Switch from the prototype of duplicating the SQL filter as a runtime
and keeping them in sync on each event that might invalidate the current
filtered entries to a simpler approach of requesting the filter again
and doing the diff to detect the new changes.
Also add a new reset API to model the new entries design requirements.
The new approach shows less corner-case to handle and follows one source
of truth concept making debugging and future maintenance easier.
Other changes
- Fix pending mocking to work with multiple calls
- Refactor tests to account for the new changes
Updates status-desktop #12120
This commit introduces the first steps towards implementing a session-based activity API to support dynamic updates of the current visualized filter in the wallet activity service. This change is necessary to move away from static paginated filtering, which was previously done in SQL, to a more dynamic approach that can handle updates in real-time.
The main changes include:
- Add basic `EventActivitySessionUpdated` support for pending transactions.
- Added a `TODO.md` file outlining the plan and requirements for dynamic activity updates.
- New session-related API to the `activity.Service`
- `session.go` contains the logic for session management and event processing related to activity updates.
- Add test case for incremental filter updates.
The commit also includes:
- various other minor changes and refactoring to support the new session-based approach.
- Deprecation notices added to the `api.go` file for methods that are no longer used by the status-desktop application.
- Clarification comments added to the `scheduler.go` file regarding replacement policies.
Updates: #12120
ghstack-source-id: a61ef74184
Pull Request resolved: https://github.com/status-im/status-go/pull/4480
The reading of the amount for pending transactions was done in the same
way as for transfers table. However, the transfers table has a string
hex representation of the amount, while the pending transactions table
has a binary representation of the amount (*big.Int). This was
triggering the not int warning and value was missing.
Updates status-desktop #12120
In case both to/from addresses are present in the list we were using
the same logic as for transfers. However, this doesn't make sense given
that we can have only one entry in pending activity.
The following cases are still covered
- When the receiver is in addresses we get received
- When both receiver and sender are in the list will get sent
- When the sender is on the list we will get sent
Updates status-desktop #12120
keep restarting on error non-stop, removed some tests accordingly.
Fixed a flaky test for loadBlocksAndTransfers command.
Added tests for async.AtomicGroup.
Made tranfersCommand FiniteCommandWithErrorCounter to prevent infinite
restart.
- favourite column removed from the saved_addresses table
- favourite property removed from the SavedAddress struct
- ens name removed from the primary key, the primary key now is composed of address and is_test columns
- ens parameter removed from wakuext_deleteSavedAddress
- wallet_getSavedAddresses moved to wakuext_getSavedAddresses (to keep them all in a single place)
- saved addresses related endpoints removed from the wallet service, even they logically belong there, a reason for that
is avoiding emitting sync message if one uses calls from the wallet service, while that's not the case in ext service. Once
we refactor this and introduce devices syncing mechanism in the wallet service, we should not only these but other wallet
related endpoints move there (removed: wallet_getSavedAddresses, wallet_addSavedAddress and wallet_deleteSavedAddress).
Affected area:
Saved addresses
Reading the Nonce from the local cache may be incorrect if the tx is made out of the Status app or
if Status app sends a tx prepared by the dapp (via WalletConnect). A submitted tx with a wrong Nonce
results in a failing tx, that's why we need to read the Nonce from the network.
mock it in tests.
Made a configurable timeout interval for Commander interface.
Added tests to verify loadBlocksAndTransfers command is stopped
correctly on max errors limit reached
proper handling of errors and commands restart.
Now:
- Infinite commands started only once and never restarted, stoped on
context.Done.
- Finite commands are joined into AtomicGroup to stop the rest in the
group in case one command fails. Otherwise other commands in the group
will continue running and the failed command is not retried to
restart. Fixed goroutine leakage in case of failure of some commands
`getLogs` for multiple accounts simultaneously. For now only used for
new transfers detection. Detection of `new` transfers has been changed,
now they are searched from head and forward. Previously they were
searched from last scanned block forward.
- `WalletConnectTransfer` identified as a new transfer type
- Wallet-related endpoints that logically belong to the wallet moved from the wallet connect service
- Wallet connect service now receives `transfer.TransactionManager` instead of `transactions.Transactor`
- Deadlock issue when trying to send the tx with the wrong nonce fixed
Add new APIs to track if valid pairings are available to be used
by application not to run WalletConnect SDK if not needed.
Closes status-desktop: #12794
Refactored transfers loading to reduce blockchain RPC requests (getBaseFee, getTransaction,
getTransactionReceipt) by reusing preloaded transaction and block fee.
Split extraction of subtransaction from logs and from ETH transfer into
different methods.
Refactored log_parser to extract sender and receiver addresses
uniformly for different transfer types.
Replaced info logs with debug where needed.
closes#4221
Stefan
Roman Volosovskyi
Dario Gabriel Lipicar
Anthony Laibe
Ivan Belyakov
Siddarth Kumar
Richard Ramos
Sale Djenic
Anthony Laibe
Cuteivist
Stefan Dunca