Remove hardcoded notifications capability

[pascalwengerter]

Just found out that the ownCloud web frontend incorrectly checks for notifications since the notifications capability in Reva, for historical reasons and contrary to the behavior in other capability cases, returns an array with a "disable" string (

reva/tests/oc-integration-tests/local/frontend.toml

Line 110 in a8c6140

endpoints = ["disable"]

).

To make it consistent we shouldn't return anything here until Reva/oCIS is capable of handling notifications.

Glad to take care of this myself when the time comes but also don't want to cause a breaking change in the Semver for this minor nitpicky issue

[labkode]

Hi @pascalwengerter, what do you need? To omit the value or to set it to something else?

[pascalwengerter]

Hi @pascalwengerter, what do you need? To omit the value or to set it to something else?

I'd vote in favor of omitting it since from my understanding that's how non-boolean values are being treated so far (?)

[labkode]

@pascalwengerter from the code looks like this capability is not a boolean.

   if h.c.Capabilities.Notifications == nil {
                h.c.Capabilities.Notifications = &data.CapabilitiesNotifications{}
        }
        if h.c.Capabilities.Notifications.Endpoints == nil {
                h.c.Capabilities.Notifications.Endpoints = []string{"list", "get", "delete"}
        }

The default would be an empty array.

[pascalwengerter]

@pascalwengerter from the code looks like this capability is not a boolean.

   if h.c.Capabilities.Notifications == nil {
                h.c.Capabilities.Notifications = &data.CapabilitiesNotifications{}
        }
        if h.c.Capabilities.Notifications.Endpoints == nil {
                h.c.Capabilities.Notifications.Endpoints = []string{"list", "get", "delete"}
        }

The default would be an empty array.

Fine with me, too. Checking for an array length feels way cleaner than doing if capabilities.notifications.endpoints[0] != "disable" (which even contains a grammatical error)

[labkode]

@pascalwengerter great, so we need the oc QA team to evaluate if changing the test to use an empty array will not break other things, I'll create a PR to check what goes wrong.