On Tue, 10 Sep 2024, Carlos Eduardo Gallo Filho <gcarlos@xxxxxxxxxxx> wrote:
> +/* Try to init a framebuffer without setting its format */
> +static void drm_test_framebuffer_init_bad_format(struct kunit *test)
> +{
> + struct drm_framebuffer_test_priv *priv = test->priv;
> + struct drm_device *dev = &priv->dev;
> + struct drm_framebuffer fb1 = { .dev = dev, .format = NULL };
> + struct drm_framebuffer_funcs funcs = { };
> + int ret;
> +
> + /* Fails if fb.format isn't set */
> + ret = drm_framebuffer_init(dev, &fb1, &funcs);
Not only does this fail, it spits a WARN_ON_ONCE() in dmesg. Which in
turn gets flagged as a failure in the test in our CI.
What's the policy with kunit tests causing warnings? I think it's
reasonable for any CI to flag dmesg warnings. We shouldn't be hitting
those. Filtering the warnigs is a tricky business.
BR,
Jani.
> + KUNIT_EXPECT_EQ(test, ret, -EINVAL);
> +}
--
Jani Nikula, Intel
