feat(engine)!: syncing published templates #1222
Conversation
…h-template-as-substate
…h-template-as-substate # Conflicts: # dan_layer/consensus/src/consensus_constants.rs # dan_layer/consensus_tests/src/support/harness.rs # dan_layer/engine/src/transaction/error.rs # dan_layer/engine/src/transaction/processor.rs
… template only when block is committed
…h-template-as-substate
…h-template-as-substate # Conflicts: # applications/tari_dan_app_utilities/src/base_layer_scanner.rs # dan_layer/engine_types/src/substate.rs
…h-template-as-substate
…tate' into feature/published-template-sync
…tate' into feature/published-template-sync
…tate' into feature/publish-template-as-substate
…tate' into feature/published-template-sync
…hed-template-sync # Conflicts: # dan_layer/consensus/src/hotstuff/worker.rs
| info!(target: LOG_TARGET, "Template manager initializing"); | ||
| // Template manager | ||
| let (tx_template_sync, rx_template_sync) = broadcast::channel::<TemplateSyncRequest>(100); |
There was a problem hiding this comment.
use an unbounded mpsc channel. Broadcast channel is not appropriate as it is typically used for events and if there are more than 100 requests some of them may start to be dropped (broadcast channel lagging)
| use tari_engine_types::substate::SubstateId; | ||
|
|
||
| /// General trait for different state stores to do operations on substates. | ||
| pub trait SubstateStore { |
There was a problem hiding this comment.
This doesnt seem right. Firstly, this is in the SQLite implementation of the state store which will be a problem when we switch to rocksDB. Templates can always be version 0, so not sure this is necessary, I'll need to look more closely at the rest of the code. There is currently no special process for upgrading the template code, you simply publish a new template at v0.
| } | ||
| if let Some(TemplateResult::Template(fetched_template)) = self.fetch_template(address).optional()? { | ||
| template = Some(fetched_template); | ||
| } |
| template = Some(fetched_template); | ||
| } | ||
| // sleeping here to not overload the local database while waiting for the template to be ready | ||
| thread::sleep(std::time::Duration::from_millis(100)); |
There was a problem hiding this comment.
Not ideal. Will think about how we should handle pending templates in a transaction.
Description
Motivation and Context
How Has This Been Tested?
What process can a PR reviewer use to test or verify this change?
Breaking Changes