return none if subscription returns early

[gregdhill]

Signed-off-by: Gregory Hill gregorydhill@outlook.com

[niklasad1]

The proper action should be to indicate that the subscription is dropped and need to restarted because after it returns None it shouldn't be polled anymore because the channel has been terminated.

If you poll if after None it will panic

[gregdhill]

Which is exactly why I would expect it to return None, to indicate the stream has closed (much like with StreamExt).

[niklasad1]

Alright, that makes sense to me but perhaps we should document it because it's not obvious to me ^^

[ascjones]

LGTM

[ascjones]

I think our CI is broken, all test runs are hanging now, possibly due to recent client changes.

[niklasad1]

I get this locally:

[2021-03-16T12:52:42Z ERROR sc_finality_grandpa] GRANDPA voter error: Safety("`voter_commands_rx` was closed.")
[2021-03-16T12:52:42Z ERROR sc_service::task_manager] Essential task `grandpa-voter` failed. Shutting down service.
test frame::balances::tests::test_state_read_free_balance ... ok
test tests::test_getting_read_proof ... ok
test tests::test_iter ... ok
test tests::test_tx_transfer_balance ... ok
test tests::test_insert_key ... ok
test frame::balances::tests::test_transfer_subscription ... ok
[2021-03-16T12:52:48Z ERROR sc_finality_grandpa] GRANDPA voter error: Safety("`voter_commands_rx` was closed.")
[2021-03-16T12:52:48Z ERROR sc_service::task_manager] Essential task `grandpa-voter` failed. Shutting down service.
test frame::balances::tests::test_basic_transfer ... ok
test frame::sudo::tests::test_sudo_unchecked_weight ... ok
test frame::sudo::tests::test_sudo ... ok
[2021-03-16T12:53:12Z ERROR sc_finality_grandpa] GRANDPA voter error: Client(DatabaseError(DatabaseError(Custom { kind: Other, error: Error { message: "IO error: No such file or directoryWhile open a file for appending: /tmp/subxt-.QfQWaR1EUaUu/db/000033.log: No such file or directory" } })))
[2021-03-16T12:53:12Z ERROR sc_service::task_manager] Essential task `grandpa-voter` failed. Shutting down service.
test frame::balances::tests::test_transfer_error has been running for over 60 seconds

[gregdhill]

I don't think the CI issue is related to the client changes as it was working before on this pipeline, also worked locally for me with these latest changes. Is it possible it has something to do with the environment? There's an annotation on the last successful run that ubuntu-latest was scheduled to be upgraded.

[ascjones]

@gregdhill please merge master, it should fix CI