- Change incorrect "key sharing" references to "key forwarding".
- Explain that this is a feature that needs to be enabled.
- Regenerate the key forwarding algorithm diagram.
- Provide a spec link and mention alternative names.
Use a Bloom filter to keep track of which events we have already reported to the parent UTD hook.
We load the data from database on startup, and flush it out immediately on every update.
This patch renames `TimelineEventHandler::add` to `add_item`.
This patch also removes the `should_add` argument. The `add_item`
is called conditionally everytime. I believe this is much cleaner.
Otherwise the method should have been called `maybe_add_item` with a
return type or something that indicates whether the item is added. This
patch makes it clear and remove one possible state in `add_item`.
This patch updates `TimelineEventHandler` to re-use the public API and
avoid using `TimelineInnerMeta::next_internal_id`. The consequence is
that `next_internal_id` can now be private instead of public. It's
better to have isolated API in this code.
This patch removes a useless clone of `TimelineInner::items`. This clone
was technically cheap because it's a `Vector` behind the scene, which is
cheap to clone, but still, it's not necessary at all to clone it.
Because of the task cancellation that can happen at any place in the
code base, it's really hard to predict in which situation a
day-divider-adjuster should have run or not, so just demote the assert
to an error. The intent is that, if we see errors with day dividers in
the future, they'll be reported along rageshakes and we'll notice this
in the logs.
Using async when not required will increase compile times, and propagate async-ness to the callers, transitively.
See also the [lint description](https://rust-lang.github.io/rust-clippy/master/#/unused_async).
Since we only had a few false positives, I've enabled it by default.
Also reunify two methods that did the same thing, with slightly
different semantics, and test the one that wasn't tested at all.
Note that `is_editable()` was already exposed to the FFI and used in EX
apps.
It was used after a previous local echo couldn't be sent (i.e. the
sending queue failed to send it). However, it was slightly incorrect to
mark those as cancelled, since those local echoes would still have
corresponding items in the sending queue, and would be retried as soon
as the sending queue was reenabled and could send events.
Instead, this is removed: it means that previously cancelled events
would be in the NotSentYet state, which is correct. (At the limit, we
could even get rid of the SendingFailed variant, since the sending queue
will automatically retry as soon as possible.)
The test relied on the fact that sending an event from a given timeline
is not observable from another timeline. Indeed, it sent a message using
a first timeline object, then constructed a second timeline object and
expected only the remote event to be in there.
Now, the sending queue is shared across all instances of a Room, thus
all instances of a Timeline, and the second timeline can see the local
echo for the message sent by the first timeline.
The "fix" is thus in the test structure itself: when waiting for the
remote echo to be there, check that the timeline item doesn't pertain to
a local echo, i.e. is a remote echo.
Technically, the test already passed before the change in the builder,
because `TimelineEventHandler::handle_event` already filters out local
events on non-live timelines, but it's a waste of resources to even
spawn the local echo listener task in that case, so let's not do it.
By keeping a reference to the FFI Client, we make sure that the SDK
Client is dropped while in a tokio runtime context. This makes it
possible for an `Encryption` (FFI) object to outlive the `Client` (FFI)
object without crashing (because deadpool requires to be in a tokio
runtime context when dropping).
Following https://github.com/matrix-org/matrix-rust-sdk/pull/3473, we
made it so that if there's a base-path but no username, then we'll
create a sqlite database.
Unfortunately, this introduced a bad side-effect: for the temporary
client used during homeserver resolution, this would create a temporary
sqlite database.
The "fix" is to not set the base path for the temporary client, and only
for the other caller of `new_client_builder()`, manually. The long term
fix is to get rid of the `AuthenticationService` so we can test it
properly.
This patch updates Ruma to 75e8829 so that `RoomSummary::heroes` is now
a `Vec<OwnedUserId>` instead of a `Vec<String>`. This patch updates our
code accordingly by removing the parsing of heroes as `OwnedUserId`.
Without this, the configs from `.cargo/config.toml` were not read in CI
tasks, causing false positives when running Clippy on CI (i.e. there
were issues observed when compiling locally that were not found when
compiling remotely).
Not entirely sure why it's needed, because I'm seeing the issues when
I'm using `cargo xtask ci clippy` locally, with nothing changed.
Benjamin Bouvier