On 30/09/2016 12:29, Greg Kurz wrote:
> On Fri, 30 Sep 2016 11:13:33 +0200
> Laurent Vivier <lviv...@redhat.com> wrote:
>
>> On 30/09/2016 10:33, Greg Kurz wrote:
>>> On Thu, 29 Sep 2016 19:15:05 +0200
>>> Laurent Vivier <lviv...@redhat.com> wrote:
>>>
>>>> [...]
>>>> static void pci_nop(void)
>>>> {
>>>> - qvirtio_9p_start();
>>>> - qvirtio_9p_stop();
>>>> + QOSState *qs;
>>>> +
>>>> + qs = qvirtio_9p_start();
>>>> + g_assert(qs);
>>>
>>> The appropriate macro to use here is: g_assert_nonnull().
>>
>> OK
>>
>>>
>>> BTW, how can qs be NULL ?
>>
>> we should not know what happens in qtest_pc_boot() (or
>> qtest_spapr_boot(), or qtest_XXX_boot())
>>
>
> What is the point in letting qtest_XXX_boot() return NULL then if
> it is always followed by g_assert() in the test program code ?
>
> I'd rather move the assertion there and document that it cannot
> return NULL, since it is always unrecoverable in the test code.
You're right it looks better as you say. I'm going to change that.
Thanks,
Laurent
