[web] Olm session creation between own devices
ClosedPublic
Actions

Authored by michal on Mar 15 2024, 10:38 AM.

Details

Reviewers

kamil
tomek

Commits

rCOMM90a9d4b9e87e: [web] Olm session creation between own devices

Summary

ENG-7221 : Tunnelbroker session creation on web

Now that we have a proper olm session persistance on web we can implement testing screens for peer to peer message sending, and creating sessions between own devices.

getContentSigningKey and createOlmSessionsWithOwnDevices functions have been moved from native to lib and updated to use the cross-platform APIs.

Depends on D11340

Test Plan

Created sessions from web to native and tried sending messages both ways. Repeated with native creating the session to web. In both cases the messages where successfully ecnrypted/decrypted.

Diff Detail

Repository

rCOMM Comm

Lint

No Lint Coverage

Unit

No Test Coverage

Event Timeline

michal created this revision.Mar 15 2024, 10:38 AM

Herald added a subscriber: ashoat. · View Herald TranscriptMar 15 2024, 10:38 AM

michal added a child revision: D11342: [web/native] Move notificationsSessionCreator to olmAPI.Mar 15 2024, 10:40 AM

Harbormaster completed remote builds in B27537: Diff 38106.Mar 15 2024, 10:53 AM

michal requested review of this revision.Mar 15 2024, 10:53 AM

kamil accepted this revision.Mar 18 2024, 6:09 AM

kamil added inline comments.

lib/utils/crypto-utils.js
17–18 ↗	(On Diff #38106)	can be merged
63 ↗	(On Diff #38106)	we can add a log that key's set is missing for this device

This revision is now accepted and ready to land.Mar 18 2024, 6:09 AM

Fixes

Harbormaster completed remote builds in B27588: Diff 38177.Mar 19 2024, 8:44 AM

ashoat added inline comments.Mar 19 2024, 5:50 PM

lib/utils/crypto-utils.js
47	Thought I had while reading the code, that's not relevant to this diff, and doesn't affect the current (imperative) usage of this function in the debug menus: We'll probably eventually want this function to only create Olm sessions with the devices that don't currently already have sessions. There's a risk that it fails to create a session (if one-time keys are missing, or perhaps a race condition), so we should be able to call it again later and create only the sessions that were missed.