This reworks tracking the Olm sessions a particular task is updating to avoid
deadlocks. By ensuring we synchronously mark all sessions a task cares about as
in progress from the start, we know that no other tasks will own updating a
session in common, which avoids deadlocks across multiple tasks that might be
working on a shared set of devices.
Fixes https://github.com/vector-im/element-web/issues/16194
- allow applications to pre-send decryption keys before the message is sent
- establish new olm sessions with a shorter timeout first, and then re-try in
the background with a longer timeout without blocking message sending
- only exporting account and P2P sessions
- test is halfway done:
- it only prints the export result instead of running assertions on it
- there are no sessions to export
Note: to run only the added test:
node_modules/.bin/jest spec/unit/crypto/algorithms/olm.spec.js --testEnvironment node --testNamePattern OlmDevice
The earlier commit, d3ce0cb82f, has most of the juicy details on this. In addition to d3ce's changes, we also:
* Use `TestClient` in many integration tests due to subtle behaviour changes in imports when switching to ES6. Namely the behaviour where setting the request function is less reliable in the way we did it, but `TestClient` is very reliable.
* We now use the Olm loader more often to avoid having to maintain so much duplicate code. This makes the imports slightly easier to read.
The code to migrate from the `sessionStore` to `cryptoStore` originally appeared
in https://github.com/matrix-org/matrix-js-sdk/pull/584 (2017-12-06). At this
point, it seems safe to assume most sessions that need migrating have already
done so. Removing this code simplifies store handling and removes the
`sessionStore` from most places in JS SDK.
Standardize on importing `olm-loader` rather than pasting the same boilerplate
in different tests. Importantly, `spec/unit/crypto.spec.js` did not include any
loading approach, so it would only find Olm if some other test loaded it first.
Signed-off-by: J. Ryan Stinnett <jryans@gmail.com>
* Start a new Olm sessions with a device when we get an undecryptable
message on it.
* Send a dummy message on that sessions such that the other end knows
about it.
* Re-send any outstanding keyshare requests for that device.
Also includes a unit test for megolm that isn't very related but came
out as a result anyway.
Includes https://github.com/matrix-org/matrix-js-sdk/pull/776
Fixes https://github.com/vector-im/riot-web/issues/3822
wasm Olm has a new interface: it now has an init method that needs
to be called and the promise it returns waited on before the Olm
module is used. Support that, and allow Crypto etc to be imported
whether Olm is enabled or not. Change whether olm is enabled to
be async since now it will be unavailable if the async module init
fails. Don't call getOlmVersion() until the Olm.init() is done.
* OlmDevice.hasInboundSessionKeys
* OlmDevice.getInboundGroupSessionKey
The latter means that MegolmDecryption.shareKeysWithDevice takes longer before
it sends out the keyshare, so means the unit test needed an update
... instead of having it call event.setClearData.
The main advantage of this is that it fixes a race condition, wherein apps
could see `event.isDecrypting()` to be true, but in fact the event had been
decrypted (and there was no `Event.decrypted` event on its way).
We're also fixing another race, wherein if the first attempt to decrypt failed,
a call to `attemptDecryption` would race against the first call and a second
attempt to decrypt would never happen.
This also gives a cleaner interface to MatrixEvent, at the expense of making
the `megolm` unit test a bit more hoop-jumpy.
Michael Telatynski