Re: [PULL 14/15] qdev: Base object creation on QDict rather than QemuOpts

[Markus Armbruster]

Peter Maydell <peter.maydell@linaro.org> writes:

> On Fri, 15 Oct 2021 at 16:01, Kevin Wolf <kwolf@redhat.com> wrote:
>> QDicts are both what QMP natively uses and what the keyval parser
>> produces. Going through QemuOpts isn't useful for either one, so switch
>> the main device creation function to QDicts. By sharing more code with
>> the -object/object-add code path, we can even reduce the code size a
>> bit.
>>
>> This commit doesn't remove the detour through QemuOpts from any code
>> path yet, but it allows the following commits to do so.
>>
>> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
>> Message-Id: <20211008133442.141332-15-kwolf@redhat.com>
>> Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
>> Tested-by: Peter Krempa <pkrempa@redhat.com>
>> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
>
> Hi; we discovered via a report on IRC this this commit broke
> handling of "array properties", of which one example is:
> qemu-system-x86_64 -netdev user,id=a -device rocker,len-ports=1,ports[0]=a
>
> This used to work, and now fails with
>  qemu-system-x86_64: -device rocker,len-ports=1,ports[0]=a: Property
> 'rocker.ports[0]' not found
>
> I think this happens because array properties have the
> requirement that the len-foo property is set first before
> any of the foo[n] properties can be set. In the old code
> I guess we used to set properties from the command line
> in the order they were specified, whereas in the new code
> we end up in object_set_properties_from_qdict() which
> tries to set them in whatever order the qdict hash table
> provides them, which turns out to be the wrong one :-(
>
> Any suggestions for how to address this ?

My initial (knee-jerk) reaction to breaking array properties: Faster,
Pussycat! Kill! Kill!

Back to serious: replace the implementation of QDict so it iterates in
order?