With the recent introduction of pending states, the community requests
logic became more complex. This commit simplifies the flow and
appropriately delegates logic to its corresponding abstraction levels:
messenger, manager and community. Additionally, it eliminates
redundancies in notifications and request-saving mechanism.
- use protected topics for communities
- associate chats to pubsub topics and populate these depending if the chat belongs to a community or not
- mailserver functions should be aware of pubsub topics
- generate private key for pubsub topic protection when creating a community
- add shard cluster and index to communities
- setup shards for existing communities
- distribute pubsubtopic password
- fix: do not send the requests to join and cancel in the protected topic
- fix: undefined shard values for backward compatibility
- refactor: use shard message in protobuffers
- share requests to join with new privileged roles during reevaluating member role
- share requests to join with new members, joined the community as TOKEN_MASTER, ADMIN
- share requests to join revealed addresses to ADMINS and TOKEN_MASTERS
- refactor common test functionality to make them more predictable
- removed unused CommunityToken protobuf
This commit skips deleting a keystore file for account which are not marked as fully operable
and also skips deleting master key keystore file if a keypair is non operable. It also takes into
consideration the operable property of an account when adding/updating/handling a keypair/keycard.
u3
When we received a remove event from a private group chat out of order,
the chat would not be created.
This was causing some issues if later on we received the previous event.
This commit changes the behavior so that a chat is created.
When we switched to marking removed accounts, the handler remained unchanged. It kept
checking for the account existence in the received `protobuf.SyncKeypair` message, instead
of checking a removed flag.
This is a bigger change in how community membership requests are handled
among admins, token masters, owners, and control nodes.
Prior to this commit, all privileged users, also known as
`EventSenders`, were able to accept and reject community membership
requests and those changes would be applied by all users.
This commit changes this behaviour such that:
1. EventSenders can make a decision (accept, reject), but merely forward
their decision to the control node, which ultimately has to confirm
it
2. EventSenders are no longer removing or adding members to and from
communities
3. When an eventsender signaled a decision, the membership request will
enter a pending state (acceptedPending or rejectedPending)
4. Once a decision was made by one eventsender, no other eventsender can
override that decision
This implementation is covered with a bunch of tests:
- Ensure that decision made by event sender is shared with other event
senders
- `testAcceptMemberRequestToJoinResponseSharedWithOtherEventSenders()`
- `testRejectMemberRequestToJoinResponseSharedWithOtherEventSenders()`
- Ensure memebrship request stays pending, until control node has
confirmed decision by event senders
- `testAcceptMemberRequestToJoinNotConfirmedByControlNode()`
- `testRejectMemberRequestToJoinNotConfirmedByControlNode()`
- Ensure that decision made by event sender cannot be overriden by other
event senders
- `testEventSenderCannotOverrideRequestToJoinState()`
These test cases live in three test suites for different event sender
types respectively
- `OwnerWithoutCommunityKeyCommunityEventsSuite`
- `TokenMasterCommunityEventsSuite`
- `AdminCommunityEventsSuite`
In addition to the changes mentioned above, there's also a smaller
changes that ensures membership requests to *not* attached revealed wallet
addresses when the requests are sent to event senders (in addition to
control nodes).
Requests send to a control node will still include revealed addresses as
the control node needs them to verify token permissions.
This commit does not yet handle the case of event senders attempting to
kick and ban members.
Similar to accepting and rejecting membership requests, kicking and
banning need a new pending state. However, we don't track such state in
local databases yet so those two cases will be handled in future commit
to not have this commit grow larger.
- Fixed redundant permissions check. If community is set to auto-accept,
then permissions would be checked twice, in
`HandleCommunityRequestToJoin` and `AcceptRequestToJoinCommunity`.
Mitigated it by returning from `HandleCommunityRequestToJoin` immediately
in case of auto-accept.
- Extracted `accountsSatisfyPermissionsToJoin` to remove code
duplication and simplify the logic.
* feat: don't remove sent mutual state messages on accepting a CR
* fix: don't send mutual state message for a new contact
* chore: move mutual state messages to `addContact`
* fix: use one chat for mutual state messages and contact requests
* fix: change `added` mutual state updatede messages to `accepted`
* feat: Use different content type for each mutual state event system message
* chore: use constants for mutual event system messages test, review fixes
* chore: fix tests related to local contacts map
This is the second step of improvements over keypairs/keycards/accounts.
- `SyncKeycardAction` protobuf removed
- `SyncKeypair` protobuf is used for syncing keycards state as well as for all
keycards related changes
- `last_update_clock` column removed from `keypairs` table cause as well as
for accounts, any keycard related change is actually a change made on a related
keypair, thus a keypair's clock keeps the clock of the last change
- `position` column added to `keypairs` table, needed to display keycards in
the same order accross devices
This is the first step of improvements over keypairs/keycards/accounts.
- `SyncKeypairFull` protobuf removed
- `SyncKeypair` protobuf is used for syncing all but the watch only accounts
- `SyncAccount` is used only for syncing watch only accounts
- related keycards are synced together with a keypair
- on any keypair change (either it's just a keypair name or any change made over an
account which belongs to that keypair) entire keypair is synced including related keycards
- on any watch only account related change, that account is synced with all its details
fix flaky test: TestRetrieveBlockedContact
resolve conflict when rebase origin/develop
Feat/sync activity center notification (#3581)
* feat: sync activity center notification
* add test
* fix lint issue
* fix failed test
* addressed feedback from sale
* fix failed test
* addressed feedback from ilmotta
go generate ./protocol/migrations/sqlite/...
feat: add updated_at for syncing activity center notification
* feat: add mutual state update system message
* feat: send mutual state update on accepting CR
* feat: send mutual state update when removing a contact
* fix: don't send MutualStateUpdateMessage over wire
* fix: mutual state update message text fixed
* fix: new clock to ensure system message after CR and add chat to the response
* feat: add AC notification for contact removal
* feat: replace "sent" mutual state system message with "added"
Fixes an issue where if a group chat was first received from a non-contact, and later received from a contact, it still wouldn't save it as active.
That's because we checked if we were **newly** added instead of just if we were added. That meant that in the case I described above, the chat would then never have the chance to be set active.
* fix(community): stop re-joining comm when receiving a sync community msg
Fixes an issue with chats being reset. Since joining a community resaves the chats with the synced default value, it resets the sate of the chats, losing the unread messages, the muted state and more.
The solution is to block the re-joining of the community. In the case of the sync, we catch that error and just continue on.
* fix(import): fix HandleImport not saving the chat
Doesn't change much, but it could have caused issues in the future, so since we might have modified the chat, we make sure to save them
Also adds a test
* fix tests
- old `accounts` table is moved/mapped to `keypairs` and `keypairs_accounts`
- `keycards` table has foreign key which refers to `keypairs.key_uid`
- `Keypair` introduced as a new type
- api endpoints updated according to this change
* fix: create a CR on contact sync with received CR state
* fix: create a CR on contact sync with sent CR state
* Review fixes
* Fix: ignore own contact installation or syncing
* sync local deleted messages
* rebase
* add REPLACE
* fix lint
* defer rows.Close() / rename function
* add local pair test
* replace unused clock with _
* fix(mentions): deleting or editing a mention should remove the mention
* test(edit): add a test for mentions in edits
* test(delete): add test for deleting a message with a mention
There were a couple of issues on how we handle pinned messages:
1) Clock of the message was only checked when saving, meaning that the
client would receive potentially updates that were not to be
processed.
2) We relied on the client to generate a notification for a pinned
message by sending a normal message through the wire. This PR changes
the behavior so that the notification is generated locally, either on
response to a network event or client event.
3) When deleting a message, we pull all the replies/pinned notifications
and send them over to the client so they know that those messages
needs updating.
* fix: Contact requests flows
fix: pending CR notification
fix: use CR as message with provided text
* fix: Remove legacy default contact request
* fix: Add test case sending CR after removal from contacts
* fix: refactor contact request tests to have common steps
* fix: activate chat on sender side after receiveing the CR
* chore: Return defaultContactRequestID function
* fix: force activate chat on reciever's side
* fix: ensure AC notification's name for CR notifiaction match contact's primary name
Changes applied here introduce:
- improvements to sync wallet accounts among devices (including all account types)
- backing up wallet accounts to and fetch them from waku (an information about received
wallet accounts is sent via `waku.backedup.wallet-account` signal to a client)
* Community request to join changes
* Fix read state for request to join notification
* Bring back deleted notification when updated with response
* Update Request timeout to 7 days
* Update VERSION
* fix(image-album): make sure to delete all images part of an album
* test(delete): add test that deletes a message part of an album
* test(delete): add test where the signal to delete images is after
also adds the handling of deleteForMe
* fix(delete): add album deletion handling for deleteForMe
- `keypair_name` added to `accounts` table, all accounts derived from the
same master key have the same keypair name and also no two keypairs share
the same keypair name (keypair name is unique per keypair)
- `last_used_derivation_index` added to `accounts` table, cause we need
to maintain the highest index been used for the derivations made within
the same keypair
This commit extends the `CommunityRequestToJoin` with `RevealedAddresses` which represent wallet addresses and signatures provided by the sender, to proof a community owner ownership of those wallet addresses.
**Note: This only works with keystore files maanged by status-go**
At high level, the follwing happens:
1. User instructs Status to send a request to join to a community. By adding a password hash to the instruction, Status will try to unlock the users keystore and verify each wallet account.
2. For every verified wallet account, a signature is created for the following payload, using each wallet's private key
``` keccak256(chatkey + communityID + requestToJoinID) ``` A map of walletAddress->signature is then attached to the community request to join, which will be sent to the community owner
3. The owner node receives the request, and if the community requires users to hold tokens to become a member, it will check and verify whether the given wallet addresses are indeed owned by the sender. If any signature provided by the request cannot be recovered, the request is immediately declined by the owner.
4. The verified addresses are then added to the owner node's database such that, once the request should be accepted, the addresses can be used to check on chain whether they own the necessary funds to fulfill the community's permissions
The checking of required funds is **not** part of this commit. It will be added in a follow-up commit.
`type` column is set for all rows to appropriate value. Before this change
accounts which were generated from the keypair created importing seed phrase
had `generated` value for the `type`.
According to above, a function for generating an account sets the `type`
based on the passed derive from address.
Adds a new column named `deleted` to the table `activity_center_notifications`.
Related PR in Mobile https://github.com/status-im/status-mobile/pull/15106 for a lot more details of the feature.
Why? Relying on the `dismissed` column for soft deletion is no longer viable because the mobile & desktop clients should display dismissed notifications (sometimes), hence the need for a new column to truly represent soft deletion.
* Outgoing contact requests
* Test fix
* Test fix
* Fixes
* Bugfixes
* Bugfixes
* Almost there
* Removed the activity center notification
* Test update
* Almost ready
* Fixes
* Fixes
* feat: Add seen/unseen activity center state
* feat: ActivityCenterState for grouping ActivityCenter unread messages cnt and seen state
* feat: always use messenger's addActivityCenterNotification & add state to the response
* Remove unused activity center endpoints form api and fix test
This commit fixes an issue where when accepting a contact request
the other end would display an extra notification.
It also changes WaitOnResponse to collect results. This should make
tests less flaky, since sometimes messages are processed in different
batches.
Now we need to be though exact on what we expect from the response (i.e
use == instead of >, otherwise the same behavior applies)
This uncovered a couple of issues with messenger.Merge, so I have moved
the struct to use a map based collection instead of an array.
In general, any time a piece of state is updated in the backend, that
should be propagated to the client through signals.
In this case, when a request was accepted, the client wasn't notified,
requiring them to re-fetch the accepted requests and causing
inconsistent state between status-go and client.
Mykhailo Prakhov
Patryk Osmaczko
Jonathan Rainville
richΛrd
Alexander
Boris Melnik
Sale Djenic
Igor Sirotin
Ibrahem Khalil
Andrea Maria Piana
flexsurfer
Pascal Precht
Anthony Laibe
frank
Mikhail Rogachev
Khushboo-dev-cpp
Roman Volosovskyi
Volodymyr Kozieiev
Mohamed Javid
yqrashawn
Icaro Motta
Vitaliy Vlasov