Hi,

I agree w/ Erik - however, I think the test has a bug, if the installations are not directly associated with a Variant. Should be easy to fix.

BTW. qmx, you mentioned the "setDeviceType()", but that is really just plain text metadata, like here:
https://github.com/aerogear/aerogear-push-helloworld/blob/master/ios/HelloWorld/AGAppDelegate.m#L106 

The validator is triggered by the actual variant type (installation.getVariant().getType())

-Matthias



On Fri, Feb 13, 2015 at 8:17 AM, Erik Jan de Wit <edewit@redhat.com> wrote:
Hi,

Don’t know if it’s really a bug, but in order to validate that a device token is valid we need to know what variant type it belongs to. It makes no sense for a installation to be persisted without a Variant, because then you cannot send any messages to that device. So if you try to persist a installation without a variant the device token is not valid seems to be to be a valid statement.

This was introduced when we added cordova, there it’s easy to make a mistake and have the wrong variantID / secret in your settings. Resulting in a iOS device registering under an android variant. With the deviceToken validation this can no longer happen.

Does this make sense?

Cheers,
        Erik Jan

On 13 Feb,2015, at 2:02 , Douglas Campos <qmx@qmx.me> wrote:

> Howdy!
>
> I was doing the migration work and found out something funny:
>
> For Installations, we have `installation.setDeviceType()`, but during
> Installation `persist()`, the DeviceTokenValidator is called, which in
> turn looks for a Variant.
>
> So after fiddling with some orm settings suddenly a bunch of tests
> started to break, which led me to this snippet[1]:
>
> // disabled
> Installation android3 = new Installation();
> android3.setAlias("foo@bar.org");
> android3.setDeviceToken("543234234890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890");
> android3.setDeviceType("Android Tablet");
> android3.setEnabled(false);
>
> installationDao.create(android3);
>
> See, our tests are creating an installation without associating it to a
> variant first, then it blows up here[2]:
>
> // DeviceTokenValidator#isValid()
> if (installation.getVariant() == null || installation.getVariant().getType() == null || deviceToken == null) {
>    return false;
> }
>
> This is smelling like a bug to me - kinda like the validator wasn't
> actually running during tests. Am I crazy?
>
> [1]:https://github.com/aerogear/aerogear-unifiedpush-server/blob/master/model/jpa/src/test/java/org/jboss/aerogear/unifiedpush/jpa/InstallationDaoTest.java#L133-L140
> [2]:https://github.com/aerogear/aerogear-unifiedpush-server/blob/master/model/api/src/main/java/org/jboss/aerogear/unifiedpush/api/validation/DeviceTokenValidator.java#L60
>
> --
> qmx
> _______________________________________________
> aerogear-dev mailing list
> aerogear-dev@lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/aerogear-dev


_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev



--
Matthias Wessendorf

blog: http://matthiaswessendorf.wordpress.com/
sessions: http://www.slideshare.net/mwessendorf
twitter: http://twitter.com/mwessendorf