On 12/06/2015 10:56 PM, Peter Xu wrote:
> v5 changes:
> - patch 1
>   - comment English fix [Fam]
> - patch 2
>   - pass has_detach=true always in hmp_dump_guest_memory [Paolo]
> - patch 3
>   - always use local_err and error_propagate() when need to check
>     the result [Fam]
> - patch 8
>   - add "DumpQueryResult" in DUMP_COMPLETED event [Eric]
>     (since DumpQueryResult is introduced in patch 10, so doing it in
>     patch 10 for convenience. Please let me know if I should not do
>     this, e.g., if patch re-ordering is required)

All patches should build in isolation.  It looks like you met that goal
(you introduce 'DUMP_COMPLETED' event in 8 without a 'result' member,
then modify it in 10), so that it at least builds.  But it results in
churn, in that you have multiple different definitions of
'DUMP_COMPLETED' over the life of the series.

It's not a requirement to rework things since each step builds, but if I
were writing the series, I do find it conceptually easier to supply
patches in an order that minimizes churn (the first patch that
introduces a type uses its final form, rather than going through several
iterations of that type).  So on that grounds, introducing
DumpQueryResult as a separate patch, before either DUMP_COMPLETED or
query-dump, might be easier to review, if there is a reason for a v6 spin.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org