Re: [Qemu-devel] [PATCH 27/28] sysemu: Move the VMChangeStateEntry typedef to qemu/typedefs.h

[Markus Armbruster]

Markus Armbruster <address@hidden> writes:

> Paolo Bonzini <address@hidden> writes:
>
>> On 26/07/19 14:05, Markus Armbruster wrote:
>>> +typedef struct VMChangeStateEntry VMChangeStateEntry;
>>>  typedef struct VMStateDescription VMStateDescription;
>>>  
>>
>> This is a bit borderline; I'd rather split sysemu/sysemu.h, e.g. adding
>> sysemu/runstate.h that would have VMChangeStateEntry functions.  If
>> there aren't many conflicts, perhaps you can drop this patch?
>
> Without it, the next one will be ineffective.

Actually, "ineffective" is an exaggeration, it's not *that* bad.  Two
headers are affected:

                            Before 27  After 27+28  Just 28
qapi/qapi-types-run-state.h      5500         4400     5000
sysemu/vmstate-notify.h             0          180     1000

The numbers are .o depending on the header in my "build everything"
tree, out of 6600 total (not counting tests and objects that don't
depend on qemu/osdep.h).

Mainly because hw/virtio/virtio.h (760 .o) and hw/scsi/scsi.h (180 .o)
need to include sysemu/vmstate-notify.h to get the VMChangeStateEntry
typedef.  Four more headers need it as well, but they are small
potatoes.

Splitting sysemu/runstate.h off sysemu/sysemu.h makes sense anyway, but
won't improve these numbers.

I'd keep PATCH 27.

[...]