This patch extracts fields from `RoomEventCacheState` and move them
into `RoomEventCacheStateLock`. This lock provides 2 methods: `read`
and `write`, respectively to acquire a read-only lock, and a write-only
lock, represented by the `RoomEventCacheStateLockReadGuard` and the
`RoomEventCacheStateLockWriteGuard` types.
All “public” methods on `RoomEventCacheState` now are facade to the read
and write guards.
This refactoring makes the code to compile with the last change in
`EventCacheStore::lock`, which now returns a `EventCacheStoreLockState`.
The next step is to re-load `RoomEventCacheStateLock` when the lock is
dirty! But before doing that, we need this new mechanism to centralise
the management of the store lock.
In the previous version of the thread, the following sequence of events
could happen:
- we subscribe to the thread linked chunk changes
- *then*, the events are being added to the thread linked chunk
This is an edge case where, since we've subscribed to the thread linked
chunk, the thread root will be "known" to be part of a thread, and will
be appended to the thread linked chunk.
If the events happen in the other order (first the events are added to
the thread linked chunk, then we subscribe to the changes), then the
thread root will not be part of the thread linked chunk (because when it
arrived, we didn't know it would be a thread root). As such, the thread
linked chunk state would end up being different in this case.
The solution is to make it so that the thread linked chunk is always
subscribed to *before* any events are added to it. This way, we make
sure that we'll always have the thread root in the thread linked chunk.
In this initial version, a redaction will:
- *remove* the event from the thread chunk, as does Element Web,
- update the thread summary to reflect the new number of messages in the
thread, and let us have a thread summary with 0 replies.
A next commit will adapt the code so that a thread summary with 0
replies is removed.
We actually want other event formats in those cases, and in most cases
just using `.into()` is enough to generate the proper format.
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
Instead of having one static method duplicating an underlying
`ClientBuilder` method, we can pass the builder directly to a closure,
that will replace it. Call sites are a bit more verbose, but that would
avoid having to add duplicate `MockClientBuilder` methods for each
`ClientBuilder` method.
This patch updates our `Raw` API usage since the newly added `JsonCastable` that disallows Raw casts that are known to fail deserialization.
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
This patch removes the use of `OrderTracker` because the
implementation of `EventCacheStore::load_all_chunks_metadata` for
`SqliteEventCacheStore` is the cause of severe slownesses (up to 100s
for some account).
We are going to undo this patch once the problem has been solved.
Ivan Enderlin