* python/qemu/ too picky with QMP responses?
@ 2026-11-18 7:35 Markus Armbruster
2026-11-18 18:21 ` Markus Armbruster
0 siblings, 1 reply; 2+ messages in thread
From: Markus Armbruster @ 2026-11-18 7:35 UTC (permalink / raw)
To: John Snow, Cleber Rosa; +Cc: Marc-André Lureau, qemu-devel
Marc-André and I broke python/qemu/utils/qom_common.py in commit
0eee31cf1085 (qapi/qom: add qapi-type to qom-list, qom-list-properties,
qom-list-get):
$ bld/run qom-list -s ../images/test-qmp /chardevs
Traceback (most recent call last):
File "/work/armbru/qemu/bld/pyvenv/bin/qom-list", line 8, in <module>
sys.exit(QOMList.entry_point())
~~~~~~~~~~~~~~~~~~~^^
File "/work/armbru/qemu/python/qemu/utils/qom_common.py", line 230, in entry_point
return cls.command_runner(args)
~~~~~~~~~~~~~~~~~~^^^^^^
File "/work/armbru/qemu/python/qemu/utils/qom_common.py", line 215, in command_runner
return cmd.run()
~~~~~~~^^
File "/work/armbru/qemu/python/qemu/utils/qom.py", line 179, in run
rsp = self.qom_list(self.path)
File "/work/armbru/qemu/python/qemu/utils/qom_common.py", line 192, in qom_list
return [ObjectPropertyInfo.make(x) for x in rsp]
~~~~~~~~~~~~~~~~~~~~~~~^^^
File "/work/armbru/qemu/python/qemu/utils/qom_common.py", line 52, in make
assert value.keys() <= {'name', 'type', 'description', 'default-value'}
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError
Needs fixing for 11.2. I can take care of it, but I have questions.
Here's what breaks:
@classmethod
def make(cls, value: Dict[str, Any]) -> 'ObjectPropertyInfo':
"""
Build an ObjectPropertyInfo from a Dict with an unknown shape.
"""
assert value.keys() >= {'name', 'type'}
---> assert value.keys() <= {'name', 'type', 'description', 'default-value'}
return cls(value['name'], value['type'],
value.get('description'),
value.get('default-value'))
The first assertion checks we got the members we actually require.
The second assertion checks we got no members we don't know.
The commit I quoted added member @qapi-type without updating the second
assertion.
Adding it to the set of members we know would be easy enough. But is it
appropriate? docs/devel/qapi-code-gen.rst section "Compatibility
considerations":
Clients are expected to ignore [members] they don't know.
Shouldn't we do that here?
Any objections to me hunting down and deleting such assertions?
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: python/qemu/ too picky with QMP responses?
2026-11-18 7:35 python/qemu/ too picky with QMP responses? Markus Armbruster
@ 2026-11-18 18:21 ` Markus Armbruster
0 siblings, 0 replies; 2+ messages in thread
From: Markus Armbruster @ 2026-11-18 18:21 UTC (permalink / raw)
To: John Snow, Cleber Rosa; +Cc: Marc-André Lureau, qemu-devel
Markus Armbruster <armbru@pond.sub.org> writes:
> Marc-André and I broke python/qemu/utils/qom_common.py in commit
> 0eee31cf1085 (qapi/qom: add qapi-type to qom-list, qom-list-properties,
> qom-list-get):
[...]
> The first assertion checks we got the members we actually require.
>
> The second assertion checks we got no members we don't know.
>
> The commit I quoted added member @qapi-type without updating the second
> assertion.
>
> Adding it to the set of members we know would be easy enough. But is it
> appropriate? docs/devel/qapi-code-gen.rst section "Compatibility
> considerations":
>
> Clients are expected to ignore [members] they don't know.
>
> Shouldn't we do that here?
>
> Any objections to me hunting down and deleting such assertions?
Just sent:
Subject: [PATCH] python/qemu: Don't choke on unknown members in QMP replies
Date: Wed, 18 Nov 2026 19:20:05 +0100
Message-ID: <20261118182005.1066151-1-armbru@redhat.com>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-11-18 18:22 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-18 7:35 python/qemu/ too picky with QMP responses? Markus Armbruster
2026-11-18 18:21 ` Markus Armbruster
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.