* feat(pairing)!: Add extra parameters and remove v2 compatibility
This commit includes the following changes:
I have added a flag to maintain 2.29 compatibility.
Breaking change in connection string
The local pairing code that was parsing the connection string had a few non-upgradable features:
It was strictly checking the number of parameters, throwing an error if the number was different. This made it impossible to add parameters to it without breaking.
It was strictly checking the version number. This made increasing the version number impossible as older client would just refuse to connect.
The code has been changed so that:
Two parameters have been added, installation-id and key-uid. Those are needed for the fallback flow.
I have also removed version from the payload, since it wasn't used.
This means that we don't support v1 anymore. V2 parsing is supported . Going forward there's a clear strategy on how to update the protocol (append parameters, don't change existing one).
https://www.youtube.com/watch?v=oyLBGkS5ICk Is a must watch video for understanding the strategy
Changed MessengerResponse to use internally a map of installations rather than an array (minor)
Just moving towards maps as arrays tend to lead to subtle bugs.
Moved pairing methods to messenger_pairing.go
Just moved some methods
Added 2 new methods for the fallback flow
FinishPairingThroughSeedPhraseProcess
https://github.com/status-im/status-go/pull/5567/files#diff-1ad620b07fa3bd5fbc96c9f459d88829938a162bf1aaf41c61dea6e38b488d54R29
EnableAndSyncInstallation
https://github.com/status-im/status-go/pull/5567/files#diff-1ad620b07fa3bd5fbc96c9f459d88829938a162bf1aaf41c61dea6e38b488d54R18
Flow for clients
Client A1 is logged in
Client A2 is logged out
Client A1 shows a QR code
Client A2 scans a QR code
If connection fails on A2, the user will be prompted to enter a seed phrase.
If the generated account matches the key-uid from the QR code, A2 should call FinishPairingThroughSeedPhraseProcess with the installation id passed in the QR code. This will send installation information over waku. The user should be shown its own installation id and prompted to check the other device.
Client A1 will receive new installation data through waku, if they are still on the qr code page, they should show a popup to the user showing the received installation id, and a way to Enable and Sync, which should call the EnableAndSyncInstallation endpoint. This should finish the fallback syncing flow.
Current issues
Currently I haven't tested that all the data is synced after finishing the flow. I see that the two devices are paired correctly, but for example the DisplayName is not changed on the receiving device. I haven't had time to look into it further.
* test_: add more test for connection string parser
* fix_: fix panic when parse old connection string
* test_: add comments for TestMessengerPairAfterSeedPhrase
* fix_: correct error description
* feat_:rename FinishPairingThroughSeedPhraseProcess to EnableInstallationAndPair
* fix_: delete leftover
* fix_: add UniqueKey method
* fix_: unify the response for InputConnectionStringForBootstrapping
* fix_: remove fields installationID and keyUID in GethStatusBackend
* fix_: rename messenger_pairing to messenger_pairing_and_syncing
---------
Co-authored-by: Andrea Maria Piana <andrea.maria.piana@gmail.com>
* feat_: Check for mobile data connection if setting is on
* fix_: check code control flag inside asyncRequestAllHistoricMessages
---------
Co-authored-by: Igor Sirotin <sirotin@status.im>
This commit adds support for centralized metrics.
There are two providers as of now, and we haven't quite decided which
one to go for, so for the time being both are supported.
It also introduces a new endpoint InitializeApplication that replaces
OpenAccounts
FIxes https://github.com/status-im/status-desktop/issues/15284
Somehow, the display name could get messed up and when it happened, no message could be received, because we also validate the display name on reception.
I think the problem for the particular user happened because the special display name was added before we enforced it, or through the sync.
- TokenMaster request shared addresses on restoring from a backup
- Fixes sharing members revealed adresses
- Filter out outdated ApplicationMetadataMessage_COMMUNITY_PRIVILEGED_USER_SYNC_MESSAGE messages
This commit limits the number of ephemeral keys used by the push
notification client to 3, as it was noticed that it would greatly
increase the number of filters installed and was unbound.
* feat_: aggregate filter subscriptions to do bulk subs with peer
* chore_: take possible deadlock fix in go-waku
* fix_: don't resubscribe filters unless there is a change in shard for community (#5467)
This commit removes the list of members from non token gated channels.
Unfortunately is a breaking change. I could make it non-breaking, but
we would lose any performance benefit.
For clients, the pseudo code for checking the member list of a channel
is:
```
members := channel.TokenGated ? channel.Members : community.Members
```
* feat_: add ability to enable http and ws connections from the client app
* feat_: add websocket option for api config
* fix_: use new api options in statusd
* chore_: add test for `overrideApiConfig`
Fixes https://github.com/status-im/status-desktop/issues/15175
The problem was that we used StartMembersReevaluaitonLoop in createCommunityPermission, in case the loop was never started. Indeed it worked if it was the first ever permission, because the loop would then start and members would be re-evaluated.
However, if the loop was already started, in the case where there were previous permissions, the call would just early exit, because it was already started.
The solution here is to use `ScheduleMembersReevaluaiton` like other permission functions use. To make sure a first permission still works, we call startLoop in schedule if the task doesn't exist.
* feat(community)_: Move images from community data to MediaServer
* test_: fix lint issue
* test_: add more test statements
* feat_: deprecate old API
* test_: addressed review feedback from Icaro
* fix_: addressed review feedback from Jonathan
* chore_:wrap image url in an object
Fixed logic to respect specified NFT quantities. Previously, holding one
NFT sufficed, regardless of the required count.
fixes: status-im/status-desktop#15122
chore(wallet)_: split getWalletTokenBalances into multiple functions
Removed some unused balances methods from wallet API
chore(wallet)_: refactored FetchOrGetWalletTokenBalances
- getWalletTokenBalances only returns cached ones
- update of balances is done in a separate method
chore(wallet)_: fix isVisible in getWalletTokenBalances is overwritten
It is overwritten and in some cases its value is desrespected
chore(wallet)_: simplify getWalletTokenBalance even further
chore(wallet)_: remove accountsDB from wallet.Reader
Call GetTestNetworkEnabled from NetworkManager instead
chore(wallet)_: remove rpc.Client from wallet.Reader.
Added GetActiveNetworks() method for NetworkManager
Removed adding native tokens from networks, as this is done already
in NetworkManager
chore(wallet)_: moved Persistence to token package
As it works with token_balances table, moved Persistence to token package.
Fixed TokenManager's Mark/Get previously owned tokens to use persistence
storage instead of direct SQL calls.
Introduced StorageToken that aggregates Token type, because when
Persistence moved to token package, names clash
test(wallet)_: tests for wallet.Reader.FetchorGetCachedBalances
* fix_: send request for shared addresses when restoring a community
Needed for https://github.com/status-im/status-desktop/issues/14289
When someone restores a community that they joined, it now sends a message to the control node to get their shared addresses back.
It's lighter that backing up the addresses all the time on waku and should only be needed once.
* fix_: filter communities before to avoid passing `alreadyHandled` parameter to `requestCommunityKeysAndSharedAddresses`
---------
Co-authored-by: Patryk Osmaczko <osmaczkopatryk@gmail.com>
* fix_: add status telemetry client to cli
* feat_: call telemetry when pushing an envelope
* feat_: log status version in all telemetry calls
* feat_: batch all telemetry data and send request every 10 seconds
* feat_: hash based query for outgoing messages.
* chore_: more logs
* chore_: fix comments
* chore_: do not lock when send queries
* chore_: use constant for magic number
* chore_: remove message ids from query queue after ack
* chore_: fix ack clean process
* chore_: fix message resend test
* chore_: add test for waku confirm message sent.
* chore_: fix tests.
* chore_: fix more
* chore_: set store peer id when mailserver updates
* fix_: tests
* chore_: increase max hash query length
* chore_: remove debug log of ack message
* chore_: remove automatic peer selection
* chore_: mark raw message to sent after ack
* chore_: fix test
* chore_: fix test
Prevent the inclusion of CommunityDescription with an outdated clock in
MessengerResponse to avoid false-positives in tests and reduce redundant
data exchange between status-go and clients.
Fixes the slow login in mobile devices when users have joined large communities,
such as the Status one. A user would get stuck for almost 20s in some devices.
We identified that the step to set-up filters in the messenger is potentially
expensive and that it is not critical to happen before the node.login signal is
emitted. The solution presented in this PR is to set-up filters inside
messenger.Start(), which the client already calls immediately after login.
With this change, users of the mobile app can login pretty fast even when they
joined large communities. They can immediately interact with other parts of the
app even if filter initialization is running in the background, like Wallet,
Activity Center, Settings, and Profile.
Breaking changes: in the mobile repository, we had to change where the endpoint
wakuext_startMessenger was called and the order of a few events to process
chats. So essentially ordering, but no data changes.
- Root issue https://github.com/status-im/status-mobile/issues/20059
- Related mobile PR https://github.com/status-im/status-mobile/pull/20173
That's an optimisation. Instead of fetching collectibles owners for each
member, it is fetched once, before members iteration.
It should significantly reduce amount of queries to providers.
closes: status-im/status-desktop#14914
I've renamed TorrentManager to ArchiveManager, ArchiveManager to ArchiveFileManager, TorrentContract to ArchiveService, ArchiveContract to ArchiveFileService. I've also renamed all init functions and struct fields to the appropriate archive-centric naming.
Instead of using the TorrentContract interface I've set the field to expicitly declare as *TorrentManager. This removes all the random type casting used in the tests. I also removed all the usages of buildTorrentConfig() as we build the test suite with the torrent config now.
To be honest once I started this work I quickly realised how pointless it is as archiving functionality and torrent seeding functionality are really entwined. So I'm keeping the code I've done but it is a bit pointless without spending a lot of time untangling torrenting and archiving. I'm just going to make an interface for all the functions that are used publically from TorrentManager. I think that this will be the fast way to move on from this issue. I don't like this work any more, there is a lot of work to do elsewhere and torrent is a rabbit hole filled with canned worms.
I've removed any mention or dependency of TorrentManager from Manager. There is still more work to do, but Messenger now communicates directly with a TorrentManager rather than asking the communities Manager to handle it. I've ensured that LogStdout() is only called from TorrentManager and removed entirely from Manager, this functionality seems to be some kind of debug tool specifically for torrent related functionality. Next I need to focus on functions within Messenger that call a TorrentManager and see how to isolate these from the main flows, following that I also need fix the tests that are broken. I will also need to refactor torrentClientReady() so that it is a function of TorrentManager, this may allow for pushing more functions into TorrentManager which will lead to better torrent encapsulation.
I'm not particularly proud of this work. I've just passed in all the deps as vars, there are way too many concerns handled by the TorrentManager that I don't know what is the best approach to removing them. I've even resorted to declaring an 'Publisher' interface to handle all the 'publish()' calls the TorrentManager makes. Next up I need to resolved the communities Manager API, because I've removed all of its end points. The TorrentManager needs to be a lot more simple than it is.
Code moved into new TorrentManager. This commit is BROKEN! The code is not ready to use so don't use it, lots more work to do. Biggest problem is that the torrent management in Manager is very tightly coupled to sending encrypting etc. All of that needs to be prised apart
- Logs the requestID in the response
- Log cursor details
- Refactor_: reduce number of parameters for query and fix tests
- Fix_: rename `DefaultShard` to `DefaultNonProtectedShard`
IvanBelyakoff
frank
richΛrd
Igor Sirotin
Andrea Maria Piana
Andrey Bocharnikov
Mykhailo Prakhov
Vitaly Vlasov
Patryk Osmaczko
Jonathan Rainville
Prem Chaitanya Prathi
kaichao
Michal Iskierko
Jakub Sokołowski
Arseniy Klempner
Mikhail Rogachev
Pablo Lopez
Sale Djenic
Godfrain Jacques
Icaro Motta
Samuel Hawksby-Robinson