Re: [PATCH 2/2] auto-t: work around race between ANQP and dbus Connect()

[Denis Kenzior <denkenz@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1840 bytes --]

Hi James,

On 5/27/20 3:45 PM, James Prestwood wrote:
> ANQP is performed as the scan results come in but scanning is still
> signaled as done regardless of if ANQP has finished. Autoconnect
> will wait for ANQP to complete, but as far as an explicit DBus call
> we are just assuming ANQP finished. For now, we can work around this
> in the autotest and wait a bit before issuing Connect(). The real
> fix would either be to delay setting scanning as done, or have
> network/station wait for ANQP to complete before checking the
> network settings.
> ---
>   autotests/testHotspot/hotspot_test.py | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/autotests/testHotspot/hotspot_test.py b/autotests/testHotspot/hotspot_test.py
> index 0f18fae1..389038fd 100644
> --- a/autotests/testHotspot/hotspot_test.py
> +++ b/autotests/testHotspot/hotspot_test.py
> @@ -11,6 +11,7 @@ from iwd import PSKAgent
>   from iwd import NetworkType
>   from hostapd import HostapdCLI
>   import testutil
> +from time import sleep
>   
>   class Test(unittest.TestCase):
>   
> @@ -33,6 +34,8 @@ class Test(unittest.TestCase):
>           condition = 'not obj.scanning'
>           wd.wait_for_object_condition(device, condition)
>   
> +        sleep(1)
> +

So you know this, but the issue with sleep is that you're completely at 
the mercy of the scheduler and the speed of qemu.  This might work for 
you most of the time, but then it will occasionally fail.  Or it might 
fail reliably on a slower machine running qemu.

In general I think we need to go in and eliminate all sleeps in the test 
scripts.  Or at least as many as we can.

>           ordered_network = device.get_ordered_network('Hotspot')
>   
>           self.assertEqual(ordered_network.type, NetworkType.eap)
> 

Regards,
-Denis