The `timeout` setting on the `/sync` endpoint is the maximum allowed
time for the server to send its response, because this is a poll-based
API. It means that if there is no new data to show, the server will wait
until the end of `timeout` before returning a response.
It can be an undesirable behavior when starting a client and informing
the user that we are "catching up" while waiting for the first response.
By not setting a `timeout` on the first request to `/sync`, the
homeserver should reply immediately, whether the response is empty or
not.
---------
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
Signed-off-by: Ivan Enderlin <ivan@mnt.io>
Co-authored-by: Ivan Enderlin <ivan@mnt.io>
It is supposed to be an improvement over `state`, since it allows the
server to send updates to the state that might not be reflected in the
timeline.
This is also the same behavior as in Simplified Sliding Sync.
This is MSC4222 that was accepted and is about to get merged in the
spec.
---------
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
This brings 2 important bug fixes:
- Make deprecated fields of `m.room.encrypted` optional: it seems that there are events without these fields in the wild.
- Fix deserialization of `RedactedRoomJoinRulesEventContent`. This was found by a bug report in Fractal that caused the same error as #3557 when restoring the client. So maybe we could consider that this bug is fixed? It is still possible that there is another deserialization error.
There is also a breaking change in the format of the `state` field in response to `GET /v3/sync`.
This is a breaking change as spaces are now requested through sliding sync and they need to manually be excluded from the room list by using the newly introduced non-space filter.
We use `cargo ndk -p {package_name}`, where `-p` is short for
`--package`. However, `cargo ndk` has introduced the `-p` option (see
https://github.com/bbqsrc/cargo-ndk/commit/c6b93a89a2723ff0fac99b5ac86ae6636a6cf54a),
short for `--platform`. It creates a confusion and the command line
doesn't execute properly. Let's use the long option `--package` to
clarify everything.
The `m.room.create` event contains at lot of important information for a
room, like its type (i.e. whether it is a space), its version and its
creator(s) (which are important in room version 12). So ignoring the
event completely might break a room.
Instead what we do here is simply ignore the `predecessor` field if it
is considered invalid, allowing us to access the other fields.
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
Fixes the encrypting part of
https://github.com/matrix-org/matrix-rust-sdk/issues/4147
Probably easiest to review commit-by-commit
<!-- description of the changes in this PR -->
- [x] Public API changes documented in changelogs (optional)
<!-- Sign-off, if not part of the commits -->
<!-- See CONTRIBUTING.md if you don't know what this is -->
Signed-off-by:
This brings in a new breaking change from Ruma, because not all events
are stripped in a room's stripped state. For simplicity, this still
considers the events as stripped during deserialization for now, since
this format is compatible with the other possible formats.
Signed-off-by: Kévin Commaille <zecakeh@tedomum.fr>
Jonas Platte