On 29.09.2023 02:47, Elliott Mitchell wrote:
> On Mon, Sep 25, 2023 at 08:27:31AM +0200, Jan Beulich wrote:
>> On 22.09.2023 17:42, Elliott Mitchell wrote:
>>> On Fri, Sep 22, 2023 at 10:21:21AM +0200, Jan Beulich wrote:
On 21.09.2023 18:18, Elliott Mitchell wrote:
> As such these
On Mon, Sep 25, 2023 at 08:27:31AM +0200, Jan Beulich wrote:
> On 22.09.2023 17:42, Elliott Mitchell wrote:
> > On Fri, Sep 22, 2023 at 10:21:21AM +0200, Jan Beulich wrote:
> >> On 21.09.2023 18:18, Elliott Mitchell wrote:
> >>> As such these incomplete definitions should be
> >>> in xen.h next
On 22.09.2023 17:42, Elliott Mitchell wrote:
> On Fri, Sep 22, 2023 at 10:21:21AM +0200, Jan Beulich wrote:
>> On 21.09.2023 18:18, Elliott Mitchell wrote:
>>> As such these incomplete definitions should be
>>> in xen.h next to their hypercalls, rather than spread all over.
>>
>> Perhaps
On Fri, Sep 22, 2023 at 10:21:21AM +0200, Jan Beulich wrote:
> On 21.09.2023 18:18, Elliott Mitchell wrote:
> > Hypercall wrappers need the incomplete type definitions. Only when the
> > actual structure needed.
>
> While in the first sentence "only" looks to be missing, I can't really
> make
On 21.09.2023 18:18, Elliott Mitchell wrote:
> Hypercall wrappers need the incomplete type definitions. Only when the
> actual structure needed.
While in the first sentence "only" looks to be missing, I can't really
make sense of the second (without implying what I think you mean).
> As such
Hypercall wrappers need the incomplete type definitions. Only when the
actual structure needed. As such these incomplete definitions should be
in xen.h next to their hypercalls, rather than spread all over.
trap_info_t is particularly notable since even though the hypercall is
x86-only, the