Drop 7.5s

See https://github.com/matrix-org/synapse/pull/13478#discussion_r943861128

.github/workflows/latest_deps.yml

@@ -135,11 +135,42 @@ jobs:
             /logs/**/*.log*
 
 
-  # TODO: run complement (as with twisted trunk, see #12473).
+  complement:
+    if: "${{ !failure() && !cancelled() }}"
+    runs-on: ubuntu-latest
 
-  # open an issue if the build fails, so we know about it.
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - arrangement: monolith
+            database: SQLite
+
+          - arrangement: monolith
+            database: Postgres
+
+          - arrangement: workers
+            database: Postgres
+
+    steps:
+      - name: Run actions/checkout@v2 for synapse
+        uses: actions/checkout@v2
+        with:
+          path: synapse
+
+      - name: Prepare Complement's Prerequisites
+        run: synapse/.ci/scripts/setup_complement_prerequisites.sh
+
+      - run: |
+          set -o pipefail
+          TEST_ONLY_IGNORE_POETRY_LOCKFILE=1 POSTGRES=${{ (matrix.database == 'Postgres') && 1 || '' }} WORKERS=${{ (matrix.arrangement == 'workers') && 1 || '' }} COMPLEMENT_DIR=`pwd`/complement synapse/scripts-dev/complement.sh -json 2>&1 | gotestfmt
+        shell: bash
+        name: Run Complement Tests
+
+  # Open an issue if the build fails, so we know about it.
+  # Only do this if we're not experimenting with this action in a PR.
   open-issue:
-    if: failure()
+    if: "failure() && github.event_name != 'push' && github.event_name != 'pull_request'"
     needs:
       # TODO: should mypy be included here? It feels more brittle than the other two.
       - mypy

.github/workflows/tests.yml

@@ -328,6 +328,9 @@ jobs:
           - arrangement: monolith
             database: Postgres
 
+          - arrangement: workers
+            database: Postgres
+
     steps:
       - name: Run actions/checkout@v2 for synapse
         uses: actions/checkout@v2
@@ -343,30 +346,6 @@ jobs:
         shell: bash
         name: Run Complement Tests
 
-  # XXX When complement with workers is stable, move this back into the standard
-  #     "complement" matrix above.
-  #
-  # See https://github.com/matrix-org/synapse/issues/13161
-  complement-workers:
-    if: "${{ !failure() && !cancelled() }}"
-    needs: linting-done
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Run actions/checkout@v2 for synapse
-        uses: actions/checkout@v2
-        with:
-          path: synapse
-
-      - name: Prepare Complement's Prerequisites
-        run: synapse/.ci/scripts/setup_complement_prerequisites.sh
-
-      - run: |
-          set -o pipefail
-          POSTGRES=1 WORKERS=1 COMPLEMENT_DIR=`pwd`/complement synapse/scripts-dev/complement.sh -json 2>&1 | gotestfmt
-        shell: bash
-        name: Run Complement Tests
-
   # a job which marks all the other jobs as complete, thus allowing PRs to be merged.
   tests-done:
     if: ${{ always() }}

changelog.d/12978.misc

@@ -0,0 +1 @@
+Extend the release script to automatically push a new SyTest branch, rather than having that be a manual process.

changelog.d/13160.misc

@@ -0,0 +1 @@
+Make minor clarifications to the error messages given when we fail to join a room via any server.

changelog.d/13213.misc

@@ -0,0 +1 @@
+Enable Complement CI tests in the 'latest deps' test run.

changelog.d/13221.doc

@@ -0,0 +1 @@
+Document which HTTP resources support gzip compression.

changelog.d/13230.doc

@@ -0,0 +1 @@
+Add steps describing how to elevate an existing user to administrator by manipulating the database.

changelog.d/13343.feature

@@ -0,0 +1 @@
+Add new unstable error codes `ORG.MATRIX.MSC3848.ALREADY_JOINED`, `ORG.MATRIX.MSC3848.NOT_JOINED`, and `ORG.MATRIX.MSC3848.INSUFFICIENT_POWER` described in MSC3848.

changelog.d/13346.misc

@@ -0,0 +1 @@
+Fix long-standing bugged logic which was never hit in `get_pdu` asking every remote destination even after it finds an event.

changelog.d/13353.bugfix

@@ -0,0 +1 @@
+Fix a bug in the experimental faster-room-joins support which could cause it to get stuck in an infinite loop.

changelog.d/13355.misc

@@ -0,0 +1 @@
+Faster room joins: avoid blocking when pulling events with partially missing prev events.

changelog.d/13365.bugfix

@@ -0,0 +1 @@
+Fix a bug introduced in Synapse v1.41.0 where the `/hierarchy` API returned non-standard information (a `room_id` field under each entry in `children_state`).

changelog.d/13368.misc

@@ -0,0 +1 @@
+Instrument `/messages` for understandable traces in Jaeger.

changelog.d/13370.feature

@@ -0,0 +1 @@
+Use stable prefixes for [MSC3827](https://github.com/matrix-org/matrix-spec-proposals/pull/3827).

changelog.d/13372.docker

@@ -0,0 +1 @@
+Make docker images build on armv7 by installing cryptography dependencies in the "requirements" stage. Contributed by Jasper Spaans.

changelog.d/13374.bugfix

@@ -0,0 +1 @@
+Fix a bug introduced in Synapse 0.24.0 that would respond with the wrong error status code to `/joined_members` requests when the requester is not a current member of the room. Contributed by @andrewdoh.

changelog.d/13383.misc

@@ -0,0 +1 @@
+Remove an unused argument to `get_relations_for_event`.

changelog.d/13392.bugfix

@@ -0,0 +1 @@
+Fix bug in handling of typing events for appservices. Contributed by Nick @ Beeper (@fizzadar).

changelog.d/13393.misc

@@ -0,0 +1 @@
+Add a `merge-back` command to the release script, which automates merging the correct branches after a release.

changelog.d/13397.misc

@@ -0,0 +1 @@
+Adding missing type hints to tests.

changelog.d/13403.misc

@@ -0,0 +1 @@
+Faster Room Joins: don't leave a stuck room partial state flag if the join fails.

changelog.d/13404.misc

@@ -0,0 +1 @@
+Refactor `_resolve_state_at_missing_prevs` to compute an `EventContext` instead.

changelog.d/13408.bugfix

@@ -0,0 +1 @@
+Fix a bug introduced in Synapse 1.57.0 where rooms listed in `exclude_rooms_from_sync` in the configuration file would not be properly excluded from incremental syncs.

changelog.d/13413.bugfix

@@ -0,0 +1 @@
+Faster room joins: fix a bug which caused rejected events to become un-rejected during state syncing.

changelog.d/13416.misc

@@ -0,0 +1 @@
+Faster Room Joins: prevent Synapse from answering federated join requests for a room which it has not fully joined yet.

changelog.d/13420.misc

@@ -0,0 +1 @@
+Re-enable running Complement tests against Synapse with workers.

changelog.d/13428.feature

@@ -0,0 +1 @@
+Add a module API method to translate a room alias into a room ID.

changelog.d/13429.feature

@@ -0,0 +1 @@
+Add a module API method to create a room.

changelog.d/13431.misc

@@ -0,0 +1 @@
+Refactor `_resolve_state_at_missing_prevs` to compute an `EventContext` instead.

changelog.d/13432.bugfix

@@ -0,0 +1 @@
+Faster room joins: Fix error when running out of servers to sync partial state with, so that Synapse raises the intended error instead.

changelog.d/13435.misc

@@ -0,0 +1 @@
+Prevent unnecessary lookups to any external `get_event` cache. Contributed by Nick @ Beeper (@fizzadar).

changelog.d/13437.doc

@@ -0,0 +1 @@
+Fix wrong headline for `url_preview_accept_language` in documentation.

changelog.d/13438.doc

@@ -0,0 +1 @@
+Remove redundant 'Contents' section from the Configuration Manual. Contributed by @dklimpel.

changelog.d/13439.misc

@@ -0,0 +1 @@
+Add some tracing to give more insight into local room joins.

changelog.d/13442.misc

@@ -0,0 +1 @@
+Rename class `RateLimitConfig` to `RatelimitSettings` and `FederationRateLimitConfig` to `FederationRatelimitSettings`.

changelog.d/13443.doc

@@ -0,0 +1 @@
+Update documentation for config setting `macaroon_secret_key`.

changelog.d/13447.misc

@@ -0,0 +1 @@
+Improve rebuild speed for the "synapse-workers" docker image.

changelog.d/13449.doc

@@ -0,0 +1 @@
+Update outdated information on `sso_mapping_providers` documentation.

changelog.d/13450.doc

@@ -0,0 +1 @@
+Fix example code in module documentation of `password_auth_provider_callbacks`.

changelog.d/13452.misc

@@ -0,0 +1 @@
+Fix `@tag_args` being off-by-one with the arguments when tagging a span (tracing).

changelog.d/13455.misc

@@ -0,0 +1 @@
+Add some comments about how event push actions are stored.

changelog.d/13460.misc

@@ -0,0 +1 @@
+Update type of `EventContext.rejected`.

changelog.d/13478.misc

@@ -0,0 +1 @@
+Fix the `synapse_http_server_response_time_seconds` metric not having buckets big enough for requests that take more than 10s.

docker/Dockerfile

@@ -40,7 +40,8 @@ FROM docker.io/python:${PYTHON_VERSION}-slim as requirements
 RUN \
    --mount=type=cache,target=/var/cache/apt,sharing=locked \
    --mount=type=cache,target=/var/lib/apt,sharing=locked \
- apt-get update -qq && apt-get install -yqq git \
+    apt-get update -qq && apt-get install -yqq \
+      build-essential cargo git libffi-dev libssl-dev \
     && rm -rf /var/lib/apt/lists/*
 
 # We install poetry in its own build stage to avoid its dependencies conflicting with
@@ -68,7 +69,18 @@ COPY pyproject.toml poetry.lock /synapse/
 # reason, such as when a git repository is used directly as a dependency.
 ARG TEST_ONLY_SKIP_DEP_HASH_VERIFICATION
 
-RUN /root/.local/bin/poetry export --extras all -o /synapse/requirements.txt ${TEST_ONLY_SKIP_DEP_HASH_VERIFICATION:+--without-hashes}
+# If specified, we won't use the Poetry lockfile.
+# Instead, we'll just install what a regular `pip install` would from PyPI.
+ARG TEST_ONLY_IGNORE_POETRY_LOCKFILE
+
+# Export the dependencies, but only if we're actually going to use the Poetry lockfile.
+# Otherwise, just create an empty requirements file so that the Dockerfile can
+# proceed.
+RUN if [ -z "$TEST_ONLY_IGNORE_POETRY_LOCKFILE" ]; then \
+    /root/.local/bin/poetry export --extras all -o /synapse/requirements.txt ${TEST_ONLY_SKIP_DEP_HASH_VERIFICATION:+--without-hashes}; \
+  else \
+    touch /synapse/requirements.txt; \
+  fi
 
 ###
 ### Stage 1: builder
@@ -108,8 +120,17 @@ COPY synapse /synapse/synapse/
 # ... and what we need to `pip install`.
 COPY pyproject.toml README.rst /synapse/
 
+# Repeat of earlier build argument declaration, as this is a new build stage.
+ARG TEST_ONLY_IGNORE_POETRY_LOCKFILE
+
 # Install the synapse package itself.
-RUN pip install --prefix="/install" --no-deps --no-warn-script-location /synapse
+# If we have populated requirements.txt, we don't install any dependencies
+# as we should already have those from the previous `pip install` step.
+RUN if [ -z "$TEST_ONLY_IGNORE_POETRY_LOCKFILE" ]; then \
+    pip install --prefix="/install" --no-deps --no-warn-script-location /synapse[all]; \
+  else \
+    pip install --prefix="/install" --no-warn-script-location /synapse[all]; \
+  fi
 
 ###
 ### Stage 2: runtime

docker/Dockerfile-workers

@@ -1,39 +1,62 @@
 # syntax=docker/dockerfile:1
-# Inherit from the official Synapse docker image
+
 ARG SYNAPSE_VERSION=latest
+
+# first of all, we create a base image with an nginx which we can copy into the
+# target image. For repeated rebuilds, this is much faster than apt installing
+# each time.
+
+FROM debian:bullseye-slim AS deps_base
+    RUN \
+       --mount=type=cache,target=/var/cache/apt,sharing=locked \
+       --mount=type=cache,target=/var/lib/apt,sharing=locked \
+      apt-get update -qq && \
+      DEBIAN_FRONTEND=noninteractive apt-get install -yqq --no-install-recommends \
+          redis-server nginx-light
+
+# Similarly, a base to copy the redis server from.
+#
+# The redis docker image has fewer dynamic libraries than the debian package,
+# which makes it much easier to copy (but we need to make sure we use an image
+# based on the same debian version as the synapse image, to make sure we get
+# the expected version of libc.
+FROM redis:6-bullseye AS redis_base
+
+# now build the final image, based on the the regular Synapse docker image
 FROM matrixdotorg/synapse:$SYNAPSE_VERSION
 
-# Install deps
-RUN \
-   --mount=type=cache,target=/var/cache/apt,sharing=locked \
-   --mount=type=cache,target=/var/lib/apt,sharing=locked \
-  apt-get update -qq && \
-  DEBIAN_FRONTEND=noninteractive apt-get install -yqq --no-install-recommends \
-     redis-server nginx-light
+    # Install supervisord with pip instead of apt, to avoid installing a second
+    # copy of python.
+    RUN --mount=type=cache,target=/root/.cache/pip \
+        pip install supervisor~=4.2
+    RUN mkdir -p /etc/supervisor/conf.d
 
-# Install supervisord with pip instead of apt, to avoid installing a second
-# copy of python.
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install supervisor~=4.2
+    # Copy over redis and nginx
+    COPY --from=redis_base /usr/local/bin/redis-server /usr/local/bin
 
-# Disable the default nginx sites
-RUN rm /etc/nginx/sites-enabled/default
+    COPY --from=deps_base /usr/sbin/nginx /usr/sbin
+    COPY --from=deps_base /usr/share/nginx /usr/share/nginx
+    COPY --from=deps_base /usr/lib/nginx /usr/lib/nginx
+    COPY --from=deps_base /etc/nginx /etc/nginx
+    RUN rm /etc/nginx/sites-enabled/default
+    RUN mkdir /var/log/nginx /var/lib/nginx
+    RUN chown www-data /var/log/nginx /var/lib/nginx
 
-# Copy Synapse worker, nginx and supervisord configuration template files
-COPY ./docker/conf-workers/* /conf/
+    # Copy Synapse worker, nginx and supervisord configuration template files
+    COPY ./docker/conf-workers/* /conf/
 
-# Copy a script to prefix log lines with the supervisor program name
-COPY ./docker/prefix-log /usr/local/bin/
+    # Copy a script to prefix log lines with the supervisor program name
+    COPY ./docker/prefix-log /usr/local/bin/
 
-# Expose nginx listener port
-EXPOSE 8080/tcp
+    # Expose nginx listener port
+    EXPOSE 8080/tcp
 
-# A script to read environment variables and create the necessary
-# files to run the desired worker configuration. Will start supervisord.
-COPY ./docker/configure_workers_and_start.py /configure_workers_and_start.py
-ENTRYPOINT ["/configure_workers_and_start.py"]
+    # A script to read environment variables and create the necessary
+    # files to run the desired worker configuration. Will start supervisord.
+    COPY ./docker/configure_workers_and_start.py /configure_workers_and_start.py
+    ENTRYPOINT ["/configure_workers_and_start.py"]
 
-# Replace the healthcheck with one which checks *all* the workers. The script
-# is generated by configure_workers_and_start.py.
-HEALTHCHECK --start-period=5s --interval=15s --timeout=5s \
-    CMD /bin/sh /healthcheck.sh
+    # Replace the healthcheck with one which checks *all* the workers. The script
+    # is generated by configure_workers_and_start.py.
+    HEALTHCHECK --start-period=5s --interval=15s --timeout=5s \
+        CMD /bin/sh /healthcheck.sh

docker/conf-workers/supervisord.conf.j2

@@ -19,7 +19,7 @@ username=www-data
 autorestart=true
 
 [program:redis]
-command=/usr/local/bin/prefix-log /usr/bin/redis-server /etc/redis/redis.conf --daemonize no
+command=/usr/local/bin/prefix-log /usr/local/bin/redis-server
 priority=1
 stdout_logfile=/dev/stdout
 stdout_logfile_maxbytes=0

docs/modules/password_auth_provider_callbacks.md

@@ -263,7 +263,7 @@ class MyAuthProvider:
             return None
 
         if self.credentials.get(username) == login_dict.get("my_field"):
-            return self.api.get_qualified_user_id(username)
+            return (self.api.get_qualified_user_id(username), None)
 
     async def check_pass(
         self,
@@ -280,5 +280,5 @@ class MyAuthProvider:
             return None
 
         if self.credentials.get(username) == login_dict.get("password"):
-            return self.api.get_qualified_user_id(username)
+            return (self.api.get_qualified_user_id(username), None)
 ```

docs/sso_mapping_providers.md

@@ -22,7 +22,7 @@ choose their own username.
 In the first case - where users are automatically allocated a Matrix ID - it is
 the responsibility of the mapping provider to normalise the SSO attributes and
 map them to a valid Matrix ID. The [specification for Matrix
-IDs](https://matrix.org/docs/spec/appendices#user-identifiers) has some
+IDs](https://spec.matrix.org/latest/appendices/#user-identifiers) has some
 information about what is considered valid.
 
 If the mapping provider does not assign a Matrix ID, then Synapse will
@@ -37,9 +37,10 @@ as Synapse). The Synapse config is then modified to point to the mapping provide
 ## OpenID Mapping Providers
 
 The OpenID mapping provider can be customized by editing the
-`oidc_config.user_mapping_provider.module` config option.
+[`oidc_providers.user_mapping_provider.module`](usage/configuration/config_documentation.md#oidc_providers)
+config option.
 
-`oidc_config.user_mapping_provider.config` allows you to provide custom
+`oidc_providers.user_mapping_provider.config` allows you to provide custom
 configuration options to the module. Check with the module's documentation for
 what options it provides (if any). The options listed by default are for the
 user mapping provider built in to Synapse. If using a custom module, you should
@@ -58,7 +59,7 @@ A custom mapping provider must specify the following methods:
     - This method should have the `@staticmethod` decoration.
     - Arguments:
         - `config` - A `dict` representing the parsed content of the
-          `oidc_config.user_mapping_provider.config` homeserver config option.
+          `oidc_providers.user_mapping_provider.config` homeserver config option.
            Runs on homeserver startup. Providers should extract and validate
            any option values they need here.
     - Whatever is returned will be passed back to the user mapping provider module's
@@ -102,7 +103,7 @@ A custom mapping provider must specify the following methods:
       will be returned as part of the response during a successful login.
 
       Note that care should be taken to not overwrite any of the parameters
-      usually returned as part of the [login response](https://matrix.org/docs/spec/client_server/latest#post-matrix-client-r0-login).
+      usually returned as part of the [login response](https://spec.matrix.org/latest/client-server-api/#post_matrixclientv3login).
 
 ### Default OpenID Mapping Provider
 
@@ -113,7 +114,8 @@ specified in the config. It is located at
 ## SAML Mapping Providers
 
 The SAML mapping provider can be customized by editing the
-`saml2_config.user_mapping_provider.module` config option.
+[`saml2_config.user_mapping_provider.module`](docs/usage/configuration/config_documentation.md#saml2_config)
+config option.
 
 `saml2_config.user_mapping_provider.config` allows you to provide custom
 configuration options to the module. Check with the module's documentation for

docs/usage/administration/admin_api/README.md

@@ -5,8 +5,9 @@
 Many of the API calls in the admin api will require an `access_token` for a
 server admin. (Note that a server admin is distinct from a room admin.)
 
-A user can be marked as a server admin by updating the database directly, e.g.:
+An existing user can be marked as a server admin by updating the database directly.
 
+Check your [database settings](config_documentation.md#database) in the configuration file, connect to the correct database using either `psql [database name]` (if using PostgreSQL) or `sqlite3 path/to/your/database.db` (if using SQLite) and elevate the user `@foo:bar.com` to administrator.
 ```sql
 UPDATE users SET admin = 1 WHERE name = '@foo:bar.com';
 ```

docs/usage/configuration/config_documentation.md

@@ -72,49 +72,6 @@ apply if you want your config file to be read properly. A few helpful things to
   In addition, each setting has an example of its usage, with the proper indentation
   shown.
 
-## Contents
-[Modules](#modules)
-
-[Server](#server)
-
-[Homeserver Blocking](#homeserver-blocking)
-
-[TLS](#tls)
-
-[Federation](#federation)
-
-[Caching](#caching)
-
-[Database](#database)
-
-[Logging](#logging)
-
-[Ratelimiting](#ratelimiting)
-
-[Media Store](#media-store)
-
-[Captcha](#captcha)
-
-[TURN](#turn)
-
-[Registration](#registration)
-
-[API Configuration](#api-configuration)
-
-[Signing Keys](#signing-keys)
-
-[Single Sign On Integration](#single-sign-on-integration)
-
-[Push](#push)
-
-[Rooms](#rooms)
-
-[Opentracing](#opentracing)
-
-[Workers](#workers)
-
-[Background Updates](#background-updates)
-
 ## Modules
 
 Server admins can expand Synapse's functionality with external modules.
@@ -486,7 +443,8 @@ Sub-options for each listener include:
 
    * `names`: a list of names of HTTP resources. See below for a list of valid resource names.
 
-   * `compress`: set to true to enable HTTP compression for this resource.
+   * `compress`: set to true to enable gzip compression on HTTP bodies for this resource. This is currently only supported with the
+     `client`, `consent` and `metrics` resources.
 
 * `additional_resources`: Only valid for an 'http' listener. A map of
    additional endpoints which should be loaded via dynamic modules.
@@ -1858,7 +1816,7 @@ Example configuration:
 max_spider_size: 8M
 ```
 ---
-### `url_preview_language`
+### `url_preview_accept_language`
 
 A list of values for the Accept-Language HTTP header used when
 downloading webpages during URL preview generation. This allows
@@ -2537,9 +2495,13 @@ track_appservice_user_ips: true
 ---
 ### `macaroon_secret_key`
 
-A secret which is used to sign access tokens. If none is specified,
-the `registration_shared_secret` is used, if one is given; otherwise,
-a secret key is derived from the signing key.
+A secret which is used to sign
+- access token for guest users,
+- short-term login token used during SSO logins (OIDC or SAML2) and
+- token used for unsubscribing from email notifications.
+
+If none is specified, the `registration_shared_secret` is used, if one is given;
+otherwise, a secret key is derived from the signing key.
 
 Example configuration:
 ```yaml

scripts-dev/complement.sh

@@ -101,6 +101,7 @@ if [ -z "$skip_docker_build" ]; then
     echo_if_github "::group::Build Docker image: matrixdotorg/synapse"
     docker build -t matrixdotorg/synapse \
       --build-arg TEST_ONLY_SKIP_DEP_HASH_VERIFICATION \
+      --build-arg TEST_ONLY_IGNORE_POETRY_LOCKFILE \
       -f "docker/Dockerfile" .
     echo_if_github "::endgroup::"
 

scripts-dev/release.py

@@ -32,6 +32,7 @@ import click
 import commonmark
 import git
 from click.exceptions import ClickException
+from git import GitCommandError, Repo
 from github import Github
 from packaging import version
 
@@ -55,9 +56,12 @@ def run_until_successful(
 def cli() -> None:
     """An interactive script to walk through the parts of creating a release.
 
-    Requires the dev dependencies be installed, which can be done via:
+    Requirements:
+      - The dev dependencies be installed, which can be done via:
 
-        pip install -e .[dev]
+            pip install -e .[dev]
+
+      - A checkout of the sytest repository at ../sytest
 
     Then to use:
 
@@ -75,6 +79,8 @@ def cli() -> None:
 
         # Optional: generate some nice links for the announcement
 
+        ./scripts-dev/release.py merge-back
+
         ./scripts-dev/release.py announce
 
     If the env var GH_TOKEN (or GITHUB_TOKEN) is set, or passed into the
@@ -89,10 +95,12 @@ def prepare() -> None:
     """
 
     # Make sure we're in a git repo.
-    repo = get_repo_and_check_clean_checkout()
+    synapse_repo = get_repo_and_check_clean_checkout()
+    sytest_repo = get_repo_and_check_clean_checkout("../sytest", "sytest")
 
-    click.secho("Updating git repo...")
-    repo.remote().fetch()
+    click.secho("Updating Synapse and Sytest git repos...")
+    synapse_repo.remote().fetch()
+    sytest_repo.remote().fetch()
 
     # Get the current version and AST from root Synapse module.
     current_version = get_package_version()
@@ -166,12 +174,12 @@ def prepare() -> None:
     assert not parsed_new_version.is_postrelease
 
     release_branch_name = get_release_branch_name(parsed_new_version)
-    release_branch = find_ref(repo, release_branch_name)
+    release_branch = find_ref(synapse_repo, release_branch_name)
     if release_branch:
         if release_branch.is_remote():
             # If the release branch only exists on the remote we check it out
             # locally.
-            repo.git.checkout(release_branch_name)
+            synapse_repo.git.checkout(release_branch_name)
     else:
         # If a branch doesn't exist we create one. We ask which one branch it
         # should be based off, defaulting to sensible values depending on the
@@ -187,25 +195,34 @@ def prepare() -> None:
             "Which branch should the release be based on?", default=default
         )
 
-        base_branch = find_ref(repo, branch_name)
-        if not base_branch:
-            print(f"Could not find base branch {branch_name}!")
-            click.get_current_context().abort()
+        for repo_name, repo in {"synapse": synapse_repo, "sytest": sytest_repo}.items():
+            base_branch = find_ref(repo, branch_name)
+            if not base_branch:
+                print(f"Could not find base branch {branch_name} for {repo_name}!")
+                click.get_current_context().abort()
 
-        # Check out the base branch and ensure it's up to date
-        repo.head.set_reference(base_branch, "check out the base branch")
-        repo.head.reset(index=True, working_tree=True)
-        if not base_branch.is_remote():
-            update_branch(repo)
+            # Check out the base branch and ensure it's up to date
+            repo.head.set_reference(
+                base_branch, f"check out the base branch for {repo_name}"
+            )
+            repo.head.reset(index=True, working_tree=True)
+            if not base_branch.is_remote():
+                update_branch(repo)
 
-        # Create the new release branch
-        # Type ignore will no longer be needed after GitPython 3.1.28.
-        # See https://github.com/gitpython-developers/GitPython/pull/1419
-        repo.create_head(release_branch_name, commit=base_branch)  # type: ignore[arg-type]
+            # Create the new release branch
+            # Type ignore will no longer be needed after GitPython 3.1.28.
+            # See https://github.com/gitpython-developers/GitPython/pull/1419
+            repo.create_head(release_branch_name, commit=base_branch)  # type: ignore[arg-type]
+
+        # Special-case SyTest: we don't actually prepare any files so we may
+        # as well push it now (and only when we create a release branch;
+        # not on subsequent RCs or full releases).
+        if click.confirm("Push new SyTest branch?", default=True):
+            sytest_repo.git.push("-u", sytest_repo.remote().name, release_branch_name)
 
     # Switch to the release branch and ensure it's up to date.
-    repo.git.checkout(release_branch_name)
-    update_branch(repo)
+    synapse_repo.git.checkout(release_branch_name)
+    update_branch(synapse_repo)
 
     # Update the version specified in pyproject.toml.
     subprocess.check_output(["poetry", "version", new_version])
@@ -230,15 +247,15 @@ def prepare() -> None:
     run_until_successful('dch -M -r -D stable ""', shell=True)
 
     # Show the user the changes and ask if they want to edit the change log.
-    repo.git.add("-u")
+    synapse_repo.git.add("-u")
     subprocess.run("git diff --cached", shell=True)
 
     if click.confirm("Edit changelog?", default=False):
         click.edit(filename="CHANGES.md")
 
     # Commit the changes.
-    repo.git.add("-u")
-    repo.git.commit("-m", new_version)
+    synapse_repo.git.add("-u")
+    synapse_repo.git.commit("-m", new_version)
 
     # We give the option to bail here in case the user wants to make sure things
     # are OK before pushing.
@@ -246,17 +263,21 @@ def prepare() -> None:
         print("")
         print("Run when ready to push:")
         print("")
-        print(f"\tgit push -u {repo.remote().name} {repo.active_branch.name}")
+        print(
+            f"\tgit push -u {synapse_repo.remote().name} {synapse_repo.active_branch.name}"
+        )
         print("")
         sys.exit(0)
 
     # Otherwise, push and open the changelog in the browser.
-    repo.git.push("-u", repo.remote().name, repo.active_branch.name)
+    synapse_repo.git.push(
+        "-u", synapse_repo.remote().name, synapse_repo.active_branch.name
+    )
 
     print("Opening the changelog in your browser...")
     print("Please ask others to give it a check.")
     click.launch(
-        f"https://github.com/matrix-org/synapse/blob/{repo.active_branch.name}/CHANGES.md"
+        f"https://github.com/matrix-org/synapse/blob/{synapse_repo.active_branch.name}/CHANGES.md"
     )
 
 
@@ -423,6 +444,79 @@ def upload() -> None:
     )
 
 
+def _merge_into(repo: Repo, source: str, target: str) -> None:
+    """
+    Merges branch `source` into branch `target`.
+    Pulls both before merging and pushes the result.
+    """
+
+    # Update our branches and switch to the target branch
+    for branch in [source, target]:
+        click.echo(f"Switching to {branch} and pulling...")
+        repo.heads[branch].checkout()
+        # Pull so we're up to date
+        repo.remote().pull()
+
+    assert repo.active_branch.name == target
+
+    try:
+        # TODO This seemed easier than using GitPython directly
+        click.echo(f"Merging {source}...")
+        repo.git.merge(source)
+    except GitCommandError as exc:
+        # If a merge conflict occurs, give some context and try to
+        # make it easy to abort if necessary.
+        click.echo(exc)
+        if not click.confirm(
+            f"Likely merge conflict whilst merging ({source} → {target}). "
+            f"Have you resolved it?"
+        ):
+            repo.git.merge("--abort")
+            return
+
+    # Push result.
+    click.echo("Pushing...")
+    repo.remote().push()
+
+
+@cli.command()
+def merge_back() -> None:
+    """Merge the release branch back into the appropriate branches.
+    All branches will be automatically pulled from the remote and the results
+    will be pushed to the remote."""
+
+    synapse_repo = get_repo_and_check_clean_checkout()
+    branch_name = synapse_repo.active_branch.name
+
+    if not branch_name.startswith("release-v"):
+        raise RuntimeError("Not on a release branch. This does not seem sensible.")
+
+    # Pull so we're up to date
+    synapse_repo.remote().pull()
+
+    current_version = get_package_version()
+
+    if current_version.is_prerelease:
+        # Release candidate
+        if click.confirm(f"Merge {branch_name} → develop?", default=True):
+            _merge_into(synapse_repo, branch_name, "develop")
+    else:
+        # Full release
+        sytest_repo = get_repo_and_check_clean_checkout("../sytest", "sytest")
+
+        if click.confirm(f"Merge {branch_name} → master?", default=True):
+            _merge_into(synapse_repo, branch_name, "master")
+
+        if click.confirm("Merge master → develop?", default=True):
+            _merge_into(synapse_repo, "master", "develop")
+
+        if click.confirm(f"On SyTest, merge {branch_name} → master?", default=True):
+            _merge_into(sytest_repo, branch_name, "master")
+
+        if click.confirm("On SyTest, merge master → develop?", default=True):
+            _merge_into(sytest_repo, "master", "develop")
+
+
 @cli.command()
 def announce() -> None:
     """Generate markdown to announce the release."""
@@ -469,14 +563,18 @@ def get_release_branch_name(version_number: version.Version) -> str:
     return f"release-v{version_number.major}.{version_number.minor}"
 
 
-def get_repo_and_check_clean_checkout() -> git.Repo:
+def get_repo_and_check_clean_checkout(
+    path: str = ".", name: str = "synapse"
+) -> git.Repo:
     """Get the project repo and check it's not got any uncommitted changes."""
     try:
-        repo = git.Repo()
+        repo = git.Repo(path=path)
     except git.InvalidGitRepositoryError:
-        raise click.ClickException("Not in Synapse repo.")
+        raise click.ClickException(
+            f"{path} is not a git repository (expecting a {name} repository)."
+        )
     if repo.is_dirty():
-        raise click.ClickException("Uncommitted changes exist.")
+        raise click.ClickException(f"Uncommitted changes exist in {path}.")
     return repo
 
 

synapse/api/auth.py

@@ -26,11 +26,17 @@ from synapse.api.errors import (
     Codes,
     InvalidClientTokenError,
     MissingClientTokenError,
+    UnstableSpecAuthError,
 )
 from synapse.appservice import ApplicationService
 from synapse.http import get_request_user_agent
 from synapse.http.site import SynapseRequest
-from synapse.logging.opentracing import active_span, force_tracing, start_active_span
+from synapse.logging.opentracing import (
+    active_span,
+    force_tracing,
+    start_active_span,
+    trace,
+)
 from synapse.storage.databases.main.registration import TokenLookupResult
 from synapse.types import Requester, UserID, create_requester
 
@@ -106,8 +112,11 @@ class Auth:
                 forgot = await self.store.did_forget(user_id, room_id)
                 if not forgot:
                     return membership, member_event_id
-
-        raise AuthError(403, "User %s not in room %s" % (user_id, room_id))
+        raise UnstableSpecAuthError(
+            403,
+            "User %s not in room %s" % (user_id, room_id),
+            errcode=Codes.NOT_JOINED,
+        )
 
     async def get_user_by_req(
         self,
@@ -563,6 +572,7 @@ class Auth:
 
             return query_params[0].decode("ascii")
 
+    @trace
     async def check_user_in_room_or_world_readable(
         self, room_id: str, user_id: str, allow_departed_users: bool = False
     ) -> Tuple[str, Optional[str]]:
@@ -600,8 +610,9 @@ class Auth:
                 == HistoryVisibility.WORLD_READABLE
             ):
                 return Membership.JOIN, None
-            raise AuthError(
+            raise UnstableSpecAuthError(
                 403,
                 "User %s not in room %s, and room previews are disabled"
                 % (user_id, room_id),
+                errcode=Codes.NOT_JOINED,
             )

synapse/api/constants.py

@@ -268,4 +268,4 @@ class PublicRoomsFilterFields:
     """
 
     GENERIC_SEARCH_TERM: Final = "generic_search_term"
-    ROOM_TYPES: Final = "org.matrix.msc3827.room_types"
+    ROOM_TYPES: Final = "room_types"

synapse/api/errors.py

@@ -26,6 +26,7 @@ from twisted.web import http
 from synapse.util import json_decoder
 
 if typing.TYPE_CHECKING:
+    from synapse.config.homeserver import HomeServerConfig
     from synapse.types import JsonDict
 
 logger = logging.getLogger(__name__)
@@ -80,6 +81,12 @@ class Codes(str, Enum):
     INVALID_SIGNATURE = "M_INVALID_SIGNATURE"
     USER_DEACTIVATED = "M_USER_DEACTIVATED"
 
+    # Part of MSC3848
+    # https://github.com/matrix-org/matrix-spec-proposals/pull/3848
+    ALREADY_JOINED = "ORG.MATRIX.MSC3848.ALREADY_JOINED"
+    NOT_JOINED = "ORG.MATRIX.MSC3848.NOT_JOINED"
+    INSUFFICIENT_POWER = "ORG.MATRIX.MSC3848.INSUFFICIENT_POWER"
+
     # The account has been suspended on the server.
     # By opposition to `USER_DEACTIVATED`, this is a reversible measure
     # that can possibly be appealed and reverted.
@@ -167,7 +174,7 @@ class SynapseError(CodeMessageException):
         else:
             self._additional_fields = dict(additional_fields)
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, **self._additional_fields)
 
 
@@ -213,7 +220,7 @@ class ConsentNotGivenError(SynapseError):
         )
         self._consent_uri = consent_uri
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, consent_uri=self._consent_uri)
 
 
@@ -307,6 +314,37 @@ class AuthError(SynapseError):
         super().__init__(code, msg, errcode, additional_fields)
 
 
+class UnstableSpecAuthError(AuthError):
+    """An error raised when a new error code is being proposed to replace a previous one.
+    This error will return a "org.matrix.unstable.errcode" property with the new error code,
+    with the previous error code still being defined in the "errcode" property.
+
+    This error will include `org.matrix.msc3848.unstable.errcode` in the C-S error body.
+    """
+
+    def __init__(
+        self,
+        code: int,
+        msg: str,
+        errcode: str,
+        previous_errcode: str = Codes.FORBIDDEN,
+        additional_fields: Optional[dict] = None,
+    ):
+        self.previous_errcode = previous_errcode
+        super().__init__(code, msg, errcode, additional_fields)
+
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
+        fields = {}
+        if config is not None and config.experimental.msc3848_enabled:
+            fields["org.matrix.msc3848.unstable.errcode"] = self.errcode
+        return cs_error(
+            self.msg,
+            self.previous_errcode,
+            **fields,
+            **self._additional_fields,
+        )
+
+
 class InvalidClientCredentialsError(SynapseError):
     """An error raised when there was a problem with the authorisation credentials
     in a client request.
@@ -338,8 +376,8 @@ class InvalidClientTokenError(InvalidClientCredentialsError):
         super().__init__(msg=msg, errcode="M_UNKNOWN_TOKEN")
         self._soft_logout = soft_logout
 
-    def error_dict(self) -> "JsonDict":
-        d = super().error_dict()
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
+        d = super().error_dict(config)
         d["soft_logout"] = self._soft_logout
         return d
 
@@ -362,7 +400,7 @@ class ResourceLimitError(SynapseError):
         self.limit_type = limit_type
         super().__init__(code, msg, errcode=errcode)
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(
             self.msg,
             self.errcode,
@@ -397,7 +435,7 @@ class InvalidCaptchaError(SynapseError):
         super().__init__(code, msg, errcode)
         self.error_url = error_url
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, error_url=self.error_url)
 
 
@@ -414,7 +452,7 @@ class LimitExceededError(SynapseError):
         super().__init__(code, msg, errcode)
         self.retry_after_ms = retry_after_ms
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, retry_after_ms=self.retry_after_ms)
 
 
@@ -429,7 +467,7 @@ class RoomKeysVersionError(SynapseError):
         super().__init__(403, "Wrong room_keys version", Codes.WRONG_ROOM_KEYS_VERSION)
         self.current_version = current_version
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, current_version=self.current_version)
 
 
@@ -469,7 +507,7 @@ class IncompatibleRoomVersionError(SynapseError):
 
         self._room_version = room_version
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         return cs_error(self.msg, self.errcode, room_version=self._room_version)
 
 
@@ -515,7 +553,7 @@ class UnredactedContentDeletedError(SynapseError):
         )
         self.content_keep_ms = content_keep_ms
 
-    def error_dict(self) -> "JsonDict":
+    def error_dict(self, config: Optional["HomeServerConfig"]) -> "JsonDict":
         extra = {}
         if self.content_keep_ms is not None:
             extra = {"fi.mau.msc2815.content_keep_ms": self.content_keep_ms}

synapse/api/ratelimiting.py

@@ -17,7 +17,7 @@ from collections import OrderedDict
 from typing import Hashable, Optional, Tuple
 
 from synapse.api.errors import LimitExceededError
-from synapse.config.ratelimiting import RateLimitConfig
+from synapse.config.ratelimiting import RatelimitSettings
 from synapse.storage.databases.main import DataStore
 from synapse.types import Requester
 from synapse.util import Clock
@@ -314,8 +314,8 @@ class RequestRatelimiter:
         self,
         store: DataStore,
         clock: Clock,
-        rc_message: RateLimitConfig,
-        rc_admin_redaction: Optional[RateLimitConfig],
+        rc_message: RatelimitSettings,
+        rc_admin_redaction: Optional[RatelimitSettings],
     ):
         self.store = store
         self.clock = clock

synapse/config/experimental.py

@@ -88,5 +88,5 @@ class ExperimentalConfig(Config):
         # MSC3715: dir param on /relations.
         self.msc3715_enabled: bool = experimental.get("msc3715_enabled", False)
 
-        # MSC3827: Filtering of /publicRooms by room type
-        self.msc3827_enabled: bool = experimental.get("msc3827_enabled", False)
+        # MSC3848: Introduce errcodes for specific event sending failures
+        self.msc3848_enabled: bool = experimental.get("msc3848_enabled", False)

synapse/config/ratelimiting.py

@@ -21,7 +21,7 @@ from synapse.types import JsonDict
 from ._base import Config
 
 
-class RateLimitConfig:
+class RatelimitSettings:
     def __init__(
         self,
         config: Dict[str, float],
@@ -34,7 +34,7 @@ class RateLimitConfig:
 
 
 @attr.s(auto_attribs=True)
-class FederationRateLimitConfig:
+class FederationRatelimitSettings:
     window_size: int = 1000
     sleep_limit: int = 10
     sleep_delay: int = 500
@@ -50,11 +50,11 @@ class RatelimitConfig(Config):
         # Load the new-style messages config if it exists. Otherwise fall back
         # to the old method.
         if "rc_message" in config:
-            self.rc_message = RateLimitConfig(
+            self.rc_message = RatelimitSettings(
                 config["rc_message"], defaults={"per_second": 0.2, "burst_count": 10.0}
             )
         else:
-            self.rc_message = RateLimitConfig(
+            self.rc_message = RatelimitSettings(
                 {
                     "per_second": config.get("rc_messages_per_second", 0.2),
                     "burst_count": config.get("rc_message_burst_count", 10.0),
@@ -64,9 +64,9 @@ class RatelimitConfig(Config):
         # Load the new-style federation config, if it exists. Otherwise, fall
         # back to the old method.
         if "rc_federation" in config:
-            self.rc_federation = FederationRateLimitConfig(**config["rc_federation"])
+            self.rc_federation = FederationRatelimitSettings(**config["rc_federation"])
         else:
-            self.rc_federation = FederationRateLimitConfig(
+            self.rc_federation = FederationRatelimitSettings(
                 **{
                     k: v
                     for k, v in {
@@ -80,17 +80,17 @@ class RatelimitConfig(Config):
                 }
             )
 
-        self.rc_registration = RateLimitConfig(config.get("rc_registration", {}))
+        self.rc_registration = RatelimitSettings(config.get("rc_registration", {}))
 
-        self.rc_registration_token_validity = RateLimitConfig(
+        self.rc_registration_token_validity = RatelimitSettings(
             config.get("rc_registration_token_validity", {}),
             defaults={"per_second": 0.1, "burst_count": 5},
         )
 
         rc_login_config = config.get("rc_login", {})
-        self.rc_login_address = RateLimitConfig(rc_login_config.get("address", {}))
-        self.rc_login_account = RateLimitConfig(rc_login_config.get("account", {}))
-        self.rc_login_failed_attempts = RateLimitConfig(
+        self.rc_login_address = RatelimitSettings(rc_login_config.get("address", {}))
+        self.rc_login_account = RatelimitSettings(rc_login_config.get("account", {}))
+        self.rc_login_failed_attempts = RatelimitSettings(
             rc_login_config.get("failed_attempts", {})
         )
 
@@ -101,20 +101,20 @@ class RatelimitConfig(Config):
         rc_admin_redaction = config.get("rc_admin_redaction")
         self.rc_admin_redaction = None
         if rc_admin_redaction:
-            self.rc_admin_redaction = RateLimitConfig(rc_admin_redaction)
+            self.rc_admin_redaction = RatelimitSettings(rc_admin_redaction)
 
-        self.rc_joins_local = RateLimitConfig(
+        self.rc_joins_local = RatelimitSettings(
             config.get("rc_joins", {}).get("local", {}),
             defaults={"per_second": 0.1, "burst_count": 10},
         )
-        self.rc_joins_remote = RateLimitConfig(
+        self.rc_joins_remote = RatelimitSettings(
             config.get("rc_joins", {}).get("remote", {}),
             defaults={"per_second": 0.01, "burst_count": 10},
         )
 
         # Track the rate of joins to a given room. If there are too many, temporarily
         # prevent local joins and remote joins via this server.
-        self.rc_joins_per_room = RateLimitConfig(
+        self.rc_joins_per_room = RatelimitSettings(
             config.get("rc_joins_per_room", {}),
             defaults={"per_second": 1, "burst_count": 10},
         )
@@ -124,31 +124,31 @@ class RatelimitConfig(Config):
         # * For requests received over federation this is keyed by the origin.
         #
         # Note that this isn't exposed in the configuration as it is obscure.
-        self.rc_key_requests = RateLimitConfig(
+        self.rc_key_requests = RatelimitSettings(
             config.get("rc_key_requests", {}),
             defaults={"per_second": 20, "burst_count": 100},
         )
 
-        self.rc_3pid_validation = RateLimitConfig(
+        self.rc_3pid_validation = RatelimitSettings(
             config.get("rc_3pid_validation") or {},
             defaults={"per_second": 0.003, "burst_count": 5},
         )
 
-        self.rc_invites_per_room = RateLimitConfig(
+        self.rc_invites_per_room = RatelimitSettings(
             config.get("rc_invites", {}).get("per_room", {}),
             defaults={"per_second": 0.3, "burst_count": 10},
         )
-        self.rc_invites_per_user = RateLimitConfig(
+        self.rc_invites_per_user = RatelimitSettings(
             config.get("rc_invites", {}).get("per_user", {}),
             defaults={"per_second": 0.003, "burst_count": 5},
         )
 
-        self.rc_invites_per_issuer = RateLimitConfig(
+        self.rc_invites_per_issuer = RatelimitSettings(
             config.get("rc_invites", {}).get("per_issuer", {}),
             defaults={"per_second": 0.3, "burst_count": 10},
         )
 
-        self.rc_third_party_invite = RateLimitConfig(
+        self.rc_third_party_invite = RatelimitSettings(
             config.get("rc_third_party_invite", {}),
             defaults={
                 "per_second": self.rc_message.per_second,

synapse/event_auth.py

@@ -30,7 +30,13 @@ from synapse.api.constants import (
     JoinRules,
     Membership,
 )
-from synapse.api.errors import AuthError, EventSizeError, SynapseError
+from synapse.api.errors import (
+    AuthError,
+    Codes,
+    EventSizeError,
+    SynapseError,
+    UnstableSpecAuthError,
+)
 from synapse.api.room_versions import (
     KNOWN_ROOM_VERSIONS,
     EventFormatVersions,
@@ -291,7 +297,11 @@ def check_state_dependent_auth_rules(
         invite_level = get_named_level(auth_dict, "invite", 0)
 
         if user_level < invite_level:
-            raise AuthError(403, "You don't have permission to invite users")
+            raise UnstableSpecAuthError(
+                403,
+                "You don't have permission to invite users",
+                errcode=Codes.INSUFFICIENT_POWER,
+            )
         else:
             logger.debug("Allowing! %s", event)
             return
@@ -474,7 +484,11 @@ def _is_membership_change_allowed(
             return
 
         if not caller_in_room:  # caller isn't joined
-            raise AuthError(403, "%s not in room %s." % (event.user_id, event.room_id))
+            raise UnstableSpecAuthError(
+                403,
+                "%s not in room %s." % (event.user_id, event.room_id),
+                errcode=Codes.NOT_JOINED,
+            )
 
     if Membership.INVITE == membership:
         # TODO (erikj): We should probably handle this more intelligently
@@ -484,10 +498,18 @@ def _is_membership_change_allowed(
         if target_banned:
             raise AuthError(403, "%s is banned from the room" % (target_user_id,))
         elif target_in_room:  # the target is already in the room.
-            raise AuthError(403, "%s is already in the room." % target_user_id)
+            raise UnstableSpecAuthError(
+                403,
+                "%s is already in the room." % target_user_id,
+                errcode=Codes.ALREADY_JOINED,
+            )
         else:
             if user_level < invite_level:
-                raise AuthError(403, "You don't have permission to invite users")
+                raise UnstableSpecAuthError(
+                    403,
+                    "You don't have permission to invite users",
+                    errcode=Codes.INSUFFICIENT_POWER,
+                )
     elif Membership.JOIN == membership:
         # Joins are valid iff caller == target and:
         # * They are not banned.
@@ -549,15 +571,27 @@ def _is_membership_change_allowed(
     elif Membership.LEAVE == membership:
         # TODO (erikj): Implement kicks.
         if target_banned and user_level < ban_level:
-            raise AuthError(403, "You cannot unban user %s." % (target_user_id,))
+            raise UnstableSpecAuthError(
+                403,
+                "You cannot unban user %s." % (target_user_id,),
+                errcode=Codes.INSUFFICIENT_POWER,
+            )
         elif target_user_id != event.user_id:
             kick_level = get_named_level(auth_events, "kick", 50)
 
             if user_level < kick_level or user_level <= target_level:
-                raise AuthError(403, "You cannot kick user %s." % target_user_id)
+                raise UnstableSpecAuthError(
+                    403,
+                    "You cannot kick user %s." % target_user_id,
+                    errcode=Codes.INSUFFICIENT_POWER,
+                )
     elif Membership.BAN == membership:
         if user_level < ban_level or user_level <= target_level:
-            raise AuthError(403, "You don't have permission to ban")
+            raise UnstableSpecAuthError(
+                403,
+                "You don't have permission to ban",
+                errcode=Codes.INSUFFICIENT_POWER,
+            )
     elif room_version.msc2403_knocking and Membership.KNOCK == membership:
         if join_rule != JoinRules.KNOCK and (
             not room_version.msc3787_knock_restricted_join_rule
@@ -567,7 +601,11 @@ def _is_membership_change_allowed(
         elif target_user_id != event.user_id:
             raise AuthError(403, "You cannot knock for other users")
         elif target_in_room:
-            raise AuthError(403, "You cannot knock on a room you are already in")
+            raise UnstableSpecAuthError(
+                403,
+                "You cannot knock on a room you are already in",
+                errcode=Codes.ALREADY_JOINED,
+            )
         elif caller_invited:
             raise AuthError(403, "You are already invited to this room")
         elif target_banned:
@@ -638,10 +676,11 @@ def _can_send_event(event: "EventBase", auth_events: StateMap["EventBase"]) -> b
     user_level = get_user_power_level(event.user_id, auth_events)
 
     if user_level < send_level:
-        raise AuthError(
+        raise UnstableSpecAuthError(
             403,
             "You don't have permission to post that to the room. "
             + "user_level (%d) < send_level (%d)" % (user_level, send_level),
+            errcode=Codes.INSUFFICIENT_POWER,
         )
 
     # Check state_key
@@ -716,9 +755,10 @@ def check_historical(
     historical_level = get_named_level(auth_events, "historical", 100)
 
     if user_level < historical_level:
-        raise AuthError(
+        raise UnstableSpecAuthError(
             403,
             'You don\'t have permission to send send historical related events ("insertion", "batch", and "marker")',
+            errcode=Codes.INSUFFICIENT_POWER,
         )
 
 

synapse/events/snapshot.py

@@ -11,11 +11,10 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import TYPE_CHECKING, List, Optional, Tuple, Union
+from typing import TYPE_CHECKING, List, Optional, Tuple
 
 import attr
 from frozendict import frozendict
-from typing_extensions import Literal
 
 from synapse.appservice import ApplicationService
 from synapse.events import EventBase
@@ -33,7 +32,7 @@ class EventContext:
     Holds information relevant to persisting an event
 
     Attributes:
-        rejected: A rejection reason if the event was rejected, else False
+        rejected: A rejection reason if the event was rejected, else None
 
         _state_group: The ID of the state group for this event. Note that state events
             are persisted with a state group which includes the new event, so this is
@@ -85,7 +84,7 @@ class EventContext:
     """
 
     _storage: "StorageControllers"
-    rejected: Union[Literal[False], str] = False
+    rejected: Optional[str] = None
     _state_group: Optional[int] = None
     state_group_before_event: Optional[int] = None
     _state_delta_due_to_event: Optional[StateMap[str]] = None

synapse/federation/federation_client.py

@@ -61,6 +61,7 @@ from synapse.federation.federation_base import (
 )
 from synapse.federation.transport.client import SendJoinResponse
 from synapse.http.types import QueryParams
+from synapse.logging.opentracing import trace
 from synapse.types import JsonDict, UserID, get_domain_from_id
 from synapse.util.async_helpers import concurrently_execute
 from synapse.util.caches.expiringcache import ExpiringCache
@@ -233,6 +234,7 @@ class FederationClient(FederationBase):
             destination, content, timeout
         )
 
+    @trace
     async def backfill(
         self, dest: str, room_id: str, limit: int, extremities: Collection[str]
     ) -> Optional[List[EventBase]]:
@@ -403,9 +405,9 @@ class FederationClient(FederationBase):
                         # Prime the cache
                         self._get_pdu_cache[event.event_id] = event
 
-                        # FIXME: We should add a `break` here to avoid calling every
-                        # destination after we already found a PDU (will follow-up
-                        # in a separate PR)
+                        # Now that we have an event, we can break out of this
+                        # loop and stop asking other destinations.
+                        break
 
                 except SynapseError as e:
                     logger.info(
@@ -725,6 +727,12 @@ class FederationClient(FederationBase):
         if failover_errcodes is None:
             failover_errcodes = ()
 
+        if not destinations:
+            # Give a bit of a clearer message if no servers were specified at all.
+            raise SynapseError(
+                502, f"Failed to {description} via any server: No servers specified."
+            )
+
         for destination in destinations:
             if destination == self.server_name:
                 continue
@@ -774,7 +782,7 @@ class FederationClient(FederationBase):
                     "Failed to %s via %s", description, destination, exc_info=True
                 )
 
-        raise SynapseError(502, "Failed to %s via any server" % (description,))
+        raise SynapseError(502, f"Failed to {description} via any server")
 
     async def make_membership_event(
         self,

synapse/federation/federation_server.py

@@ -469,7 +469,7 @@ class FederationServer(FederationBase):
                     )
                     for pdu in pdus_by_room[room_id]:
                         event_id = pdu.event_id
-                        pdu_results[event_id] = e.error_dict()
+                        pdu_results[event_id] = e.error_dict(self.hs.config)
                     return
 
                 for pdu in pdus_by_room[room_id]:
@@ -843,8 +843,25 @@ class FederationServer(FederationBase):
                 Codes.BAD_JSON,
             )
 
+        # Note that get_room_version throws if the room does not exist here.
         room_version = await self.store.get_room_version(room_id)
 
+        if await self.store.is_partial_state_room(room_id):
+            # If our server is still only partially joined, we can't give a complete
+            # response to /send_join, /send_knock or /send_leave.
+            # This is because we will not be able to provide the server list (for partial
+            # joins) or the full state (for full joins).
+            # Return a 404 as we would if we weren't in the room at all.
+            logger.info(
+                f"Rejecting /send_{membership_type} to %s because it's a partial state room",
+                room_id,
+            )
+            raise SynapseError(
+                404,
+                f"Unable to handle /send_{membership_type} right now; this server is not fully joined.",
+                errcode=Codes.NOT_FOUND,
+            )
+
         if membership_type == Membership.KNOCK and not room_version.msc2403_knocking:
             raise SynapseError(
                 403,

synapse/handlers/auth.py

@@ -565,7 +565,7 @@ class AuthHandler:
             except LoginError as e:
                 # this step failed. Merge the error dict into the response
                 # so that the client can have another go.
-                errordict = e.error_dict()
+                errordict = e.error_dict(self.hs.config)
 
         creds = await self.store.get_completed_ui_auth_stages(session.session_id)
         for f in flows:

synapse/handlers/federation.py

@@ -59,6 +59,7 @@ from synapse.events.validator import EventValidator
 from synapse.federation.federation_client import InvalidResponseError
 from synapse.http.servlet import assert_params_in_dict
 from synapse.logging.context import nested_logging_context
+from synapse.logging.opentracing import trace
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.module_api import NOT_SPAM
 from synapse.replication.http.federation import (
@@ -180,6 +181,7 @@ class FederationHandler:
                 "resume_sync_partial_state_room", self._resume_sync_partial_state_room
             )
 
+    @trace
     async def maybe_backfill(
         self, room_id: str, current_depth: int, limit: int
     ) -> bool:
@@ -546,9 +548,9 @@ class FederationHandler:
             )
 
             if ret.partial_state:
-                # TODO(faster_joins): roll this back if we don't manage to start the
-                #   background resync (eg process_remote_join fails)
-                #   https://github.com/matrix-org/synapse/issues/12998
+                # Mark the room as having partial state.
+                # The background process is responsible for unmarking this flag,
+                # even if the join fails.
                 await self.store.store_partial_state_room(room_id, ret.servers_in_room)
 
             try:
@@ -574,17 +576,21 @@ class FederationHandler:
                     room_id,
                 )
                 raise LimitExceededError(msg=e.msg, errcode=e.errcode, retry_after_ms=0)
-
-            if ret.partial_state:
-                # Kick off the process of asynchronously fetching the state for this
-                # room.
-                run_as_background_process(
-                    desc="sync_partial_state_room",
-                    func=self._sync_partial_state_room,
-                    initial_destination=origin,
-                    other_destinations=ret.servers_in_room,
-                    room_id=room_id,
-                )
+            finally:
+                # Always kick off the background process that asynchronously fetches
+                # state for the room.
+                # If the join failed, the background process is responsible for
+                # cleaning up — including unmarking the room as a partial state room.
+                if ret.partial_state:
+                    # Kick off the process of asynchronously fetching the state for this
+                    # room.
+                    run_as_background_process(
+                        desc="sync_partial_state_room",
+                        func=self._sync_partial_state_room,
+                        initial_destination=origin,
+                        other_destinations=ret.servers_in_room,
+                        room_id=room_id,
+                    )
 
             # We wait here until this instance has seen the events come down
             # replication (if we're using replication) as the below uses caches.
@@ -748,6 +754,23 @@ class FederationHandler:
         # (and return a 404 otherwise)
         room_version = await self.store.get_room_version(room_id)
 
+        if await self.store.is_partial_state_room(room_id):
+            # If our server is still only partially joined, we can't give a complete
+            # response to /make_join, so return a 404 as we would if we weren't in the
+            # room at all.
+            # The main reason we can't respond properly is that we need to know about
+            # the auth events for the join event that we would return.
+            # We also should not bother entertaining the /make_join since we cannot
+            # handle the /send_join.
+            logger.info(
+                "Rejecting /make_join to %s because it's a partial state room", room_id
+            )
+            raise SynapseError(
+                404,
+                "Unable to handle /make_join right now; this server is not fully joined.",
+                errcode=Codes.NOT_FOUND,
+            )
+
         # now check that we are *still* in the room
         is_in_room = await self._event_auth_handler.check_host_in_room(
             room_id, self.server_name
@@ -1539,15 +1562,16 @@ class FederationHandler:
 
         # Make an infinite iterator of destinations to try. Once we find a working
         # destination, we'll stick with it until it flakes.
+        destinations: Collection[str]
         if initial_destination is not None:
             # Move `initial_destination` to the front of the list.
             destinations = list(other_destinations)
             if initial_destination in destinations:
                 destinations.remove(initial_destination)
             destinations = [initial_destination] + destinations
-            destination_iter = itertools.cycle(destinations)
         else:
-            destination_iter = itertools.cycle(other_destinations)
+            destinations = other_destinations
+        destination_iter = itertools.cycle(destinations)
 
         # `destination` is the current remote homeserver we're pulling from.
         destination = next(destination_iter)

synapse/handlers/federation_event.py

@@ -59,6 +59,7 @@ from synapse.events import EventBase
 from synapse.events.snapshot import EventContext
 from synapse.federation.federation_client import InvalidResponseError
 from synapse.logging.context import nested_logging_context
+from synapse.logging.opentracing import trace
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet
 from synapse.replication.http.federation import (
@@ -278,7 +279,8 @@ class FederationEventHandler:
                 )
 
         try:
-            await self._process_received_pdu(origin, pdu, state_ids=None)
+            context = await self._state_handler.compute_event_context(pdu)
+            await self._process_received_pdu(origin, pdu, context)
         except PartialStateConflictError:
             # The room was un-partial stated while we were processing the PDU.
             # Try once more, with full state this time.
@@ -286,7 +288,8 @@ class FederationEventHandler:
                 "Room %s was un-partial stated while processing the PDU, trying again.",
                 room_id,
             )
-            await self._process_received_pdu(origin, pdu, state_ids=None)
+            context = await self._state_handler.compute_event_context(pdu)
+            await self._process_received_pdu(origin, pdu, context)
 
     async def on_send_membership_event(
         self, origin: str, event: EventBase
@@ -316,6 +319,7 @@ class FederationEventHandler:
             The event and context of the event after inserting it into the room graph.
 
         Raises:
+            RuntimeError if any prev_events are missing
             SynapseError if the event is not accepted into the room
             PartialStateConflictError if the room was un-partial stated in between
                 computing the state at the event and persisting it. The caller should
@@ -376,7 +380,7 @@ class FederationEventHandler:
         # need to.
         await self._event_creation_handler.cache_joined_hosts_for_event(event, context)
 
-        await self._check_for_soft_fail(event, None, origin=origin)
+        await self._check_for_soft_fail(event, context=context, origin=origin)
         await self._run_push_actions_and_persist_event(event, context)
         return event, context
 
@@ -534,32 +538,36 @@ class FederationEventHandler:
             #
             # This is the same operation as we do when we receive a regular event
             # over federation.
-            state_ids = await self._resolve_state_at_missing_prevs(destination, event)
-
-            # build a new state group for it if need be
-            context = await self._state_handler.compute_event_context(
-                event,
-                state_ids_before_event=state_ids,
+            context = await self._compute_event_context_with_maybe_missing_prevs(
+                destination, event
             )
             if context.partial_state:
                 # this can happen if some or all of the event's prev_events still have
-                # partial state - ie, an event has an earlier stream_ordering than one
-                # or more of its prev_events, so we de-partial-state it before its
-                # prev_events.
+                # partial state. We were careful to only pick events from the db without
+                # partial-state prev events, so that implies that a prev event has
+                # been persisted (with partial state) since we did the query.
                 #
-                # TODO(faster_joins): we probably need to be more intelligent, and
-                #    exclude partial-state prev_events from consideration
-                #    https://github.com/matrix-org/synapse/issues/13001
+                # So, let's just ignore `event` for now; when we re-run the db query
+                # we should instead get its partial-state prev event, which we will
+                # de-partial-state, and then come back to event.
                 logger.warning(
-                    "%s still has partial state: can't de-partial-state it yet",
+                    "%s still has prev_events with partial state: can't de-partial-state it yet",
                     event.event_id,
                 )
                 return
+
+            # since the state at this event has changed, we should now re-evaluate
+            # whether it should have been rejected. We must already have all of the
+            # auth events (from last time we went round this path), so there is no
+            # need to pass the origin.
+            await self._check_event_auth(None, event, context)
+
             await self._store.update_state_for_partial_state_event(event, context)
             self._state_storage_controller.notify_event_un_partial_stated(
                 event.event_id
             )
 
+    @trace
     async def backfill(
         self, dest: str, room_id: str, limit: int, extremities: Collection[str]
     ) -> None:
@@ -604,6 +612,7 @@ class FederationEventHandler:
             backfilled=True,
         )
 
+    @trace
     async def _get_missing_events_for_pdu(
         self, origin: str, pdu: EventBase, prevs: Set[str], min_depth: int
     ) -> None:
@@ -704,6 +713,7 @@ class FederationEventHandler:
         logger.info("Got %d prev_events", len(missing_events))
         await self._process_pulled_events(origin, missing_events, backfilled=False)
 
+    @trace
     async def _process_pulled_events(
         self, origin: str, events: Iterable[EventBase], backfilled: bool
     ) -> None:
@@ -742,6 +752,7 @@ class FederationEventHandler:
             with nested_logging_context(ev.event_id):
                 await self._process_pulled_event(origin, ev, backfilled=backfilled)
 
+    @trace
     async def _process_pulled_event(
         self, origin: str, event: EventBase, backfilled: bool
     ) -> None:
@@ -806,29 +817,55 @@ class FederationEventHandler:
             return
 
         try:
-            state_ids = await self._resolve_state_at_missing_prevs(origin, event)
-            # TODO(faster_joins): make sure that _resolve_state_at_missing_prevs does
-            #   not return partial state
-            #   https://github.com/matrix-org/synapse/issues/13002
+            try:
+                context = await self._compute_event_context_with_maybe_missing_prevs(
+                    origin, event
+                )
+                await self._process_received_pdu(
+                    origin,
+                    event,
+                    context,
+                    backfilled=backfilled,
+                )
+            except PartialStateConflictError:
+                # The room was un-partial stated while we were processing the event.
+                # Try once more, with full state this time.
+                context = await self._compute_event_context_with_maybe_missing_prevs(
+                    origin, event
+                )
 
-            await self._process_received_pdu(
-                origin, event, state_ids=state_ids, backfilled=backfilled
-            )
+                # We ought to have full state now, barring some unlikely race where we left and
+                # rejoned the room in the background.
+                if context.partial_state:
+                    raise AssertionError(
+                        f"Event {event.event_id} still has a partial resolved state "
+                        f"after room {event.room_id} was un-partial stated"
+                    )
+
+                await self._process_received_pdu(
+                    origin,
+                    event,
+                    context,
+                    backfilled=backfilled,
+                )
         except FederationError as e:
             if e.code == 403:
                 logger.warning("Pulled event %s failed history check.", event_id)
             else:
                 raise
 
-    async def _resolve_state_at_missing_prevs(
+    async def _compute_event_context_with_maybe_missing_prevs(
         self, dest: str, event: EventBase
-    ) -> Optional[StateMap[str]]:
-        """Calculate the state at an event with missing prev_events.
+    ) -> EventContext:
+        """Build an EventContext structure for a non-outlier event whose prev_events may
+        be missing.
 
-        This is used when we have pulled a batch of events from a remote server, and
-        still don't have all the prev_events.
+        This is used when we have pulled a batch of events from a remote server, and may
+        not have all the prev_events.
 
-        If we already have all the prev_events for `event`, this method does nothing.
+        To build an EventContext, we need to calculate the state before the event. If we
+        already have all the prev_events for `event`, we can simply use the state after
+        the prev_events to calculate the state before `event`.
 
         Otherwise, the missing prevs become new backwards extremities, and we fall back
         to asking the remote server for the state after each missing `prev_event`,
@@ -849,8 +886,7 @@ class FederationEventHandler:
             event: an event to check for missing prevs.
 
         Returns:
-            if we already had all the prev events, `None`. Otherwise, returns
-            the event ids of the state at `event`.
+            The event context.
 
         Raises:
             FederationError if we fail to get the state from the remote server after any
@@ -864,7 +900,7 @@ class FederationEventHandler:
         missing_prevs = prevs - seen
 
         if not missing_prevs:
-            return None
+            return await self._state_handler.compute_event_context(event)
 
         logger.info(
             "Event %s is missing prev_events %s: calculating state for a "
@@ -876,9 +912,15 @@ class FederationEventHandler:
         # resolve them to find the correct state at the current event.
 
         try:
+            # Determine whether we may be about to retrieve partial state
+            # Events may be un-partial stated right after we compute the partial state
+            # flag, but that's okay, as long as the flag errs on the conservative side.
+            partial_state_flags = await self._store.get_partial_state_events(seen)
+            partial_state = any(partial_state_flags.values())
+
             # Get the state of the events we know about
             ours = await self._state_storage_controller.get_state_groups_ids(
-                room_id, seen
+                room_id, seen, await_full_state=False
             )
 
             # state_maps is a list of mappings from (type, state_key) to event_id
@@ -924,7 +966,9 @@ class FederationEventHandler:
                 "We can't get valid state history.",
                 affected=event_id,
             )
-        return state_map
+        return await self._state_handler.compute_event_context(
+            event, state_ids_before_event=state_map, partial_state=partial_state
+        )
 
     async def _get_state_ids_after_missing_prev_event(
         self,
@@ -1093,7 +1137,7 @@ class FederationEventHandler:
         self,
         origin: str,
         event: EventBase,
-        state_ids: Optional[StateMap[str]],
+        context: EventContext,
         backfilled: bool = False,
     ) -> None:
         """Called when we have a new non-outlier event.
@@ -1115,24 +1159,18 @@ class FederationEventHandler:
 
             event: event to be persisted
 
-            state_ids: Normally None, but if we are handling a gap in the graph
-                (ie, we are missing one or more prev_events), the resolved state at the
-                event. Must not be partial state.
+            context: The `EventContext` to persist the event with.
 
             backfilled: True if this is part of a historical batch of events (inhibits
                 notification to clients, and validation of device keys.)
 
         PartialStateConflictError: if the room was un-partial stated in between
-            computing the state at the event and persisting it. The caller should retry
-            exactly once in this case. Will never be raised if `state_ids` is provided.
+            computing the state at the event and persisting it. The caller should
+            recompute `context` and retry exactly once when this happens.
         """
         logger.debug("Processing event: %s", event)
         assert not event.internal_metadata.outlier
 
-        context = await self._state_handler.compute_event_context(
-            event,
-            state_ids_before_event=state_ids,
-        )
         try:
             await self._check_event_auth(origin, event, context)
         except AuthError as e:
@@ -1144,7 +1182,7 @@ class FederationEventHandler:
             # For new (non-backfilled and non-outlier) events we check if the event
             # passes auth based on the current state. If it doesn't then we
             # "soft-fail" the event.
-            await self._check_for_soft_fail(event, state_ids, origin=origin)
+            await self._check_for_soft_fail(event, context=context, origin=origin)
 
         await self._run_push_actions_and_persist_event(event, context, backfilled)
 
@@ -1556,13 +1594,15 @@ class FederationEventHandler:
         )
 
     async def _check_event_auth(
-        self, origin: str, event: EventBase, context: EventContext
+        self, origin: Optional[str], event: EventBase, context: EventContext
     ) -> None:
         """
         Checks whether an event should be rejected (for failing auth checks).
 
         Args:
-            origin: The host the event originates from.
+            origin: The host the event originates from. This is used to fetch
+               any missing auth events. It can be set to None, but only if we are
+               sure that we already have all the auth events.
             event: The event itself.
             context:
                 The event context.
@@ -1705,7 +1745,7 @@ class FederationEventHandler:
     async def _check_for_soft_fail(
         self,
         event: EventBase,
-        state_ids: Optional[StateMap[str]],
+        context: EventContext,
         origin: str,
     ) -> None:
         """Checks if we should soft fail the event; if so, marks the event as
@@ -1716,7 +1756,7 @@ class FederationEventHandler:
 
         Args:
             event
-            state_ids: The state at the event if we don't have all the event's prev events
+            context: The `EventContext` which we are about to persist the event with.
             origin: The host the event originates from.
         """
         if await self._store.is_partial_state_room(event.room_id):
@@ -1742,11 +1782,15 @@ class FederationEventHandler:
         auth_types = auth_types_for_event(room_version_obj, event)
 
         # Calculate the "current state".
-        if state_ids is not None:
-            # If we're explicitly given the state then we won't have all the
-            # prev events, and so we have a gap in the graph. In this case
-            # we want to be a little careful as we might have been down for
-            # a while and have an incorrect view of the current state,
+        seen_event_ids = await self._store.have_events_in_timeline(prev_event_ids)
+        has_missing_prevs = bool(prev_event_ids - seen_event_ids)
+        if has_missing_prevs:
+            # We don't have all the prev_events of this event, which means we have a
+            # gap in the graph, and the new event is going to become a new backwards
+            # extremity.
+            #
+            # In this case we want to be a little careful as we might have been
+            # down for a while and have an incorrect view of the current state,
             # however we still want to do checks as gaps are easy to
             # maliciously manufacture.
             #
@@ -1759,6 +1803,7 @@ class FederationEventHandler:
                 event.room_id, extrem_ids
             )
             state_sets: List[StateMap[str]] = list(state_sets_d.values())
+            state_ids = await context.get_prev_state_ids()
             state_sets.append(state_ids)
             current_state_ids = (
                 await self._state_resolution_handler.resolve_events_with_store(
@@ -1808,7 +1853,7 @@ class FederationEventHandler:
             event.internal_metadata.soft_failed = True
 
     async def _load_or_fetch_auth_events_for_event(
-        self, destination: str, event: EventBase
+        self, destination: Optional[str], event: EventBase
     ) -> Collection[EventBase]:
         """Fetch this event's auth_events, from database or remote
 
@@ -1824,12 +1869,19 @@ class FederationEventHandler:
         Args:
             destination: where to send the /event_auth request. Typically the server
                that sent us `event` in the first place.
+
+               If this is None, no attempt is made to load any missing auth events:
+               rather, an AssertionError is raised if there are any missing events.
+
             event: the event whose auth_events we want
 
         Returns:
             all of the events listed in `event.auth_events_ids`, after deduplication
 
         Raises:
+            AssertionError if some auth events were missing and no `destination` was
+            supplied.
+
             AuthError if we were unable to fetch the auth_events for any reason.
         """
         event_auth_event_ids = set(event.auth_event_ids())
@@ -1841,6 +1893,13 @@ class FederationEventHandler:
         )
         if not missing_auth_event_ids:
             return event_auth_events.values()
+        if destination is None:
+            # this shouldn't happen: destination must be set unless we know we have already
+            # persisted the auth events.
+            raise AssertionError(
+                "_load_or_fetch_auth_events_for_event() called with no destination for "
+                "an event with missing auth_events"
+            )
 
         logger.info(
             "Event %s refers to unknown auth events %s: fetching auth chain",

synapse/handlers/message.py

@@ -41,6 +41,7 @@ from synapse.api.errors import (
     NotFoundError,
     ShadowBanError,
     SynapseError,
+    UnstableSpecAuthError,
     UnsupportedRoomVersionError,
 )
 from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
@@ -51,6 +52,7 @@ from synapse.events.builder import EventBuilder
 from synapse.events.snapshot import EventContext
 from synapse.events.validator import EventValidator
 from synapse.handlers.directory import DirectoryHandler
+from synapse.logging import opentracing
 from synapse.logging.context import make_deferred_yieldable, run_in_background
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.send_event import ReplicationSendEventRestServlet
@@ -149,7 +151,11 @@ class MessageHandler:
                 "Attempted to retrieve data from a room for a user that has never been in it. "
                 "This should not have happened."
             )
-            raise SynapseError(403, "User not in room", errcode=Codes.FORBIDDEN)
+            raise UnstableSpecAuthError(
+                403,
+                "User not in room",
+                errcode=Codes.NOT_JOINED,
+            )
 
         return data
 
@@ -319,8 +325,10 @@ class MessageHandler:
                 room_id, user_id, allow_departed_users=True
             )
             if membership != Membership.JOIN:
-                raise NotImplementedError(
-                    "Getting joined members after leaving is not implemented"
+                raise SynapseError(
+                    code=403,
+                    errcode=Codes.FORBIDDEN,
+                    msg="Getting joined members while not being a current member of the room is forbidden.",
                 )
 
         users_with_profile = await self.store.get_users_in_room_with_profiles(room_id)
@@ -334,7 +342,11 @@ class MessageHandler:
                     break
             else:
                 # Loop fell through, AS has no interested users in room
-                raise AuthError(403, "Appservice not in room")
+                raise UnstableSpecAuthError(
+                    403,
+                    "Appservice not in room",
+                    errcode=Codes.NOT_JOINED,
+                )
 
         return {
             user_id: {
@@ -1135,6 +1147,10 @@ class EventCreationHandler:
             context = await self.state.compute_event_context(
                 event,
                 state_ids_before_event=state_map_for_event,
+                # TODO(faster_joins): check how MSC2716 works and whether we can have
+                #   partial state here
+                #   https://github.com/matrix-org/synapse/issues/13003
+                partial_state=False,
             )
         else:
             context = await self.state.compute_event_context(event)
@@ -1359,9 +1375,10 @@ class EventCreationHandler:
         # and `state_groups` because they have `prev_events` that aren't persisted yet
         # (historical messages persisted in reverse-chronological order).
         if not event.internal_metadata.is_historical():
-            await self._bulk_push_rule_evaluator.action_for_event_by_user(
-                event, context
-            )
+            with opentracing.start_active_span("calculate_push_actions"):
+                await self._bulk_push_rule_evaluator.action_for_event_by_user(
+                    event, context
+                )
 
         try:
             # If we're a worker we need to hit out to the master.
@@ -1448,9 +1465,10 @@ class EventCreationHandler:
             state = await state_entry.get_state(
                 self._storage_controllers.state, StateFilter.all()
             )
-            joined_hosts = await self.store.get_joined_hosts(
-                event.room_id, state, state_entry
-            )
+            with opentracing.start_active_span("get_joined_hosts"):
+                joined_hosts = await self.store.get_joined_hosts(
+                    event.room_id, state, state_entry
+                )
 
             # Note that the expiry times must be larger than the expiry time in
             # _external_cache_joined_hosts_updates.

synapse/handlers/pagination.py

@@ -24,6 +24,7 @@ from synapse.api.errors import SynapseError
 from synapse.api.filtering import Filter
 from synapse.events.utils import SerializeEventConfig
 from synapse.handlers.room import ShutdownRoomResponse
+from synapse.logging.opentracing import trace
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage.state import StateFilter
 from synapse.streams.config import PaginationConfig
@@ -416,6 +417,7 @@ class PaginationHandler:
 
             await self._storage_controllers.purge_events.purge_room(room_id)
 
+    @trace
     async def get_messages(
         self,
         requester: Requester,

synapse/handlers/relations.py

@@ -19,6 +19,7 @@ import attr
 from synapse.api.constants import RelationTypes
 from synapse.api.errors import SynapseError
 from synapse.events import EventBase, relation_from_event
+from synapse.logging.opentracing import trace
 from synapse.storage.databases.main.relations import _RelatedEvent
 from synapse.types import JsonDict, Requester, StreamToken, UserID
 from synapse.visibility import filter_events_for_client
@@ -73,7 +74,6 @@ class RelationsHandler:
         room_id: str,
         relation_type: Optional[str] = None,
         event_type: Optional[str] = None,
-        aggregation_key: Optional[str] = None,
         limit: int = 5,
         direction: str = "b",
         from_token: Optional[StreamToken] = None,
@@ -89,7 +89,6 @@ class RelationsHandler:
             room_id: The room the event belongs to.
             relation_type: Only fetch events with this relation type, if given.
             event_type: Only fetch events with this event type, if given.
-            aggregation_key: Only fetch events with this aggregation key, if given.
             limit: Only fetch the most recent `limit` events.
             direction: Whether to fetch the most recent first (`"b"`) or the
                 oldest first (`"f"`).
@@ -122,7 +121,6 @@ class RelationsHandler:
             room_id=room_id,
             relation_type=relation_type,
             event_type=event_type,
-            aggregation_key=aggregation_key,
             limit=limit,
             direction=direction,
             from_token=from_token,
@@ -364,6 +362,7 @@ class RelationsHandler:
 
         return results
 
+    @trace
     async def get_bundled_aggregations(
         self, events: Iterable[EventBase], user_id: str
     ) -> Dict[str, BundledAggregations]:

synapse/handlers/room_list.py

@@ -182,7 +182,7 @@ class RoomListHandler:
                 == HistoryVisibility.WORLD_READABLE,
                 "guest_can_join": room["guest_access"] == "can_join",
                 "join_rule": room["join_rules"],
-                "org.matrix.msc3827.room_type": room["room_type"],
+                "room_type": room["room_type"],
             }
 
             # Filter out Nones – rather omit the field altogether

synapse/handlers/room_member.py

@@ -32,6 +32,7 @@ from synapse.event_auth import get_named_level, get_power_level_event
 from synapse.events import EventBase
 from synapse.events.snapshot import EventContext
 from synapse.handlers.profile import MAX_AVATAR_URL_LEN, MAX_DISPLAYNAME_LEN
+from synapse.logging import opentracing
 from synapse.module_api import NOT_SPAM
 from synapse.storage.state import StateFilter
 from synapse.types import (
@@ -428,14 +429,14 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
                 await self._join_rate_per_room_limiter.ratelimit(
                     requester, key=room_id, update=False
                 )
-
-        result_event = await self.event_creation_handler.handle_new_client_event(
-            requester,
-            event,
-            context,
-            extra_users=[target],
-            ratelimit=ratelimit,
-        )
+        with opentracing.start_active_span("handle_new_client_event"):
+            result_event = await self.event_creation_handler.handle_new_client_event(
+                requester,
+                event,
+                context,
+                extra_users=[target],
+                ratelimit=ratelimit,
+            )
 
         if event.membership == Membership.LEAVE:
             if prev_member_event_id:
@@ -564,25 +565,26 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
         # by application services), and then by room ID.
         async with self.member_as_limiter.queue(as_id):
             async with self.member_linearizer.queue(key):
-                result = await self.update_membership_locked(
-                    requester,
-                    target,
-                    room_id,
-                    action,
-                    txn_id=txn_id,
-                    remote_room_hosts=remote_room_hosts,
-                    third_party_signed=third_party_signed,
-                    ratelimit=ratelimit,
-                    content=content,
-                    new_room=new_room,
-                    require_consent=require_consent,
-                    outlier=outlier,
-                    historical=historical,
-                    allow_no_prev_events=allow_no_prev_events,
-                    prev_event_ids=prev_event_ids,
-                    state_event_ids=state_event_ids,
-                    depth=depth,
-                )
+                with opentracing.start_active_span("update_membership_locked"):
+                    result = await self.update_membership_locked(
+                        requester,
+                        target,
+                        room_id,
+                        action,
+                        txn_id=txn_id,
+                        remote_room_hosts=remote_room_hosts,
+                        third_party_signed=third_party_signed,
+                        ratelimit=ratelimit,
+                        content=content,
+                        new_room=new_room,
+                        require_consent=require_consent,
+                        outlier=outlier,
+                        historical=historical,
+                        allow_no_prev_events=allow_no_prev_events,
+                        prev_event_ids=prev_event_ids,
+                        state_event_ids=state_event_ids,
+                        depth=depth,
+                    )
 
         return result
 
@@ -649,6 +651,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
         Returns:
             A tuple of the new event ID and stream ID.
         """
+
         content_specified = bool(content)
         if content is None:
             content = {}
@@ -1679,7 +1682,11 @@ class RoomMemberMasterHandler(RoomMemberHandler):
         ]
 
         if len(remote_room_hosts) == 0:
-            raise SynapseError(404, "No known servers")
+            raise SynapseError(
+                404,
+                "Can't join remote room because no servers "
+                "that are in the room have been provided.",
+            )
 
         check_complexity = self.hs.config.server.limit_remote_rooms.enabled
         if (

synapse/handlers/room_summary.py

@@ -28,11 +28,11 @@ from synapse.api.constants import (
     RoomTypes,
 )
 from synapse.api.errors import (
-    AuthError,
     Codes,
     NotFoundError,
     StoreError,
     SynapseError,
+    UnstableSpecAuthError,
     UnsupportedRoomVersionError,
 )
 from synapse.api.ratelimiting import Ratelimiter
@@ -175,10 +175,11 @@ class RoomSummaryHandler:
 
         # First of all, check that the room is accessible.
         if not await self._is_local_room_accessible(requested_room_id, requester):
-            raise AuthError(
+            raise UnstableSpecAuthError(
                 403,
                 "User %s not in room %s, and room previews are disabled"
                 % (requester, requested_room_id),
+                errcode=Codes.NOT_JOINED,
             )
 
         # If this is continuing a previous session, pull the persisted data.
@@ -452,7 +453,6 @@ class RoomSummaryHandler:
                 "type": e.type,
                 "state_key": e.state_key,
                 "content": e.content,
-                "room_id": e.room_id,
                 "sender": e.sender,
                 "origin_server_ts": e.origin_server_ts,
             }

synapse/handlers/sync.py

@@ -1536,15 +1536,13 @@ class SyncHandler:
         ignored_users = await self.store.ignored_users(user_id)
         if since_token:
             room_changes = await self._get_rooms_changed(
-                sync_result_builder, ignored_users, self.rooms_to_exclude
+                sync_result_builder, ignored_users
             )
             tags_by_room = await self.store.get_updated_tags(
                 user_id, since_token.account_data_key
             )
         else:
-            room_changes = await self._get_all_rooms(
-                sync_result_builder, ignored_users, self.rooms_to_exclude
-            )
+            room_changes = await self._get_all_rooms(sync_result_builder, ignored_users)
             tags_by_room = await self.store.get_tags_for_user(user_id)
 
         log_kv({"rooms_changed": len(room_changes.room_entries)})
@@ -1623,13 +1621,14 @@ class SyncHandler:
         self,
         sync_result_builder: "SyncResultBuilder",
         ignored_users: FrozenSet[str],
-        excluded_rooms: List[str],
     ) -> _RoomChanges:
         """Determine the changes in rooms to report to the user.
 
         This function is a first pass at generating the rooms part of the sync response.
         It determines which rooms have changed during the sync period, and categorises
-        them into four buckets: "knock", "invite", "join" and "leave".
+        them into four buckets: "knock", "invite", "join" and "leave". It also excludes
+        from that list any room that appears in the list of rooms to exclude from sync
+        results in the server configuration.
 
         1. Finds all membership changes for the user in the sync period (from
            `since_token` up to `now_token`).
@@ -1655,7 +1654,7 @@ class SyncHandler:
         #       _have_rooms_changed. We could keep the results in memory to avoid a
         #       second query, at the cost of more complicated source code.
         membership_change_events = await self.store.get_membership_changes_for_user(
-            user_id, since_token.room_key, now_token.room_key, excluded_rooms
+            user_id, since_token.room_key, now_token.room_key, self.rooms_to_exclude
         )
 
         mem_change_events_by_room_id: Dict[str, List[EventBase]] = {}
@@ -1862,7 +1861,6 @@ class SyncHandler:
         self,
         sync_result_builder: "SyncResultBuilder",
         ignored_users: FrozenSet[str],
-        ignored_rooms: List[str],
     ) -> _RoomChanges:
         """Returns entries for all rooms for the user.
 
@@ -1884,7 +1882,7 @@ class SyncHandler:
         room_list = await self.store.get_rooms_for_local_user_where_membership_is(
             user_id=user_id,
             membership_list=Membership.LIST,
-            excluded_rooms=ignored_rooms,
+            excluded_rooms=self.rooms_to_exclude,
         )
 
         room_entries = []
@@ -2150,7 +2148,9 @@ class SyncHandler:
                 raise Exception("Unrecognized rtype: %r", room_builder.rtype)
 
     async def get_rooms_for_user_at(
-        self, user_id: str, room_key: RoomStreamToken
+        self,
+        user_id: str,
+        room_key: RoomStreamToken,
     ) -> FrozenSet[str]:
         """Get set of joined rooms for a user at the given stream ordering.
 
@@ -2176,7 +2176,12 @@ class SyncHandler:
         # If the membership's stream ordering is after the given stream
         # ordering, we need to go and work out if the user was in the room
         # before.
+        # We also need to check whether the room should be excluded from sync
+        # responses as per the homeserver config.
         for joined_room in joined_rooms:
+            if joined_room.room_id in self.rooms_to_exclude:
+                continue
+
             if not joined_room.event_pos.persisted_after(room_key):
                 joined_room_ids.add(joined_room.room_id)
                 continue

synapse/handlers/typing.py

@@ -489,8 +489,15 @@ class TypingNotificationEventSource(EventSource[int, JsonDict]):
             handler = self.get_typing_handler()
 
             events = []
-            for room_id in handler._room_serials.keys():
-                if handler._room_serials[room_id] <= from_key:
+
+            # Work on a copy of things here as these may change in the handler while
+            # waiting for the AS `is_interested_in_room` call to complete.
+            # Shallow copy is safe as no nested data is present.
+            latest_room_serial = handler._latest_room_serial
+            room_serials = handler._room_serials.copy()
+
+            for room_id, serial in room_serials.items():
+                if serial <= from_key:
                     continue
 
                 if not await service.is_interested_in_room(room_id, self._main_store):
@@ -498,7 +505,7 @@ class TypingNotificationEventSource(EventSource[int, JsonDict]):
 
                 events.append(self._make_event_for(room_id))
 
-            return events, handler._latest_room_serial
+            return events, latest_room_serial
 
     async def get_new_events(
         self,

synapse/http/request_metrics.py

@@ -43,6 +43,24 @@ response_timer = Histogram(
     "synapse_http_server_response_time_seconds",
     "sec",
     ["method", "servlet", "tag", "code"],
+    buckets=(
+        0.005,
+        0.01,
+        0.025,
+        0.05,
+        0.1,
+        0.25,
+        0.5,
+        1.0,
+        2.5,
+        5.0,
+        10.0,
+        30.0,
+        60.0,
+        120.0,
+        180.0,
+        "+Inf",
+    ),
 )
 
 response_ru_utime = Counter(

synapse/http/server.py

@@ -58,6 +58,7 @@ from synapse.api.errors import (
     SynapseError,
     UnrecognizedRequestError,
 )
+from synapse.config.homeserver import HomeServerConfig
 from synapse.http.site import SynapseRequest
 from synapse.logging.context import defer_to_thread, preserve_fn, run_in_background
 from synapse.logging.opentracing import active_span, start_active_span, trace_servlet
@@ -155,15 +156,16 @@ def is_method_cancellable(method: Callable[..., Any]) -> bool:
     return getattr(method, "cancellable", False)
 
 
-def return_json_error(f: failure.Failure, request: SynapseRequest) -> None:
+def return_json_error(
+    f: failure.Failure, request: SynapseRequest, config: Optional[HomeServerConfig]
+) -> None:
     """Sends a JSON error response to clients."""
 
     if f.check(SynapseError):
         # mypy doesn't understand that f.check asserts the type.
         exc: SynapseError = f.value  # type: ignore
         error_code = exc.code
-        error_dict = exc.error_dict()
-
+        error_dict = exc.error_dict(config)
         logger.info("%s SynapseError: %s - %s", request, error_code, exc.msg)
     elif f.check(CancelledError):
         error_code = HTTP_STATUS_REQUEST_CANCELLED
@@ -450,7 +452,7 @@ class DirectServeJsonResource(_AsyncResource):
         request: SynapseRequest,
     ) -> None:
         """Implements _AsyncResource._send_error_response"""
-        return_json_error(f, request)
+        return_json_error(f, request, None)
 
 
 @attr.s(slots=True, frozen=True, auto_attribs=True)
@@ -575,6 +577,14 @@ class JsonResource(DirectServeJsonResource):
 
         return callback_return
 
+    def _send_error_response(
+        self,
+        f: failure.Failure,
+        request: SynapseRequest,
+    ) -> None:
+        """Implements _AsyncResource._send_error_response"""
+        return_json_error(f, request, self.hs.config)
+
 
 class DirectServeHtmlResource(_AsyncResource):
     """A resource that will call `self._async_on_<METHOD>` on new requests,

synapse/logging/opentracing.py

@@ -901,6 +901,11 @@ def trace(func: Callable[P, R]) -> Callable[P, R]:
 def tag_args(func: Callable[P, R]) -> Callable[P, R]:
     """
     Tags all of the args to the active span.
+
+    Args:
+        func: `func` is assumed to be a method taking a `self` parameter, or a
+            `classmethod` taking a `cls` parameter. In either case, a tag is not
+            created for this parameter.
     """
 
     if not opentracing:
@@ -909,8 +914,14 @@ def tag_args(func: Callable[P, R]) -> Callable[P, R]:
     @wraps(func)
     def _tag_args_inner(*args: P.args, **kwargs: P.kwargs) -> R:
         argspec = inspect.getfullargspec(func)
-        for i, arg in enumerate(argspec.args[1:]):
-            set_tag("ARG_" + arg, str(args[i]))  # type: ignore[index]
+        # We use `[1:]` to skip the `self` object reference and `start=1` to
+        # make the index line up with `argspec.args`.
+        #
+        # FIXME: We could update this handle any type of function by ignoring the
+        #   first argument only if it's named `self` or `cls`. This isn't fool-proof
+        #   but handles the idiomatic cases.
+        for i, arg in enumerate(args[1:], start=1):  # type: ignore[index]
+            set_tag("ARG_" + argspec.args[i], str(arg))
         set_tag("args", str(args[len(argspec.args) :]))  # type: ignore[index]
         set_tag("kwargs", str(kwargs))
         return func(*args, **kwargs)

synapse/module_api/__init__.py

@@ -1452,6 +1452,81 @@ class ModuleApi:
             start_timestamp, end_timestamp
         )
 
+    async def lookup_room_alias(self, room_alias: str) -> Tuple[str, List[str]]:
+        """
+        Get the room ID associated with a room alias.
+
+        Added in Synapse v1.65.0.
+
+        Args:
+            room_alias: The alias to look up.
+
+        Returns:
+            A tuple of:
+                The room ID (str).
+                Hosts likely to be participating in the room ([str]).
+
+        Raises:
+            SynapseError if room alias is invalid or could not be found.
+        """
+        alias = RoomAlias.from_string(room_alias)
+        (room_id, hosts) = await self._hs.get_room_member_handler().lookup_room_alias(
+            alias
+        )
+
+        return room_id.to_string(), hosts
+
+    async def create_room(
+        self,
+        user_id: str,
+        config: JsonDict,
+        ratelimit: bool = True,
+        creator_join_profile: Optional[JsonDict] = None,
+    ) -> Tuple[str, Optional[str]]:
+        """Creates a new room.
+
+        Added in Synapse v1.65.0.
+
+        Args:
+            user_id:
+                The user who requested the room creation.
+            config : A dict of configuration options. See "Request body" of:
+                https://spec.matrix.org/latest/client-server-api/#post_matrixclientv3createroom
+            ratelimit: set to False to disable the rate limiter for this specific operation.
+
+            creator_join_profile:
+                Set to override the displayname and avatar for the creating
+                user in this room. If unset, displayname and avatar will be
+                derived from the user's profile. If set, should contain the
+                values to go in the body of the 'join' event (typically
+                `avatar_url` and/or `displayname`.
+
+        Returns:
+                A tuple containing: 1) the room ID (str), 2) if an alias was requested,
+                the room alias (str), otherwise None if no alias was requested.
+
+        Raises:
+            ResourceLimitError if server is blocked to some resource being
+            exceeded.
+            RuntimeError if the user_id does not refer to a local user.
+            SynapseError if the user_id is invalid, room ID couldn't be stored, or
+            something went horribly wrong.
+        """
+        if not self.is_mine(user_id):
+            raise RuntimeError(
+                "Tried to create a room as a user that isn't local to this homeserver",
+            )
+
+        requester = create_requester(user_id)
+        room_id_and_alias, _ = await self._hs.get_room_creation_handler().create_room(
+            requester=requester,
+            config=config,
+            ratelimit=ratelimit,
+            creator_join_profile=creator_join_profile,
+        )
+
+        return room_id_and_alias["room_id"], room_id_and_alias.get("room_alias", None)
+
 
 class PublicRoomListManager:
     """Contains methods for adding to, removing from and querying whether a room

synapse/rest/client/register.py

@@ -33,7 +33,7 @@ from synapse.api.ratelimiting import Ratelimiter
 from synapse.config import ConfigError
 from synapse.config.emailconfig import ThreepidBehaviour
 from synapse.config.homeserver import HomeServerConfig
-from synapse.config.ratelimiting import FederationRateLimitConfig
+from synapse.config.ratelimiting import FederationRatelimitSettings
 from synapse.config.server import is_threepid_reserved
 from synapse.handlers.auth import AuthHandler
 from synapse.handlers.ui_auth import UIAuthSessionDataConstants
@@ -325,7 +325,7 @@ class UsernameAvailabilityRestServlet(RestServlet):
         self.registration_handler = hs.get_registration_handler()
         self.ratelimiter = FederationRateLimiter(
             hs.get_clock(),
-            FederationRateLimitConfig(
+            FederationRatelimitSettings(
                 # Time window of 2s
                 window_size=2000,
                 # Artificially delay requests if rate > sleep_limit/window_size

synapse/rest/client/versions.py

@@ -95,8 +95,8 @@ class VersionsRestServlet(RestServlet):
                     "org.matrix.msc3026.busy_presence": self.config.experimental.msc3026_enabled,
                     # Supports receiving private read receipts as per MSC2285
                     "org.matrix.msc2285": self.config.experimental.msc2285_enabled,
-                    # Supports filtering of /publicRooms by room type MSC3827
-                    "org.matrix.msc3827": self.config.experimental.msc3827_enabled,
+                    # Supports filtering of /publicRooms by room type as per MSC3827
+                    "org.matrix.msc3827.stable": True,
                     # Adds support for importing historical messages as per MSC2716
                     "org.matrix.msc2716": self.config.experimental.msc2716_enabled,
                     # Adds support for jump to date endpoints (/timestamp_to_event) as per MSC3030

synapse/state/__init__.py

@@ -255,7 +255,7 @@ class StateHandler:
         self,
         event: EventBase,
         state_ids_before_event: Optional[StateMap[str]] = None,
-        partial_state: bool = False,
+        partial_state: Optional[bool] = None,
     ) -> EventContext:
         """Build an EventContext structure for a non-outlier event.
 
@@ -270,10 +270,18 @@ class StateHandler:
                 it can't be calculated from existing events. This is normally
                 only specified when receiving an event from federation where we
                 don't have the prev events, e.g. when backfilling.
-            partial_state: True if `state_ids_before_event` is partial and omits
-                non-critical membership events
+            partial_state:
+                `True` if `state_ids_before_event` is partial and omits non-critical
+                membership events.
+                `False` if `state_ids_before_event` is the full state.
+                `None` when `state_ids_before_event` is not provided. In this case, the
+                flag will be calculated based on `event`'s prev events.
         Returns:
             The event context.
+
+        Raises:
+            RuntimeError if `state_ids_before_event` is not provided and one or more
+                prev events are missing or outliers.
         """
 
         assert not event.internal_metadata.is_outlier()
@@ -298,12 +306,14 @@ class StateHandler:
                 )
             )
 
+            # the partial_state flag must be provided
+            assert partial_state is not None
         else:
             # otherwise, we'll need to resolve the state across the prev_events.
 
             # partial_state should not be set explicitly in this case:
             # we work it out dynamically
-            assert not partial_state
+            assert partial_state is None
 
             # if any of the prev-events have partial state, so do we.
             # (This is slightly racy - the prev-events might get fixed up before we use
@@ -313,13 +323,13 @@ class StateHandler:
             incomplete_prev_events = await self.store.get_partial_state_events(
                 prev_event_ids
             )
-            if any(incomplete_prev_events.values()):
+            partial_state = any(incomplete_prev_events.values())
+            if partial_state:
                 logger.debug(
                     "New/incoming event %s refers to prev_events %s with partial state",
                     event.event_id,
                     [k for (k, v) in incomplete_prev_events.items() if v],
                 )
-                partial_state = True
 
             logger.debug("calling resolve_state_groups from compute_event_context")
             # we've already taken into account partial state, so no need to wait for
@@ -426,6 +436,10 @@ class StateHandler:
 
         Returns:
             The resolved state
+
+        Raises:
+            RuntimeError if we don't have a state group for one or more of the events
+               (ie. they are outliers or unknown)
         """
         logger.debug("resolve_state_groups event_ids %s", event_ids)
 

synapse/storage/controllers/state.py

@@ -29,6 +29,7 @@ from typing import (
 
 from synapse.api.constants import EventTypes
 from synapse.events import EventBase
+from synapse.logging.opentracing import trace
 from synapse.storage.state import StateFilter
 from synapse.storage.util.partial_state_events_tracker import (
     PartialCurrentStateTracker,
@@ -82,13 +83,15 @@ class StateStorageController:
         return state_group_delta.prev_group, state_group_delta.delta_ids
 
     async def get_state_groups_ids(
-        self, _room_id: str, event_ids: Collection[str]
+        self, _room_id: str, event_ids: Collection[str], await_full_state: bool = True
     ) -> Dict[int, MutableStateMap[str]]:
         """Get the event IDs of all the state for the state groups for the given events
 
         Args:
             _room_id: id of the room for these events
             event_ids: ids of the events
+            await_full_state: if `True`, will block if we do not yet have complete
+               state at these events.
 
         Returns:
             dict of state_group_id -> (dict of (type, state_key) -> event id)
@@ -100,7 +103,9 @@ class StateStorageController:
         if not event_ids:
             return {}
 
-        event_to_groups = await self.get_state_group_for_events(event_ids)
+        event_to_groups = await self.get_state_group_for_events(
+            event_ids, await_full_state=await_full_state
+        )
 
         groups = set(event_to_groups.values())
         group_to_state = await self.stores.state._get_state_for_groups(groups)
@@ -175,6 +180,7 @@ class StateStorageController:
 
         return self.stores.state._get_state_groups_from_groups(groups, state_filter)
 
+    @trace
     async def get_state_for_events(
         self, event_ids: Collection[str], state_filter: Optional[StateFilter] = None
     ) -> Dict[str, StateMap[EventBase]]:
@@ -221,6 +227,7 @@ class StateStorageController:
 
         return {event: event_to_state[event] for event in event_ids}
 
+    @trace
     async def get_state_ids_for_events(
         self,
         event_ids: Collection[str],
@@ -283,6 +290,7 @@ class StateStorageController:
         )
         return state_map[event_id]
 
+    @trace
     async def get_state_ids_for_event(
         self, event_id: str, state_filter: Optional[StateFilter] = None
     ) -> StateMap[str]:
@@ -323,6 +331,7 @@ class StateStorageController:
             groups, state_filter or StateFilter.all()
         )
 
+    @trace
     async def get_state_group_for_events(
         self,
         event_ids: Collection[str],
@@ -334,6 +343,10 @@ class StateStorageController:
             event_ids: events to get state groups for
             await_full_state: if true, will block if we do not yet have complete
                state at these events.
+
+        Raises:
+            RuntimeError if we don't have a state group for one or more of the events
+               (ie. they are outliers or unknown)
         """
         if await_full_state:
             await self._partial_state_events_tracker.await_full_state(event_ids)

synapse/storage/databases/main/event_push_actions.py

@@ -265,7 +265,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
             counts.notify_count += row[1]
             counts.unread_count += row[2]
 
-        # Next we need to count highlights, which aren't summarized
+        # Next we need to count highlights, which aren't summarised
         sql = """
             SELECT COUNT(*) FROM event_push_actions
             WHERE user_id = ?
@@ -280,7 +280,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
 
         # Finally we need to count push actions that aren't included in the
         # summary returned above, e.g. recent events that haven't been
-        # summarized yet, or the summary is empty due to a recent read receipt.
+        # summarised yet, or the summary is empty due to a recent read receipt.
         stream_ordering = max(stream_ordering, summary_stream_ordering)
         notify_count, unread_count = self._get_notif_unread_count_for_user_room(
             txn, room_id, user_id, stream_ordering
@@ -304,6 +304,17 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
 
         Does not consult `event_push_summary` table, which may include push
         actions that have been deleted from `event_push_actions` table.
+
+        Args:
+            txn: The database transaction.
+            room_id: The room ID to get unread counts for.
+            user_id: The user ID to get unread counts for.
+            stream_ordering: The (exclusive) minimum stream ordering to consider.
+            max_stream_ordering: The (inclusive) maximum stream ordering to consider.
+                If this is not given, then no maximum is applied.
+
+        Return:
+            A tuple of the notif count and unread count in the given range.
         """
 
         # If there have been no events in the room since the stream ordering,
@@ -383,27 +394,27 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         ) -> List[Tuple[str, str, int, str, bool]]:
             # find rooms that have a read receipt in them and return the next
             # push actions
-            sql = (
-                "SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,"
-                "   ep.highlight "
-                " FROM ("
-                "   SELECT room_id,"
-                "       MAX(stream_ordering) as stream_ordering"
-                "   FROM events"
-                "   INNER JOIN receipts_linearized USING (room_id, event_id)"
-                "   WHERE receipt_type = 'm.read' AND user_id = ?"
-                "   GROUP BY room_id"
-                ") AS rl,"
-                " event_push_actions AS ep"
-                " WHERE"
-                "   ep.room_id = rl.room_id"
-                "   AND ep.stream_ordering > rl.stream_ordering"
-                "   AND ep.user_id = ?"
-                "   AND ep.stream_ordering > ?"
-                "   AND ep.stream_ordering <= ?"
-                "   AND ep.notif = 1"
-                " ORDER BY ep.stream_ordering ASC LIMIT ?"
-            )
+            sql = """
+                SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,
+                    ep.highlight
+                FROM (
+                    SELECT room_id,
+                        MAX(stream_ordering) as stream_ordering
+                    FROM events
+                    INNER JOIN receipts_linearized USING (room_id, event_id)
+                    WHERE receipt_type = 'm.read' AND user_id = ?
+                    GROUP BY room_id
+                ) AS rl,
+                    event_push_actions AS ep
+                WHERE
+                    ep.room_id = rl.room_id
+                    AND ep.stream_ordering > rl.stream_ordering
+                    AND ep.user_id = ?
+                    AND ep.stream_ordering > ?
+                    AND ep.stream_ordering <= ?
+                    AND ep.notif = 1
+                ORDER BY ep.stream_ordering ASC LIMIT ?
+            """
             args = [user_id, user_id, min_stream_ordering, max_stream_ordering, limit]
             txn.execute(sql, args)
             return cast(List[Tuple[str, str, int, str, bool]], txn.fetchall())
@@ -418,23 +429,23 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         def get_no_receipt(
             txn: LoggingTransaction,
         ) -> List[Tuple[str, str, int, str, bool]]:
-            sql = (
-                "SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,"
-                "   ep.highlight "
-                " FROM event_push_actions AS ep"
-                " INNER JOIN events AS e USING (room_id, event_id)"
-                " WHERE"
-                "   ep.room_id NOT IN ("
-                "     SELECT room_id FROM receipts_linearized"
-                "       WHERE receipt_type = 'm.read' AND user_id = ?"
-                "       GROUP BY room_id"
-                "   )"
-                "   AND ep.user_id = ?"
-                "   AND ep.stream_ordering > ?"
-                "   AND ep.stream_ordering <= ?"
-                "   AND ep.notif = 1"
-                " ORDER BY ep.stream_ordering ASC LIMIT ?"
-            )
+            sql = """
+                SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,
+                    ep.highlight
+                FROM event_push_actions AS ep
+                INNER JOIN events AS e USING (room_id, event_id)
+                WHERE
+                    ep.room_id NOT IN (
+                        SELECT room_id FROM receipts_linearized
+                        WHERE receipt_type = 'm.read' AND user_id = ?
+                        GROUP BY room_id
+                    )
+                    AND ep.user_id = ?
+                    AND ep.stream_ordering > ?
+                    AND ep.stream_ordering <= ?
+                    AND ep.notif = 1
+                ORDER BY ep.stream_ordering ASC LIMIT ?
+            """
             args = [user_id, user_id, min_stream_ordering, max_stream_ordering, limit]
             txn.execute(sql, args)
             return cast(List[Tuple[str, str, int, str, bool]], txn.fetchall())
@@ -490,28 +501,28 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         def get_after_receipt(
             txn: LoggingTransaction,
         ) -> List[Tuple[str, str, int, str, bool, int]]:
-            sql = (
-                "SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,"
-                "  ep.highlight, e.received_ts"
-                " FROM ("
-                "   SELECT room_id,"
-                "       MAX(stream_ordering) as stream_ordering"
-                "   FROM events"
-                "   INNER JOIN receipts_linearized USING (room_id, event_id)"
-                "   WHERE receipt_type = 'm.read' AND user_id = ?"
-                "   GROUP BY room_id"
-                ") AS rl,"
-                " event_push_actions AS ep"
-                " INNER JOIN events AS e USING (room_id, event_id)"
-                " WHERE"
-                "   ep.room_id = rl.room_id"
-                "   AND ep.stream_ordering > rl.stream_ordering"
-                "   AND ep.user_id = ?"
-                "   AND ep.stream_ordering > ?"
-                "   AND ep.stream_ordering <= ?"
-                "   AND ep.notif = 1"
-                " ORDER BY ep.stream_ordering DESC LIMIT ?"
-            )
+            sql = """
+                SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,
+                    ep.highlight, e.received_ts
+                FROM (
+                    SELECT room_id,
+                        MAX(stream_ordering) as stream_ordering
+                    FROM events
+                    INNER JOIN receipts_linearized USING (room_id, event_id)
+                    WHERE receipt_type = 'm.read' AND user_id = ?
+                    GROUP BY room_id
+                ) AS rl,
+                event_push_actions AS ep
+                INNER JOIN events AS e USING (room_id, event_id)
+                WHERE
+                    ep.room_id = rl.room_id
+                    AND ep.stream_ordering > rl.stream_ordering
+                    AND ep.user_id = ?
+                    AND ep.stream_ordering > ?
+                    AND ep.stream_ordering <= ?
+                    AND ep.notif = 1
+                ORDER BY ep.stream_ordering DESC LIMIT ?
+            """
             args = [user_id, user_id, min_stream_ordering, max_stream_ordering, limit]
             txn.execute(sql, args)
             return cast(List[Tuple[str, str, int, str, bool, int]], txn.fetchall())
@@ -526,23 +537,23 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         def get_no_receipt(
             txn: LoggingTransaction,
         ) -> List[Tuple[str, str, int, str, bool, int]]:
-            sql = (
-                "SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,"
-                "   ep.highlight, e.received_ts"
-                " FROM event_push_actions AS ep"
-                " INNER JOIN events AS e USING (room_id, event_id)"
-                " WHERE"
-                "   ep.room_id NOT IN ("
-                "     SELECT room_id FROM receipts_linearized"
-                "       WHERE receipt_type = 'm.read' AND user_id = ?"
-                "       GROUP BY room_id"
-                "   )"
-                "   AND ep.user_id = ?"
-                "   AND ep.stream_ordering > ?"
-                "   AND ep.stream_ordering <= ?"
-                "   AND ep.notif = 1"
-                " ORDER BY ep.stream_ordering DESC LIMIT ?"
-            )
+            sql = """
+                SELECT ep.event_id, ep.room_id, ep.stream_ordering, ep.actions,
+                    ep.highlight, e.received_ts
+                FROM event_push_actions AS ep
+                INNER JOIN events AS e USING (room_id, event_id)
+                WHERE
+                    ep.room_id NOT IN (
+                        SELECT room_id FROM receipts_linearized
+                        WHERE receipt_type = 'm.read' AND user_id = ?
+                        GROUP BY room_id
+                    )
+                    AND ep.user_id = ?
+                    AND ep.stream_ordering > ?
+                    AND ep.stream_ordering <= ?
+                    AND ep.notif = 1
+                ORDER BY ep.stream_ordering DESC LIMIT ?
+            """
             args = [user_id, user_id, min_stream_ordering, max_stream_ordering, limit]
             txn.execute(sql, args)
             return cast(List[Tuple[str, str, int, str, bool, int]], txn.fetchall())
@@ -769,12 +780,12 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         # [10, <none>, 20], we should treat this as being equivalent to
         # [10, 10, 20].
         #
-        sql = (
-            "SELECT received_ts FROM events"
-            " WHERE stream_ordering <= ?"
-            " ORDER BY stream_ordering DESC"
-            " LIMIT 1"
-        )
+        sql = """
+            SELECT received_ts FROM events
+            WHERE stream_ordering <= ?
+            ORDER BY stream_ordering DESC
+            LIMIT 1
+        """
 
         while range_end - range_start > 0:
             middle = (range_end + range_start) // 2
@@ -802,14 +813,14 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         self, stream_ordering: int
     ) -> Optional[int]:
         def f(txn: LoggingTransaction) -> Optional[Tuple[int]]:
-            sql = (
-                "SELECT e.received_ts"
-                " FROM event_push_actions AS ep"
-                " JOIN events e ON ep.room_id = e.room_id AND ep.event_id = e.event_id"
-                " WHERE ep.stream_ordering > ? AND notif = 1"
-                " ORDER BY ep.stream_ordering ASC"
-                " LIMIT 1"
-            )
+            sql = """
+                SELECT e.received_ts
+                FROM event_push_actions AS ep
+                JOIN events e ON ep.room_id = e.room_id AND ep.event_id = e.event_id
+                WHERE ep.stream_ordering > ? AND notif = 1
+                ORDER BY ep.stream_ordering ASC
+                LIMIT 1
+            """
             txn.execute(sql, (stream_ordering,))
             return cast(Optional[Tuple[int]], txn.fetchone())
 
@@ -858,10 +869,13 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         Any push actions which predate the user's most recent read receipt are
         now redundant, so we can remove them from `event_push_actions` and
         update `event_push_summary`.
+
+        Returns true if all new receipts have been processed.
         """
 
         limit = 100
 
+        # The (inclusive) receipt stream ID that was previously processed..
         min_receipts_stream_id = self.db_pool.simple_select_one_onecol_txn(
             txn,
             table="event_push_summary_last_receipt_stream_id",
@@ -871,6 +885,14 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
 
         max_receipts_stream_id = self._receipts_id_gen.get_current_token()
 
+        # The (inclusive) event stream ordering that was previously summarised.
+        old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
+            txn,
+            table="event_push_summary_stream_ordering",
+            keyvalues={},
+            retcol="stream_ordering",
+        )
+
         sql = """
             SELECT r.stream_id, r.room_id, r.user_id, e.stream_ordering
             FROM receipts_linearized AS r
@@ -895,13 +917,6 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         )
         rows = txn.fetchall()
 
-        old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
-            txn,
-            table="event_push_summary_stream_ordering",
-            keyvalues={},
-            retcol="stream_ordering",
-        )
-
         # For each new read receipt we delete push actions from before it and
         # recalculate the summary.
         for _, room_id, user_id, stream_ordering in rows:
@@ -920,10 +935,13 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
                 (room_id, user_id, stream_ordering),
             )
 
+            # Fetch the notification counts between the stream ordering of the
+            # latest receipt and what was previously summarised.
             notif_count, unread_count = self._get_notif_unread_count_for_user_room(
                 txn, room_id, user_id, stream_ordering, old_rotate_stream_ordering
             )
 
+            # Replace the previous summary with the new counts.
             self.db_pool.simple_upsert_txn(
                 txn,
                 table="event_push_summary",
@@ -956,10 +974,12 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
         return len(rows) < limit
 
     def _rotate_notifs_txn(self, txn: LoggingTransaction) -> bool:
-        """Archives older notifications into event_push_summary. Returns whether
-        the archiving process has caught up or not.
+        """Archives older notifications (from event_push_actions) into event_push_summary.
+
+        Returns whether the archiving process has caught up or not.
         """
 
+        # The (inclusive) event stream ordering that was previously summarised.
         old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
             txn,
             table="event_push_summary_stream_ordering",
@@ -974,7 +994,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
             SELECT stream_ordering FROM event_push_actions
             WHERE stream_ordering > ?
             ORDER BY stream_ordering ASC LIMIT 1 OFFSET ?
-        """,
+            """,
             (old_rotate_stream_ordering, self._rotate_count),
         )
         stream_row = txn.fetchone()
@@ -993,19 +1013,31 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
 
         logger.info("Rotating notifications up to: %s", rotate_to_stream_ordering)
 
-        self._rotate_notifs_before_txn(txn, rotate_to_stream_ordering)
+        self._rotate_notifs_before_txn(
+            txn, old_rotate_stream_ordering, rotate_to_stream_ordering
+        )
 
         return caught_up
 
     def _rotate_notifs_before_txn(
-        self, txn: LoggingTransaction, rotate_to_stream_ordering: int
+        self,
+        txn: LoggingTransaction,
+        old_rotate_stream_ordering: int,
+        rotate_to_stream_ordering: int,
     ) -> None:
-        old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
-            txn,
-            table="event_push_summary_stream_ordering",
-            keyvalues={},
-            retcol="stream_ordering",
-        )
+        """Archives older notifications (from event_push_actions) into event_push_summary.
+
+        Any event_push_actions between old_rotate_stream_ordering (exclusive) and
+        rotate_to_stream_ordering (inclusive) will be added to the event_push_summary
+        table.
+
+        Args:
+            txn: The database transaction.
+            old_rotate_stream_ordering: The previous maximum event stream ordering.
+            rotate_to_stream_ordering: The new maximum event stream ordering to summarise.
+
+        Returns whether the archiving process has caught up or not.
+        """
 
         # Calculate the new counts that should be upserted into event_push_summary
         sql = """
@@ -1093,9 +1125,9 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
     async def _remove_old_push_actions_that_have_rotated(
         self,
     ) -> None:
-        """Clear out old push actions that have been summarized."""
+        """Clear out old push actions that have been summarised."""
 
-        # We want to clear out anything that older than a day that *has* already
+        # We want to clear out anything that is older than a day that *has* already
         # been rotated.
         rotated_upto_stream_ordering = await self.db_pool.simple_select_one_onecol(
             table="event_push_summary_stream_ordering",
@@ -1119,7 +1151,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas
                 SELECT stream_ordering FROM event_push_actions
                 WHERE stream_ordering <= ? AND highlight = 0
                 ORDER BY stream_ordering ASC LIMIT 1 OFFSET ?
-            """,
+                """,
                 (
                     max_stream_ordering_to_delete,
                     batch_size,
@@ -1215,16 +1247,18 @@ class EventPushActionsStore(EventPushActionsWorkerStore):
 
             # NB. This assumes event_ids are globally unique since
             # it makes the query easier to index
-            sql = (
-                "SELECT epa.event_id, epa.room_id,"
-                " epa.stream_ordering, epa.topological_ordering,"
-                " epa.actions, epa.highlight, epa.profile_tag, e.received_ts"
-                " FROM event_push_actions epa, events e"
-                " WHERE epa.event_id = e.event_id"
-                " AND epa.user_id = ? %s"
-                " AND epa.notif = 1"
-                " ORDER BY epa.stream_ordering DESC"
-                " LIMIT ?" % (before_clause,)
+            sql = """
+                SELECT epa.event_id, epa.room_id,
+                    epa.stream_ordering, epa.topological_ordering,
+                    epa.actions, epa.highlight, epa.profile_tag, e.received_ts
+                FROM event_push_actions epa, events e
+                WHERE epa.event_id = e.event_id
+                    AND epa.user_id = ? %s
+                    AND epa.notif = 1
+                ORDER BY epa.stream_ordering DESC
+                LIMIT ?
+            """ % (
+                before_clause,
             )
             txn.execute(sql, args)
             return cast(

synapse/storage/databases/main/events.py

@@ -1490,7 +1490,7 @@ class PersistEventsStore:
                     event.sender,
                     "url" in event.content and isinstance(event.content["url"], str),
                     event.get_state_key(),
-                    context.rejected or None,
+                    context.rejected,
                 )
                 for event, context in events_and_contexts
             ),

synapse/storage/databases/main/events_worker.py

@@ -600,7 +600,11 @@ class EventsWorkerStore(SQLBaseStore):
         Returns:
             map from event id to result
         """
-        event_entry_map = await self._get_events_from_cache(
+        # Shortcut: check if we have any events in the *in memory* cache - this function
+        # may be called repeatedly for the same event so at this point we cannot reach
+        # out to any external cache for performance reasons. The external cache is
+        # checked later on in the `get_missing_events_from_cache_or_db` function below.
+        event_entry_map = self._get_events_from_local_cache(
             event_ids,
         )
 
@@ -632,7 +636,9 @@ class EventsWorkerStore(SQLBaseStore):
 
         if missing_events_ids:
 
-            async def get_missing_events_from_db() -> Dict[str, EventCacheEntry]:
+            async def get_missing_events_from_cache_or_db() -> Dict[
+                str, EventCacheEntry
+            ]:
                 """Fetches the events in `missing_event_ids` from the database.
 
                 Also creates entries in `self._current_event_fetches` to allow
@@ -657,10 +663,18 @@ class EventsWorkerStore(SQLBaseStore):
                 # the events have been redacted, and if so pulling the redaction event
                 # out of the database to check it.
                 #
+                missing_events = {}
                 try:
-                    missing_events = await self._get_events_from_db(
+                    # Try to fetch from any external cache. We already checked the
+                    # in-memory cache above.
+                    missing_events = await self._get_events_from_external_cache(
                         missing_events_ids,
                     )
+                    # Now actually fetch any remaining events from the DB
+                    db_missing_events = await self._get_events_from_db(
+                        missing_events_ids - missing_events.keys(),
+                    )
+                    missing_events.update(db_missing_events)
                 except Exception as e:
                     with PreserveLoggingContext():
                         fetching_deferred.errback(e)
@@ -679,7 +693,7 @@ class EventsWorkerStore(SQLBaseStore):
             # cancellations, since multiple `_get_events_from_cache_or_db` calls can
             # reuse the same fetch.
             missing_events: Dict[str, EventCacheEntry] = await delay_cancellation(
-                get_missing_events_from_db()
+                get_missing_events_from_cache_or_db()
             )
             event_entry_map.update(missing_events)
 
@@ -754,7 +768,54 @@ class EventsWorkerStore(SQLBaseStore):
     async def _get_events_from_cache(
         self, events: Iterable[str], update_metrics: bool = True
     ) -> Dict[str, EventCacheEntry]:
-        """Fetch events from the caches.
+        """Fetch events from the caches, both in memory and any external.
+
+        May return rejected events.
+
+        Args:
+            events: list of event_ids to fetch
+            update_metrics: Whether to update the cache hit ratio metrics
+        """
+        event_map = self._get_events_from_local_cache(
+            events, update_metrics=update_metrics
+        )
+
+        missing_event_ids = (e for e in events if e not in event_map)
+        event_map.update(
+            await self._get_events_from_external_cache(
+                events=missing_event_ids,
+                update_metrics=update_metrics,
+            )
+        )
+
+        return event_map
+
+    async def _get_events_from_external_cache(
+        self, events: Iterable[str], update_metrics: bool = True
+    ) -> Dict[str, EventCacheEntry]:
+        """Fetch events from any configured external cache.
+
+        May return rejected events.
+
+        Args:
+            events: list of event_ids to fetch
+            update_metrics: Whether to update the cache hit ratio metrics
+        """
+        event_map = {}
+
+        for event_id in events:
+            ret = await self._get_event_cache.get_external(
+                (event_id,), None, update_metrics=update_metrics
+            )
+            if ret:
+                event_map[event_id] = ret
+
+        return event_map
+
+    def _get_events_from_local_cache(
+        self, events: Iterable[str], update_metrics: bool = True
+    ) -> Dict[str, EventCacheEntry]:
+        """Fetch events from the local, in memory, caches.
 
         May return rejected events.
 
@@ -766,7 +827,7 @@ class EventsWorkerStore(SQLBaseStore):
 
         for event_id in events:
             # First check if it's in the event cache
-            ret = await self._get_event_cache.get(
+            ret = self._get_event_cache.get_local(
                 (event_id,), None, update_metrics=update_metrics
             )
             if ret:
@@ -788,7 +849,7 @@ class EventsWorkerStore(SQLBaseStore):
 
                 # We add the entry back into the cache as we want to keep
                 # recently queried events in the cache.
-                await self._get_event_cache.set((event_id,), cache_entry)
+                self._get_event_cache.set_local((event_id,), cache_entry)
 
         return event_map
 
@@ -2110,11 +2171,29 @@ class EventsWorkerStore(SQLBaseStore):
     def _get_partial_state_events_batch_txn(
         txn: LoggingTransaction, room_id: str
     ) -> List[str]:
+        # we want to work through the events from oldest to newest, so
+        # we only want events whose prev_events do *not* have partial state - hence
+        # the 'NOT EXISTS' clause in the below.
+        #
+        # This is necessary because ordering by stream ordering isn't quite enough
+        # to ensure that we work from oldest to newest event (in particular,
+        # if an event is initially persisted as an outlier and later de-outliered,
+        # it can end up with a lower stream_ordering than its prev_events).
+        #
+        # Typically this means we'll only return one event per batch, but that's
+        # hard to do much about.
+        #
+        # See also: https://github.com/matrix-org/synapse/issues/13001
         txn.execute(
             """
             SELECT event_id FROM partial_state_events AS pse
                 JOIN events USING (event_id)
-            WHERE pse.room_id = ?
+            WHERE pse.room_id = ? AND
+               NOT EXISTS(
+                  SELECT 1 FROM event_edges AS ee
+                     JOIN partial_state_events AS prev_pse ON (prev_pse.event_id=ee.prev_event_id)
+                     WHERE ee.event_id=pse.event_id
+               )
             ORDER BY events.stream_ordering
             LIMIT 100
             """,

synapse/storage/databases/main/relations.py

@@ -62,7 +62,6 @@ class RelationsWorkerStore(SQLBaseStore):
         room_id: str,
         relation_type: Optional[str] = None,
         event_type: Optional[str] = None,
-        aggregation_key: Optional[str] = None,
         limit: int = 5,
         direction: str = "b",
         from_token: Optional[StreamToken] = None,
@@ -76,7 +75,6 @@ class RelationsWorkerStore(SQLBaseStore):
             room_id: The room the event belongs to.
             relation_type: Only fetch events with this relation type, if given.
             event_type: Only fetch events with this event type, if given.
-            aggregation_key: Only fetch events with this aggregation key, if given.
             limit: Only fetch the most recent `limit` events.
             direction: Whether to fetch the most recent first (`"b"`) or the
                 oldest first (`"f"`).
@@ -105,10 +103,6 @@ class RelationsWorkerStore(SQLBaseStore):
             where_clause.append("type = ?")
             where_args.append(event_type)
 
-        if aggregation_key:
-            where_clause.append("aggregation_key = ?")
-            where_args.append(aggregation_key)
-
         pagination_clause = generate_pagination_where_clause(
             direction=direction,
             column_names=("topological_ordering", "stream_ordering"),

synapse/storage/databases/main/room.py

@@ -207,7 +207,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
     def _construct_room_type_where_clause(
         self, room_types: Union[List[Union[str, None]], None]
     ) -> Tuple[Union[str, None], List[str]]:
-        if not room_types or not self.config.experimental.msc3827_enabled:
+        if not room_types:
             return None, []
         else:
             # We use None when we want get rooms without a type

synapse/storage/databases/main/roommember.py

@@ -896,7 +896,7 @@ class RoomMemberWorkerStore(EventsWorkerStore):
         # We don't update the event cache hit ratio as it completely throws off
         # the hit ratio counts. After all, we don't populate the cache if we
         # miss it here
-        event_map = await self._get_events_from_cache(
+        event_map = self._get_events_from_local_cache(
             member_event_ids, update_metrics=False
         )
 

synapse/storage/databases/main/state.py

@@ -419,13 +419,15 @@ class StateGroupWorkerStore(EventsWorkerStore, SQLBaseStore):
         # anything that was rejected should have the same state as its
         # predecessor.
         if context.rejected:
-            assert context.state_group == context.state_group_before_event
+            state_group = context.state_group_before_event
+        else:
+            state_group = context.state_group
 
         self.db_pool.simple_update_txn(
             txn,
             table="event_to_state_groups",
             keyvalues={"event_id": event.event_id},
-            updatevalues={"state_group": context.state_group},
+            updatevalues={"state_group": state_group},
         )
 
         self.db_pool.simple_delete_one_txn(
@@ -440,7 +442,7 @@ class StateGroupWorkerStore(EventsWorkerStore, SQLBaseStore):
         txn.call_after(
             self._get_state_group_for_event.prefill,
             (event.event_id,),
-            context.state_group,
+            state_group,
         )
 
 

synapse/storage/databases/main/stream.py

@@ -58,6 +58,7 @@ from twisted.internet import defer
 from synapse.api.filtering import Filter
 from synapse.events import EventBase
 from synapse.logging.context import make_deferred_yieldable, run_in_background
+from synapse.logging.opentracing import trace
 from synapse.storage._base import SQLBaseStore
 from synapse.storage.database import (
     DatabasePool,
@@ -1346,6 +1347,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore):
 
         return rows, next_token
 
+    @trace
     async def paginate_room_events(
         self,
         room_id: str,

synapse/streams/events.py

@@ -21,6 +21,7 @@ from synapse.handlers.presence import PresenceEventSource
 from synapse.handlers.receipts import ReceiptEventSource
 from synapse.handlers.room import RoomEventSource
 from synapse.handlers.typing import TypingNotificationEventSource
+from synapse.logging.opentracing import trace
 from synapse.streams import EventSource
 from synapse.types import StreamToken
 
@@ -69,6 +70,7 @@ class EventSources:
         )
         return token
 
+    @trace
     async def get_current_token_for_pagination(self, room_id: str) -> StreamToken:
         """Get the current token for a given room to be used to paginate
         events.

synapse/util/caches/lrucache.py

@@ -834,9 +834,26 @@ class AsyncLruCache(Generic[KT, VT]):
     ) -> Optional[VT]:
         return self._lru_cache.get(key, update_metrics=update_metrics)
 
+    async def get_external(
+        self,
+        key: KT,
+        default: Optional[T] = None,
+        update_metrics: bool = True,
+    ) -> Optional[VT]:
+        # This method should fetch from any configured external cache, in this case noop.
+        return None
+
+    def get_local(
+        self, key: KT, default: Optional[T] = None, update_metrics: bool = True
+    ) -> Optional[VT]:
+        return self._lru_cache.get(key, update_metrics=update_metrics)
+
     async def set(self, key: KT, value: VT) -> None:
         self._lru_cache.set(key, value)
 
+    def set_local(self, key: KT, value: VT) -> None:
+        self._lru_cache.set(key, value)
+
     async def invalidate(self, key: KT) -> None:
         # This method should invalidate any external cache and then invalidate the LruCache.
         return self._lru_cache.invalidate(key)

synapse/util/ratelimitutils.py

@@ -21,7 +21,7 @@ from typing import Any, DefaultDict, Iterator, List, Set
 from twisted.internet import defer
 
 from synapse.api.errors import LimitExceededError
-from synapse.config.ratelimiting import FederationRateLimitConfig
+from synapse.config.ratelimiting import FederationRatelimitSettings
 from synapse.logging.context import (
     PreserveLoggingContext,
     make_deferred_yieldable,
@@ -36,7 +36,7 @@ logger = logging.getLogger(__name__)
 
 
 class FederationRateLimiter:
-    def __init__(self, clock: Clock, config: FederationRateLimitConfig):
+    def __init__(self, clock: Clock, config: FederationRatelimitSettings):
         def new_limiter() -> "_PerHostRatelimiter":
             return _PerHostRatelimiter(clock=clock, config=config)
 
@@ -63,7 +63,7 @@ class FederationRateLimiter:
 
 
 class _PerHostRatelimiter:
-    def __init__(self, clock: Clock, config: FederationRateLimitConfig):
+    def __init__(self, clock: Clock, config: FederationRatelimitSettings):
         """
         Args:
             clock

synapse/visibility.py

@@ -23,6 +23,7 @@ from synapse.api.constants import EventTypes, HistoryVisibility, Membership
 from synapse.events import EventBase
 from synapse.events.snapshot import EventContext
 from synapse.events.utils import prune_event
+from synapse.logging.opentracing import trace
 from synapse.storage.controllers import StorageControllers
 from synapse.storage.databases.main import DataStore
 from synapse.storage.state import StateFilter
@@ -51,6 +52,7 @@ MEMBERSHIP_PRIORITY = (
 _HISTORY_VIS_KEY: Final[Tuple[str, str]] = (EventTypes.RoomHistoryVisibility, "")
 
 
+@trace
 async def filter_events_for_client(
     storage: StorageControllers,
     user_id: str,

tests/handlers/test_directory.py

@@ -481,17 +481,13 @@ class TestCreatePublishedRoomACL(unittest.HomeserverTestCase):
 
         return config
 
-    def prepare(
-        self, reactor: MemoryReactor, clock: Clock, hs: HomeServer
-    ) -> HomeServer:
+    def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None:
         self.allowed_user_id = self.register_user(self.allowed_localpart, "pass")
         self.allowed_access_token = self.login(self.allowed_localpart, "pass")
 
         self.denied_user_id = self.register_user("denied", "pass")
         self.denied_access_token = self.login("denied", "pass")
 
-        return hs
-
     def test_denied_without_publication_permission(self) -> None:
         """
         Try to create a room, register an alias for it, and publish it,
@@ -575,9 +571,7 @@ class TestRoomListSearchDisabled(unittest.HomeserverTestCase):
 
     servlets = [directory.register_servlets, room.register_servlets]
 
-    def prepare(
-        self, reactor: MemoryReactor, clock: Clock, hs: HomeServer
-    ) -> HomeServer:
+    def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None:
         room_id = self.helper.create_room_as(self.user_id)
 
         channel = self.make_request(
@@ -588,8 +582,6 @@ class TestRoomListSearchDisabled(unittest.HomeserverTestCase):
         self.room_list_handler = hs.get_room_list_handler()
         self.directory_handler = hs.get_directory_handler()
 
-        return hs
-
     def test_disabling_room_list(self) -> None:
         self.room_list_handler.enable_room_list_search = True
         self.directory_handler.enable_room_list_search = True

tests/handlers/test_federation.py

@@ -14,6 +14,7 @@
 import logging
 from typing import cast
 from unittest import TestCase
+from unittest.mock import Mock, patch
 
 from twisted.test.proto_helpers import MemoryReactor
 
@@ -22,6 +23,7 @@ from synapse.api.errors import AuthError, Codes, LimitExceededError, SynapseErro
 from synapse.api.room_versions import RoomVersions
 from synapse.events import EventBase, make_event_from_dict
 from synapse.federation.federation_base import event_from_pdu_json
+from synapse.federation.federation_client import SendJoinResult
 from synapse.logging.context import LoggingContext, run_in_background
 from synapse.rest import admin
 from synapse.rest.client import login, room
@@ -30,7 +32,7 @@ from synapse.util import Clock
 from synapse.util.stringutils import random_string
 
 from tests import unittest
-from tests.test_utils import event_injection
+from tests.test_utils import event_injection, make_awaitable
 
 logger = logging.getLogger(__name__)
 
@@ -280,13 +282,21 @@ class FederationTestCase(unittest.FederatingHomeserverTestCase):
 
             # we poke this directly into _process_received_pdu, to avoid the
             # federation handler wanting to backfill the fake event.
+            state_handler = self.hs.get_state_handler()
+            context = self.get_success(
+                state_handler.compute_event_context(
+                    event,
+                    state_ids_before_event={
+                        (e.type, e.state_key): e.event_id for e in current_state
+                    },
+                    partial_state=False,
+                )
+            )
             self.get_success(
                 federation_event_handler._process_received_pdu(
                     self.OTHER_SERVER_NAME,
                     event,
-                    state_ids={
-                        (e.type, e.state_key): e.event_id for e in current_state
-                    },
+                    context,
                 )
             )
 
@@ -448,3 +458,121 @@ class EventFromPduTestCase(TestCase):
                 },
                 RoomVersions.V6,
             )
+
+
+class PartialJoinTestCase(unittest.FederatingHomeserverTestCase):
+    def test_failed_partial_join_is_clean(self) -> None:
+        """
+        Tests that, when failing to partial-join a room, we don't get stuck with
+        a partial-state flag on a room.
+        """
+
+        fed_handler = self.hs.get_federation_handler()
+        fed_client = fed_handler.federation_client
+
+        room_id = "!room:example.com"
+        membership_event = make_event_from_dict(
+            {
+                "room_id": room_id,
+                "type": "m.room.member",
+                "sender": "@alice:test",
+                "state_key": "@alice:test",
+                "content": {"membership": "join"},
+            },
+            RoomVersions.V10,
+        )
+
+        mock_make_membership_event = Mock(
+            return_value=make_awaitable(
+                (
+                    "example.com",
+                    membership_event,
+                    RoomVersions.V10,
+                )
+            )
+        )
+
+        EVENT_CREATE = make_event_from_dict(
+            {
+                "room_id": room_id,
+                "type": "m.room.create",
+                "sender": "@kristina:example.com",
+                "state_key": "",
+                "depth": 0,
+                "content": {"creator": "@kristina:example.com", "room_version": "10"},
+                "auth_events": [],
+                "origin_server_ts": 1,
+            },
+            room_version=RoomVersions.V10,
+        )
+        EVENT_CREATOR_MEMBERSHIP = make_event_from_dict(
+            {
+                "room_id": room_id,
+                "type": "m.room.member",
+                "sender": "@kristina:example.com",
+                "state_key": "@kristina:example.com",
+                "content": {"membership": "join"},
+                "depth": 1,
+                "prev_events": [EVENT_CREATE.event_id],
+                "auth_events": [EVENT_CREATE.event_id],
+                "origin_server_ts": 1,
+            },
+            room_version=RoomVersions.V10,
+        )
+        EVENT_INVITATION_MEMBERSHIP = make_event_from_dict(
+            {
+                "room_id": room_id,
+                "type": "m.room.member",
+                "sender": "@kristina:example.com",
+                "state_key": "@alice:test",
+                "content": {"membership": "invite"},
+                "depth": 2,
+                "prev_events": [EVENT_CREATOR_MEMBERSHIP.event_id],
+                "auth_events": [
+                    EVENT_CREATE.event_id,
+                    EVENT_CREATOR_MEMBERSHIP.event_id,
+                ],
+                "origin_server_ts": 1,
+            },
+            room_version=RoomVersions.V10,
+        )
+        mock_send_join = Mock(
+            return_value=make_awaitable(
+                SendJoinResult(
+                    membership_event,
+                    "example.com",
+                    state=[
+                        EVENT_CREATE,
+                        EVENT_CREATOR_MEMBERSHIP,
+                        EVENT_INVITATION_MEMBERSHIP,
+                    ],
+                    auth_chain=[
+                        EVENT_CREATE,
+                        EVENT_CREATOR_MEMBERSHIP,
+                        EVENT_INVITATION_MEMBERSHIP,
+                    ],
+                    partial_state=True,
+                    servers_in_room=["example.com"],
+                )
+            )
+        )
+
+        with patch.object(
+            fed_client, "make_membership_event", mock_make_membership_event
+        ), patch.object(fed_client, "send_join", mock_send_join):
+            # Join and check that our join event is rejected
+            # (The join event is rejected because it doesn't have any signatures)
+            join_exc = self.get_failure(
+                fed_handler.do_invite_join(["example.com"], room_id, "@alice:test", {}),
+                SynapseError,
+            )
+        self.assertIn("Join event was rejected", str(join_exc))
+
+        store = self.hs.get_datastores().main
+
+        # Check that we don't have a left-over partial_state entry.
+        self.assertFalse(
+            self.get_success(store.is_partial_state_room(room_id)),
+            f"Stale partial-stated room flag left over for {room_id} after a"
+            f" failed do_invite_join!",
+        )

tests/module_api/test_api.py

@@ -635,6 +635,76 @@ class ModuleApiTestCase(HomeserverTestCase):
             [{"set_tweak": "sound", "value": "default"}]
         )
 
+    def test_lookup_room_alias(self) -> None:
+        """Test that modules can resolve a room alias to a room ID."""
+        password = "password"
+        user_id = self.register_user("user", password)
+        access_token = self.login(user_id, password)
+        room_alias = "my-alias"
+        reference_room_id = self.helper.create_room_as(
+            tok=access_token, extra_content={"room_alias_name": room_alias}
+        )
+        self.assertIsNotNone(reference_room_id)
+
+        (room_id, _) = self.get_success(
+            self.module_api.lookup_room_alias(
+                f"#{room_alias}:{self.module_api.server_name}"
+            )
+        )
+
+        self.assertEqual(room_id, reference_room_id)
+
+    def test_create_room(self) -> None:
+        """Test that modules can create a room."""
+        # First test user validation (i.e. user is local).
+        self.get_failure(
+            self.module_api.create_room(
+                user_id=f"@user:{self.module_api.server_name}abc",
+                config={},
+                ratelimit=False,
+            ),
+            RuntimeError,
+        )
+
+        # Now do the happy path.
+        user_id = self.register_user("user", "password")
+        access_token = self.login(user_id, "password")
+
+        room_id, room_alias = self.get_success(
+            self.module_api.create_room(
+                user_id=user_id, config={"room_alias_name": "foo-bar"}, ratelimit=False
+            )
+        )
+
+        # Check room creator.
+        channel = self.make_request(
+            "GET",
+            f"/_matrix/client/v3/rooms/{room_id}/state/m.room.create",
+            access_token=access_token,
+        )
+        self.assertEqual(channel.code, 200, channel.result)
+        self.assertEqual(channel.json_body["creator"], user_id)
+
+        # Check room alias.
+        self.assertEquals(room_alias, f"#foo-bar:{self.module_api.server_name}")
+
+        # Let's try a room with no alias.
+        room_id, room_alias = self.get_success(
+            self.module_api.create_room(user_id=user_id, config={}, ratelimit=False)
+        )
+
+        # Check room creator.
+        channel = self.make_request(
+            "GET",
+            f"/_matrix/client/v3/rooms/{room_id}/state/m.room.create",
+            access_token=access_token,
+        )
+        self.assertEqual(channel.code, 200, channel.result)
+        self.assertEqual(channel.json_body["creator"], user_id)
+
+        # Check room alias.
+        self.assertIsNone(room_alias)
+
 
 class ModuleApiWorkerTestCase(BaseMultiWorkerStreamTestCase):
     """For testing ModuleApi functionality in a multi-worker setup"""

tests/rest/admin/test_room.py

@@ -1772,6 +1772,21 @@ class RoomTestCase(unittest.HomeserverTestCase):
             tok=admin_user_tok,
         )
 
+    def test_get_joined_members_after_leave_room(self) -> None:
+        """Test that requesting room members after leaving the room raises a 403 error."""
+
+        # create the room
+        user = self.register_user("foo", "pass")
+        user_tok = self.login("foo", "pass")
+        room_id = self.helper.create_room_as(user, tok=user_tok)
+        self.helper.leave(room_id, user, tok=user_tok)
+
+        # delete the rooms and get joined roomed membership
+        url = f"/_matrix/client/r0/rooms/{room_id}/joined_members"
+        channel = self.make_request("GET", url.encode("ascii"), access_token=user_tok)
+        self.assertEqual(HTTPStatus.FORBIDDEN, channel.code, msg=channel.json_body)
+        self.assertEqual(Codes.FORBIDDEN, channel.json_body["errcode"])
+
 
 class JoinAliasRoomTestCase(unittest.HomeserverTestCase):
 
@@ -1873,7 +1888,10 @@ class JoinAliasRoomTestCase(unittest.HomeserverTestCase):
         )
 
         self.assertEqual(HTTPStatus.NOT_FOUND, channel.code, msg=channel.json_body)
-        self.assertEqual("No known servers", channel.json_body["error"])
+        self.assertEqual(
+            "Can't join remote room because no servers that are in the room have been provided.",
+            channel.json_body["error"],
+        )
 
     def test_room_is_not_valid(self) -> None:
         """

tests/rest/client/test_relations.py

@@ -1060,6 +1060,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase):
                     participated, bundled_aggregations.get("current_user_participated")
                 )
                 # The latest thread event has some fields that don't matter.
+                self.assertIn("latest_event", bundled_aggregations)
                 self.assert_dict(
                     {
                         "content": {
@@ -1072,7 +1073,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase):
                         "sender": self.user2_id,
                         "type": "m.room.test",
                     },
-                    bundled_aggregations.get("latest_event"),
+                    bundled_aggregations["latest_event"],
                 )
 
             return assert_thread
@@ -1112,6 +1113,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase):
             self.assertEqual(2, bundled_aggregations.get("count"))
             self.assertTrue(bundled_aggregations.get("current_user_participated"))
             # The latest thread event has some fields that don't matter.
+            self.assertIn("latest_event", bundled_aggregations)
             self.assert_dict(
                 {
                     "content": {
@@ -1124,7 +1126,7 @@ class BundledAggregationsTestCase(BaseRelationsTestCase):
                     "sender": self.user_id,
                     "type": "m.room.test",
                 },
-                bundled_aggregations.get("latest_event"),
+                bundled_aggregations["latest_event"],
             )
             # Check the unsigned field on the latest event.
             self.assert_dict(

tests/rest/client/test_rooms.py

@@ -496,7 +496,7 @@ class RoomStateTestCase(RoomBase):
 
         self.assertEqual(HTTPStatus.OK, channel.code, msg=channel.result["body"])
         self.assertCountEqual(
-            [state_event["type"] for state_event in channel.json_body],
+            [state_event["type"] for state_event in channel.json_list],
             {
                 "m.room.create",
                 "m.room.power_levels",
@@ -2070,7 +2070,6 @@ class PublicRoomsRoomTypeFilterTestCase(unittest.HomeserverTestCase):
 
         config = self.default_config()
         config["allow_public_rooms_without_auth"] = True
-        config["experimental_features"] = {"msc3827_enabled": True}
         self.hs = self.setup_test_homeserver(config=config)
         self.url = b"/_matrix/client/r0/publicRooms"
 
@@ -2123,13 +2122,13 @@ class PublicRoomsRoomTypeFilterTestCase(unittest.HomeserverTestCase):
         chunk, count = self.make_public_rooms_request([None])
 
         self.assertEqual(count, 1)
-        self.assertEqual(chunk[0].get("org.matrix.msc3827.room_type", None), None)
+        self.assertEqual(chunk[0].get("room_type", None), None)
 
     def test_returns_only_space_based_on_filter(self) -> None:
         chunk, count = self.make_public_rooms_request(["m.space"])
 
         self.assertEqual(count, 1)
-        self.assertEqual(chunk[0].get("org.matrix.msc3827.room_type", None), "m.space")
+        self.assertEqual(chunk[0].get("room_type", None), "m.space")
 
     def test_returns_both_rooms_and_space_based_on_filter(self) -> None:
         chunk, count = self.make_public_rooms_request(["m.space", None])

tests/rest/client/test_sync.py

@@ -948,3 +948,24 @@ class ExcludeRoomTestCase(unittest.HomeserverTestCase):
 
         self.assertNotIn(self.excluded_room_id, channel.json_body["rooms"]["invite"])
         self.assertIn(self.included_room_id, channel.json_body["rooms"]["invite"])
+
+    def test_incremental_sync(self) -> None:
+        """Tests that activity in the room is properly filtered out of incremental
+        syncs.
+        """
+        channel = self.make_request("GET", "/sync", access_token=self.tok)
+        self.assertEqual(channel.code, 200, channel.result)
+        next_batch = channel.json_body["next_batch"]
+
+        self.helper.send(self.excluded_room_id, tok=self.tok)
+        self.helper.send(self.included_room_id, tok=self.tok)
+
+        channel = self.make_request(
+            "GET",
+            f"/sync?since={next_batch}",
+            access_token=self.tok,
+        )
+        self.assertEqual(channel.code, 200, channel.result)
+
+        self.assertNotIn(self.excluded_room_id, channel.json_body["rooms"]["join"])
+        self.assertIn(self.included_room_id, channel.json_body["rooms"]["join"])