On 1/31/19 3:55 AM, Kevin Wolf wrote:
> Am 31.01.2019 um 03:03 hat John Snow geschrieben:
>> On 1/30/19 12:27 PM, Kevin Wolf wrote:
>>> Am 14.01.2019 um 17:25 hat Eric Blake geschrieben:
>>>> From: John Snow <js...@redhat.com>
>>>>
>>>> New interface, new smoke test.
>>>>
>>>> Signed-off-by: John Snow <js...@redhat.com>
>>>> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsement...@virtuozzo.com>
>>>> Reviewed-by: Eric Blake <ebl...@redhat.com>
>>>> Message-Id: <20181221093529.23855-12-js...@redhat.com>
>>>> [eblake: fix last-minute change to echo text]
>>>> Signed-off-by: Eric Blake <ebl...@redhat.com>
>>>
>>> This test fails for me (with Python 3) because I get a different dict
>>> order.
>>>
>>> Probably related to the OrderedDict that qmp_log() uses, so log() uses
>>> sort_keys=False and the unordered dicts that are contained in the outer
>>> OrderedDict stay unordered.
>>>
>>> Kevin
>>>
>>
>> OK, the problem is that ordered_kwargs() is not recursively ordering
>> those kwargs beneath list entries, so transactions are avoiding the sort.
>>
>> Patch coming up, thank you for the report and my apologies for breaking
>> iotests so often this cycle.
>>
>> (While I'm here, how do you configure iotests to use your python3
>> binary? I tried at configure time but that breaks the build for me with
>> some "magic number" errors. I could toy with it after by editing
>> common.env, but should the build work with python3?)
>
> Yes, just at configure time:
>
> ./configure --target-list=x86_64-softmmu --python=/usr/bin/python3
>
> I'm building and running tests in-tree, if that makes a difference.
>
> But maybe it's actually not so bad if at least one of us still tests
> with Python 2 as long as we haven't officially removed support for
> that...
>
> Kevin
>
jhuston@probe (master) ~/s/q/b/git> make
GEN qapi-gen
Traceback (most recent call last):
File "/home/bos/jhuston/src/qemu/scripts/qapi-gen.py", line 8, in <module>
import argparse
ImportError: bad magic number in 'argparse': b'\x03\xf3\r\n'
make: *** [Makefile:540: qapi-gen-timestamp] Error 1
Ah, I see ... there are stale .pyc and __pycache__ files hiding around
that break things when you switch from 2 to 3 and `make distclean`
doesn't clear those out.
Probably ought to.
