Re: [PATCH] coroutine: add libucontext as external library

[Peter Maydell]

On Tue, 9 Mar 2021 at 09:59, Joelle van Dyne <j@getutm.app> wrote:
>
> On Tue, Mar 9, 2021 at 1:36 AM Daniel P. Berrangé <berrange@redhat.com> wrote:
> > The README here:
> >
> >   https://github.com/kaniini/libucontext
> >
> > indicates that it is intentionally limiting what registers it saves
> > and restores, explicitly excluding FPU.
> >
> > Peter & Paolo expressed concern about this, indicating FPU reg support
> > was a requirement for QEMU:
> >
> >    https://lists.gnu.org/archive/html/qemu-devel/2021-01/msg05525.html
> >
> Does it make a difference if this is provided as an option and not as
> a replacement? Would it make sense to add some warning at configure
> time? Right now none of the concurrency backends are supported on iOS
> and it's possible support will go away on macOS as well in the future.
> QEMU would not be able to run at all.

We don't currently support iOS; if we add it we definitely don't
want to add it as a "we know this has buggy coroutine support"
target, because that's a path to weird hard-to-diagnose bugs.
Right now macOS works fine without libucontext; if we ever do need
to use libucontext on macOS in future we'd want to get the FPU
support etc fixed first.

Adding FPU handling for aarch64 to libucontext should not be too
difficult (because the FPU is guaranteed to exist you don't need to
do the hardware capability detection the README talks about).

thanks
-- PMM