* Remove refs to deprecated method
`MatrixEvent.isEncryptedDisabledForUnverifiedDevices` is now deprecated
* Report new posthog code for withheld message keys
* Add flag for tests to avoid double-reporting check
Some of the tests were flaking. Our best guess is that it's failing to track
some events due to false positives in the Bloom filter used to guard against
double-reporting. So we add a flag to disable using that in tests (except for
tests that test that functionality).
* invert logic to avoid double negative
* persist previously-reported event IDs as a bloom filter
* Pin to older `@types/seedrandom`
... to work around https://github.com/Callidon/bloom-filters/issues/72
* Inline `DecryptionFailureTracker.addDecryptionFailure`
* Remove redundant TRACK_INTERVAL
There really doesn't seem to be much point to this batching up of decryption
failure reports. We still call the analytics callback the same number of times.
* Rename `trackedEvents` to `reportedEvents`
* Fix incorrect documentation on `visibleEvents`
This *does* overlap with `failures`.
* Combine `addFailure` and `reportFailure`
* Calculate client properties before starting reporting
* Clear localstorage after each test
... otherwise they interfere
* Remove redundant comment
* Ensure that reports are cleared on a logout/login cycle
* make private const private and const
---------
Co-authored-by: Richard van der Hoff <richard@matrix.org>
* Inline `DecryptionFailureTracker.addDecryptionFailure`
* Remove redundant TRACK_INTERVAL
There really doesn't seem to be much point to this batching up of decryption
failure reports. We still call the analytics callback the same number of times.
* Rename `trackedEvents` to `reportedEvents`
* Fix incorrect documentation on `visibleEvents`
This *does* overlap with `failures`.
* Combine `addFailure` and `reportFailure`
* Calculate client properties before starting reporting
* report time to decrypt an event if it's slow
* add more stats
* fix some tests
* keep original timestamp, and report non-visible decryption failures
* add statistic for whether the user trusts their own device
* revert local playwright docker changes
* apply changes from review
* also remove logout event handler on logout
* apply changes from review
* make eventDecrypted callback private
* Use different messages for UTDs sent before login
* Playwright test for historical events
* Add some tests
* initial work on pre-join UTDs
* add playwright tests and remove old pre-join UTD logic
* run i18n script
* fix type error
* use different error code in PostHog for pre-join UTD
* don't remove old pre-join UTD logic yet
---------
Co-authored-by: Richard van der Hoff <richard@matrix.org>
* `DecryptionFailureTracker`: stronger typing
Use `DecryptionFailureCode` rather than string
* `DecryptionFailureTracker`: remove use of `DecryptionError`
The second argument to `MatrixEventEvent.Decrypted` callbacks is deprecatedf,
and we can get the info we need direct from the event. This means that we no
longer need to reference the internal `DecryptionError` class in the js-sdk.
* `DecryptionFailureTracker`: use a different Posthog code for historical UTDs
* Update for new UTD error codes
* Remove all mentions of Piwik
* Kill off all consumer of the old Piwik Analytics module
* Simplify ModalManager interface
* i18n
* Attempt to fix old e2e tests
* Remove unused component
* Iterate PR
* Add matrix-analytics-events as a dependency
* Make IEvent look like a stub definition
* Update pageview tracking to track screens, using a hypothetical definition of a screen event
* Remove distinction between pseudo and anon tracking, will need to rework it considering stubs
