Re: [Qemu-devel] [PATCH v15 08/23] monitor: Let generated code validate arguments

[Markus Armbruster]

Marc-André Lureau <address@hidden> writes:

> On Thu, Apr 28, 2016 at 4:09 PM, Markus Armbruster <address@hidden> wrote:
>>
>> * device_add, qmp_capabilities
>>
>>   Not QAPIfied, need everything.
>
> Some related commits from https://github.com/elmarco/qemu/commits/qapi
>
>
> https://github.com/elmarco/qemu/commit/e6828a0a926e20d3ee91ea956152b610ca30d2b8
>
> https://github.com/elmarco/qemu/commit/ba871db1c8d0edd0cbe7be716952f72c9df519a2
>
>>
>> * netdev_add, query-qmp-schema
>>
>>   Need .args_type because of 'gen': false.
>>
>> * client_migrate_info, dump-guest-memory, query-dump, getfd, closefd,
>>   add-fd, remove-fd, query-fdsets, migrate-set-capabilities
>>
>>   Define .params and/or .help.  Does this make any sense?
>>
>> A comment explaining which members need to be set would be nice.
>>
>> Have you checked Marc-André's work for overlap?  Cc'ing him.
>
> yeah, it's conflicting with my work. Unfortunately, since I have been
> asked to wait since a long time, I am no longer on top of things, nor
> what I need to wait for. I have the feeling we really want
> qmp_dispatch() in monitor.c:
> https://github.com/elmarco/qemu/commit/af4158993088ee2bbf9087c440990880d7943eb1.
>
> Eric, I have been waiting for your series to settle for getting back
> to qapi, but it keeps growing ;) I hope you are familiar with the work
> I did in the branch so we don't duplicate efforts.

I'd very much prefer to give precedence to Marc-André's solution.

I think this patch is in this series just because the previous patch
notes a redundancy, and promises its cleanup.  What about this:

* We shelve this patch for now, and s/a later patch/a future patch/ in
  the previous patch's commit message.

* We continue to flush Eric's QAPI queue.  If we come to a point where
  we really want this patch, we insert it, and apologize to Marc-André.
  But hopefully we come to a point where Marc-André's work fits first!
  We'll refer to this patch when reviewing his solution, to make sure
  we're not losing anything.

Okay?