[PATCH v6 12/12] sandbox: add test of wdt-uclass' watchdog_reset()

[Rasmus Villemoes]

On 31/08/2021 10.17, Stefan Roese wrote:
> Hi Rasmus,
> 
> I've pulled this patchset now into next [1] and have run it through
> CI via Azure. Here an error occurs:
> 
> https://dev.azure.com/sr0718/u-boot/_build/results?buildId=109&view=logs&j=50449d1b-398e-53ae-48fa-6bf338edeb51&t=97605dd2-f5a5-5dd7-2118-315ffdc8bcd6&l=533
> 
> 
> Could you please take a look at this?

I'm pretty confident that has nothing to do with my patches, but is a
broken (or, to put it more gently, fragile) test.

It does

// fetch the emulated device's current base_time value, setting it to 0
        old_base_time = sandbox_i2c_rtc_get_set_base_time(emul, 0);

// fetch the emuluated device's current base_time value, don't (-1) set
// a new value, check that we got 0
        ut_asserteq(0, sandbox_i2c_rtc_get_set_base_time(emul, -1));

// call the device's ->reset method
        /* Resetting the RTC should put he base time back to normal */
        ut_assertok(dm_rtc_reset(dev));
// fetch the new base_time value, without altering it
        base_time = sandbox_i2c_rtc_get_set_base_time(emul, -1);
// and check that the base time was put back to "normal"
        ut_asserteq(old_base_time, base_time);

The thing is, the ->reset method does

static void reset_time(struct udevice *dev)
{
        struct sandbox_i2c_rtc_plat_data *plat = dev_get_plat(dev);
        struct rtc_time now;

        os_localtime(&now);
        plat->base_time = rtc_mktime(&now);

It's inevitable that this will cause occasional spurious failures. I can
trivially reproduce it with

=> while ut dm rtc_reset ; do echo . ; done

it fails after a few screenfuls of successes.

Rasmus