Re: [Qemu-devel] [PATCH RFC v2 41/47] qom: Don't use 'gen': false for qom-get, qom-set, object-add

[Markus Armbruster <armbru@redhat.com>]

Eric Blake <eblake@redhat.com> writes:

> On 07/01/2015 02:22 PM, Markus Armbruster wrote:
>> With the previous commit, the generated marshalers just work, and save
>> us a bit of handwritten code.
>> 
>
> Generated code grows, because you are now generating the hook :)
>
>  qmp-commands.h |    6 ++
>  qmp-marshal.c  |  144 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 150 insertions(+)
>
>
>> Signed-off-by: Markus Armbruster <armbru@redhat.com>
>> ---
>>  include/monitor/monitor.h |  3 ---
>>  qapi-schema.json          |  9 +++------
>>  qmp-commands.hx           |  6 +++---
>>  qmp.c                     | 20 +++++++-------------
>>  scripts/qapi.py           |  1 +
>>  5 files changed, 14 insertions(+), 25 deletions(-)
>> 
>
>> +++ b/qmp.c
>> @@ -229,11 +229,9 @@ ObjectPropertyInfoList *qmp_qom_list(const char *path, Error **errp)
>>  }
>>  
>>  /* FIXME: teach qapi about how to pass through Visitors */
>> -void qmp_qom_set(QDict *qdict, QObject **ret, Error **errp)
>> +void qmp_qom_set(const char *path, const char *property, QObject *value,
>> +                 Error **errp)
>
> The FIXME seems stale now.

I'm not 100% sure I understand the intent of the FIXME, but my best
guess is "do what this patch does".  I'll drop it.

>> -void qmp_object_add(QDict *qdict, QObject **ret, Error **errp)
>> +void qmp_object_add(const char *type, const char *id,
>> +                    bool has_props, QObject *props, Error **errp)
>>  {
>> -    const char *type = qdict_get_str(qdict, "qom-type");
>> -    const char *id = qdict_get_str(qdict, "id");
>> -    QObject *props = qdict_get(qdict, "props");
>>      const QDict *pdict = NULL;
>>      QmpInputVisitor *qiv;
>>  
>
> Continuing:
>
>     if (props) {
>         pdict = qobject_to_qdict(props);
>         if (!pdict) {
>             error_setg(errp, QERR_INVALID_PARAMETER_TYPE, "props", "dict");
>             return;
>         }
>     }
>
> I know that we guarantee that all pointers are initialized to NULL in
> our marshaler, so this is correct; but wouldn't it be more idiomatic to
> write 'if (has_props)' as the condition?

We do similar things elsewhere.  Moreover:

> (And it would make a nice followup project for someone to figure out how
> to get rid of have_FOO arguments for strings and objects where NULL is a
> nice witness; it is only integers and booleans that require them - but
> doing that will touch a lot of the tree, and is a series of its own)

Shouldn't be hard, and I really want it done.

Will change all the places that test has_PTR to just PTR.  More reason
to simply test PTR now.

> What I pointed out is minor, so you can fix it and add:
> Reviewed-by: Eric Blake <eblake@redhat.com>

Thanks!