0ce3wRr27mqyj9gPcJt3Lw@mail.gmail.com> (raw)
In-Reply-To: <20210315073233.edxuqhu7m2lmxgzg@sirius.home.kraxel.org>

2021年3月15日(月) 16:32 Gerd Hoffmann <kraxel@redhat.com>:
>
> On Sun, Mar 14, 2021 at 04:01:47PM +0900, Akihiko Odaki wrote:
> > ui/cocoa used to raise all keys before it resigns active to prevent a
> > stuck key problem caused by key up events it does not see while it is
> > inactive. The problem is solved by checking -[NSEvent modifierFlags] in
> > commit 6d73bb643aa725348aabe6a885ac5fb0b7f70252, which is better
> > because it handles the case that key *down* events are missed while it
> > is inactive.
>
> Well, I think it is a good idea to virtually lift all keys when the
> application goes inactive.  Does this cause any actual problems?

No, but ui/cocoa already has so many states and I don't think there is
a room for extra complexity, especially when considering the code
checking -[NSEvent modifierFlags]. Keyboard event management concerns
iothread mutex, different handling of modifier and normal keys and
some special keyboard shortcuts. -[QemuCocoaView raiseAllKeys]
introduces exceptional behaviors to them just to raise normal keys,
something ui/cocoa didn't before introducing QemuKbdState.

Regards,
Akihiko Odaki

>
> Worst case should be that we send an extra keyup + keydown if the
> qemu goes through a active -> inactive -> active cycle while a modifier
> key is down, which you probably wouldn't even notice unless you log all
> key events inside the guest.
>
> take care,
>   Gerd
>