Re: [RFC PATCH-for-7.0 v2] cocoa: run qemu_init in the main thread

[Peter Maydell]

On Wed, 16 Mar 2022 at 19:29, Peter Maydell <peter.maydell@linaro.org> wrote:
>
> On Wed, 16 Mar 2022 at 17:31, Paolo Bonzini <pbonzini@redhat.com> wrote:
> >
> > On 3/16/22 17:22, Akihiko Odaki wrote:
> > > I was thinking that it may be better to let softmmu/main.c do the
> > > details if it involves the internals of qemu_main() like qemu_main_loop().
> > >
> > > More concretely, softmmu/main.c would provide a function to register a
> > > function pointer to take over the main thread. main() implemented in
> > > softmmu/main.c would call qemu_init(). If a function pointer gets
> > > registered in qemu_init(), it would create a thread for main loop and
> > > call the registered function pointer. Otherwise, it would directly call
> > > qemu_main_loop().
> > >
> > > It would be a semantically appropriate division of ui/cocoa.m and
> > > softmmu/main.c. It would also be beneficial for end-users as it would
> > > also allow to isolate ui/cocoa.m into a separate module when
> > > --enable-modules in the future. (With "In the future", I mean sometime
> > > when we have time to hack Meson build files and some details we cannot
> > > fill by 7.0.)
> >
> > I would like this for 7.1.
> >
> > Basically rename qemu_main_loop to qemu_default_main_loop, and
> > cocoa_display_init would do
> >
> >      qemu_main_loop = qemu_cocoa_main_loop;
> >
> > qemu_cocoa_main_loop would include the bulk of the current main of
> > ui/cocoa.m.  Seems like a good idea.
>
> Speaking of 7.1, is cocoa currently completely broken, ie in need
> of an interim fix for 7.0 ? If so, which of the various patches/approaches
> should it be ?

To answer the first half of my question, yes, the cocoa UI is
currently completely broken as it asserts on startup.

-- PMM