Potential problem with Platform networking

[cmelchior]

This seems to be the root cause of some of the problems we are seeing with our Sync Tests on Jenkins.

It can be reproduced in the following way:

1. Checkout `cm/fix-progress-test, from Fix race condition in ProgressListenerTest #1633
2. Modify AsymmetricSyncTests setup to enable platform networking:

    @BeforeTest
    fun setup() {
        app = TestApp(this::class.simpleName, appName = TEST_APP_FLEX, builder = { builder ->
            builder.usePlatformNetworking(false)
            builder
        })
        val (email, password) = TestHelper.randomEmail() to "password1234"
        val user = runBlocking {
            app.createUserAndLogIn(email, password)
        }
        config = SyncConfiguration.Builder(
            user,
            schema = FLEXIBLE_SYNC_SCHEMA
        ).errorHandler { session, error ->
            println(error)
        }.initialSubscriptions {
            it.query<DeviceParent>().subscribe()
        }.build()
        realm = Realm.open(config)
    }

1. Start the local test server using ./start_local_test_server.sh
2. Run AsymmetricSyncTests.asymmetricSchema() from inside Android Studio.
3. This will now fail with a timeout for uploadAllLocalChanges().
4. Rerunning the test will now succeed. Rerunning step 3 and 4 with platform networking disabled will work for this test.
5. Running the AsymmetricSyncTest.insert() from a fresh server with Platform networking disabled will instead result in an AutoClientResetFailed(1028) error being thrown. More details are reported here: Update server version on all entries in the sync history during client reset with recovery realm-core#7291 (review)

So right now it is a bit unclear if this is a problem with the server, the Sync Client, Client Reset, Platform Networking or all of the above. But something is clearly not working as intended

[sync-by-unito]

➤ cmelchior commented:

This is currently blocking a release from our main branch. We might need to try to roll back commits until it will start working again, but that will probably be quite disruptive. But the Core/Server team is currently looking at logs, so hopefully they can figure out what is wrong.

[nirinchev]

This was fixed in #1645.