Open slumber opened 1 year ago
cc @muharem
The issue we were observing was that with the new logic from https://github.com/paritytech/cumulus/pull/2658/, https://github.com/paritytech/cumulus/pull/2501, https://github.com/paritytech/cumulus/pull/2551 , that integration tests for all runtimes began to fail when integrating the new Aura hook, and the root cause of this failure was that AuraExt
wasn't initialized before set_validation_data
was called.
The change in behavior from these PRs is that ParachainSystem::set_validation_data
can now require some other pallets to initialized properly (i.e. when calling the configured ConsensusHook
).
I see that the logic in decl_test_parachains
is directly invoking on_initialize
and set_validation_data
, so I see two solutions going forward:
AuraExt
to the pallets_extra
field for the macro for each runtime and ensure it is initialized
- Alter the macro to initialize all pallets
If that isn't done currently, it sounds like the tests are not being setup correctly.
Tests in question are
parachains/integration-tests
. @rphmeier suggested the issue is related to https://github.com/paritytech/cumulus/blob/31b1469d830aaa7b0441d4d60f0c0c17be0cf114/xcm/xcm-emulator/src/lib.rs#L417-L420For instance, explicit panic in
aura-ext
pallet on initialization isn't triggered. Related to #2658