Catching EnsureOlmSessionsForDevicesAction errors #4229
Conversation
…cryption flow - we currently can't do much but log here as we've asynchronously start the fallback flow, catching the error at least stops a hard crash
| @@ -146,29 +146,36 @@ internal class EventDecryptor @Inject constructor( | |||
|
|
|||
| // offload this from crypto thread (?) | |||
| cryptoCoroutineScope.launch(coroutineDispatchers.computation) { | |||
| val ensured = ensureOlmSessionsForDevicesAction.handle(mapOf(senderId to listOf(deviceInfo)), force = true) | |||
| runCatching { ensureOlmSessionsForDevicesAction.handle(mapOf(senderId to listOf(deviceInfo)), force = true) }.fold( | |||
| onSuccess = { sendDummyToDevice(ensured = it, deviceInfo, senderId) }, | |||
There was a problem hiding this comment.
extracted the logic out to its own function to avoid too many levels of indentation
There was a problem hiding this comment.
OK.
Could you also update sendToDeviceTask.execute(sendToDeviceParams) and use executeRetry (3 retries?)
ALso in the EnsureOlmSessionsForDevicesAction, we should probably executeWith retry the claim oneTimeKeysForUsersDeviceTask.execute(claimParams)
There was a problem hiding this comment.
LGTM, maybe @BillCarsonFr can have a look too.
Can you add a small file for the changelog please?
| withContext(coroutineDispatchers.io) { | ||
| val sendToDeviceParams = SendToDeviceTask.Params(EventType.ENCRYPTED, sendToDeviceMap) | ||
| try { | ||
| sendToDeviceTask.executeRetry(sendToDeviceParams, remainingRetry = SEND_TO_DEVICE_RETRY_COUNT) |
There was a problem hiding this comment.
sendToDeviceTask is now executed with the ability to retry 3 times
| @@ -72,7 +74,7 @@ internal class EnsureOlmSessionsForDevicesAction @Inject constructor( | |||
| Timber.i("## CRYPTO | claimOneTimeKeysForUsersDevices() : $usersDevicesToClaim") | |||
|
|
|||
| val claimParams = ClaimOneTimeKeysForUsersDeviceTask.Params(usersDevicesToClaim) | |||
| val oneTimeKeys = oneTimeKeysForUsersDeviceTask.execute(claimParams) | |||
| val oneTimeKeys = oneTimeKeysForUsersDeviceTask.executeRetry(claimParams, remainingRetry = ONE_TIME_KEYS_RETRY_COUNT) | |||
There was a problem hiding this comment.
oneTimeKeysForUsersDeviceTask is now executed with the ability to retry 3 times
bmarty
deleted the
feature/adm/decrypt-event-dummy-keys-fallback-crash
branch
Fixes #3608
When failing to decrypt a
MXCRYPTO_ALGORITHM_OLMevent due toBAD_ENCRYPTED_MESSAGEthe app can crash if theensureOlmSessionsForDevicesActionlogic fails, which can happen if theclaimOneTimeKeysForUsersDeviceshttp request fails due to being offline.sendToDeviceTask&oneTimeKeysForUsersDeviceTaskexecutions to retry 3 timesThe error handling logic is asynchronous executed meaning the main decryption flow isn't able to react to any further errors.
I've naively fixed this crash by simply catching the source of the error, similar to how the surrounding logic behaves.