-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Add type annotations to synapse.metrics
#10847
Conversation
3254187
to
5447ea9
Compare
5447ea9
to
94ef546
Compare
94ef546
to
f058e4f
Compare
@@ -53,7 +66,7 @@ | |||
|
|||
class RegistryProxy: | |||
@staticmethod | |||
def collect(): | |||
def collect() -> Iterable[Metric]: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I would think this needs to be a Generator
, but maybe not?
This is true for a variety of spots we use yield
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(Or maybe I can significantly simplify type hints in other spots!)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I opted for using the most basic type that callers want, to minimize assumptions, rather than documenting the concrete return type.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All Generators are Iterators (and all Iterators are Iterable), and in this case this seems cleaner than saying a Generator[Metric, None, None]
.
@@ -179,7 +179,7 @@ def __getattr__(self, name): | |||
|
|||
|
|||
# The type of entry which goes on our after_callbacks and exception_callbacks lists. | |||
_CallbackListEntry = Tuple[Callable[..., None], Iterable[Any], Dict[str, Any]] | |||
_CallbackListEntry = Tuple[Callable[..., object], Iterable[Any], Dict[str, Any]] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Curious why this is object
and not Any
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's to make mypy complain if we ever try to use the return values of the callbacks.
Everything is assignable to object
, but not the other way around, whereas Any
is assignable to everything without any checks.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
object
is the supertype of everything is what I think you're saying, so we'd get shouted at if we tried to do x: int = (something returning object)
whereas Any
is entirely untyped & unchecked.
I think this makes sense.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does this mean we should generally prefer object
over Any
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I haven't a clue. I imagine it's on a case by case basis. Any
is certainly a last resort for me, since it's equivalent to not annotating types at all.
"""Decorator that wraps a function that gets called as a background | ||
process. | ||
|
||
Equivalent of calling the function with `run_as_background_process` | ||
""" | ||
|
||
def wrap_as_background_process_inner(func): | ||
# NB: Return type is incorrect and should be F with a Deferred[Optional[R]] return |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this should work, but maybe not? This makes me a bit nervous that we're not mentioning this returns a Deferred
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure what to do about this. I updated the docstring above to call the change of return type out.
We can either annotate the return type correctly and lose the types of the arguments for the decorated function,
or annotate the return type incorrectly and preserve the types of the arguments.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I feel kind of bad about lying but appreciate the dilemma. :/
I'm not really sure what's best, though. Is lying going to cause type errors for users of the function?
I would probably make the docstring louder about the return type being a lie, though.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I tried to improve the docstring further
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's fine as - is.
question though, what happens if you bound F
to return an Awaitable?
F = TypeVar("F", bound=Callable[..., Awaitable])
(or something similar?)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's fine as - is.
to be clear: I'm saying I'm happy with the PR as it stands.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
question though, what happens if you bound F to return an Awaitable?
F = TypeVar("F", bound=Callable[..., Awaitable])
Ooh. mypy throws up 3 errors which all look trivial to fix:
synapse/util/caches/expiringcache.py:86: error: Argument 2 to "run_as_background_process" has incompatible type "Callable[[], None]"; expected "Callable[..., Awaitable[Any]]" [arg-type]
synapse/app/_base.py:404: error: Value of type variable "F" of function cannot be "Callable[[VarArg(Any), KwArg(Any)], None]" [type-var]
synapse/handlers/typing.py:92: error: Value of type variable "F" of function cannot be "Callable[[FollowerTypingHandler], None]" [type-var]
So it sounds like the cleanest thing to do is to restrict run_as_background_process
to only accept Awaitable
-returning functions and drop the whole maybe_awaitable
thing. I'll make the change when I find some time.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've gotten rid of the maybe_awaitable
.
|
||
|
||
@overload | ||
def run_as_background_process( # type: ignore[misc] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What's being ignored here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's to suppress error: Overloaded function signatures 1 and 2 overlap with incompatible return types [misc]
. I shall add a comment
I initially tried a Union
but ran into a bug/deficiency in mypy where it couldn't figure out what R
was at every call site.
#10847 (comment)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If you have a tracking link to e.g. a mypy issue or something of the sort, could be good. (and maybe add a remark that you tried Union[Awaitable[R], R]
or whatever, to save someone for thinking 'ha! I can see how to do that one!' and falling into the same pit)
I'm a bit sad Union[Awaitable[R], R]
wouldn't have done the trick though, not your fault though
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I tried to improve the comments further
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wonder if updating mypy in #11006 will have changed this situation at all?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good shout! I gave it another go and mypy is still unhappy without the overloads (version 0.910?)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Shame. Worth a try!
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This all seems to make sense. I'm a little bit concerned about the lie and would probably want to ask someone else on the team about that ...
@@ -53,7 +66,7 @@ | |||
|
|||
class RegistryProxy: | |||
@staticmethod | |||
def collect(): | |||
def collect() -> Iterable[Metric]: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All Generators are Iterators (and all Iterators are Iterable), and in this case this seems cleaner than saying a Generator[Metric, None, None]
.
synapse/metrics/__init__.py
Outdated
# Placeholder for `InFlightGauge._metrics_class`, which is created dynamically | ||
_MetricsEntry = Any |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
is this common practice in our code?
Is it possible/useful to create a Protocol for it (if we have no other choice)?
Right now, Any
doesn't tell mypy (or me, or my IDE) a whole lot, though I do appreciate that you've taken the time to common it out into a common label :).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ah I see the fields are created dynamically. Makes sense, even if a bit sad.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm taking a look at adding more typing to synapse.utils
and got stuck on annotating metrics
here. I think it should be one of these dynamically-created types but can't annotate it meaningfully.
So +1 to any other smart approaches!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Python typings are really frustrating. The equivalent in typescript would be something like: this
Perhaps we can add a type parameter to InFlightGauge
and make the calling code provide an appropriate Protocol
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok. InFlightGauge
now takes a type parameter, so now those callbacks can have proper type annotations.
mypy forced me to annotate in_flight
in synapse/util/metrics.py (sorry!).
Two caveats:
- mypy won't allow the type parameter to be constrained to a
Protocol
- there's no check that the names passed to the
InFlightGauge
constructor match the attributes in the protocol
F = TypeVar("F", bound=Callable[..., Any]) | ||
|
||
|
||
def runUntilCurrentTimer(reactor: ReactorBase, func: F) -> F: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nice
synapse/metrics/_exposition.py
Outdated
@@ -41,7 +43,7 @@ | |||
MINUS_INF = float("-inf") | |||
|
|||
|
|||
def floatToGoString(d): | |||
def floatToGoString(d: Any) -> str: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
could we come up with something a bit stronger than Any
here?
Union[float, str]
(if that's what is needed)?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's typeof(Sample.value)
and I had a really hard time figuring out what that was, since prometheus_client
doesn't have typings.
Though it seems that prometheus_client
pinky promises that it's an int
or float
:
https://github.com/prometheus/client_python/blob/master/prometheus_client/samples.py#L35-L39
I shall update the function!
|
||
|
||
@overload | ||
def run_as_background_process( # type: ignore[misc] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If you have a tracking link to e.g. a mypy issue or something of the sort, could be good. (and maybe add a remark that you tried Union[Awaitable[R], R]
or whatever, to save someone for thinking 'ha! I can see how to do that one!' and falling into the same pit)
I'm a bit sad Union[Awaitable[R], R]
wouldn't have done the trick though, not your fault though
"""Decorator that wraps a function that gets called as a background | ||
process. | ||
|
||
Equivalent of calling the function with `run_as_background_process` | ||
""" | ||
|
||
def wrap_as_background_process_inner(func): | ||
# NB: Return type is incorrect and should be F with a Deferred[Optional[R]] return |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I feel kind of bad about lying but appreciate the dilemma. :/
I'm not really sure what's best, though. Is lying going to cause type errors for users of the function?
I would probably make the docstring louder about the return type being a lie, though.
@@ -179,7 +179,7 @@ def __getattr__(self, name): | |||
|
|||
|
|||
# The type of entry which goes on our after_callbacks and exception_callbacks lists. | |||
_CallbackListEntry = Tuple[Callable[..., None], Iterable[Any], Dict[str, Any]] | |||
_CallbackListEntry = Tuple[Callable[..., object], Iterable[Any], Dict[str, Any]] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
object
is the supertype of everything is what I think you're saying, so we'd get shouted at if we tried to do x: int = (something returning object)
whereas Any
is entirely untyped & unchecked.
I think this makes sense.
Didn't get this to pass `no-untyped-def`, think I'll need to watch #10847
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
going to collect some thoughts about whether we feel that it's OK to lie; otherwise LGTM
The following modules now pass `disallow_untyped_defs`: * synapse.util.caches.cached_call * synapse.util.caches.lrucache * synapse.util.caches.response_cache * synapse.util.caches.stream_change_cache * synapse.util.caches.ttlcache pass * synapse.util.daemonize * synapse.util.patch_inline_callbacks pass `no-untyped-defs` * synapse.util.versionstring Additional typing in synapse.util.metrics. Didn't get this to pass `no-untyped-defs`, think I'll need to watch #10847
|
||
|
||
@overload | ||
def run_as_background_process( # type: ignore[misc] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wonder if updating mypy in #11006 will have changed this situation at all?
What was the status of this change? I think we were keen to see it done and wouldn't want it to fall through the cracks! Edit: I guess Olivier wanted to think things over in #10847 (review) |
NB that edits don't seen by those of us that read our GH notifications via email. I think this is somewhere between @squahtx and @reivilibre to sort out currently. |
Thanks for pointing this out to me! I didn't realise I was holding it up. I think the only thing on my mind was #10847 (comment) really — if other people are happy with an innocent lie, then I'm happy to roll it through (and it sounds like, bear in mind that we can one day use |
Awaiting changes from myself, to restrict |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
Callable[..., Awaitable[R]], | ||
Callable[..., R], | ||
], | ||
func: Callable[..., Awaitable[Optional[R]]], |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No problem at all, but curious if this needed to be Optional[R]
— I naïvely expect it would be fine to let it be R
since R
is variable.
It won't matter anyway since the return position of a Callable is contravariant (fn(...) -> R
is-a fn(...) -> Optional[R]
), and this function always returns Optional[R]
regardless because of the error handling.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You're right, it ought to be Awaitable[R]
mypy gets unhappy elsewhere if I change it:
synapse/metrics/background_process_metrics.py:273: error: Argument 2 to "run_as_background_process" has incompatible type "F"; expected "Callable[..., Awaitable[R]]" [arg-type]
Synapse 1.48.0 (2021-11-30) =========================== This release removes support for the long-deprecated `trust_identity_server_for_password_resets` configuration flag. This release also fixes some performance issues with some background database updates introduced in Synapse 1.47.0. No significant changes since 1.48.0rc1. Synapse 1.48.0rc1 (2021-11-25) ============================== Features -------- - Experimental support for the thread relation defined in [MSC3440](matrix-org/matrix-spec-proposals#3440). ([\#11161](matrix-org/synapse#11161)) - Support filtering by relation senders & types per [MSC3440](matrix-org/matrix-spec-proposals#3440). ([\#11236](matrix-org/synapse#11236)) - Add support for the `/_matrix/client/v3` and `/_matrix/media/v3` APIs from Matrix v1.1. ([\#11318](matrix-org/synapse#11318), [\#11371](matrix-org/synapse#11371)) - Support the stable version of [MSC2778](matrix-org/matrix-spec-proposals#2778): the `m.login.application_service` login type. Contributed by @tulir. ([\#11335](matrix-org/synapse#11335)) - Add a new version of delete room admin API `DELETE /_synapse/admin/v2/rooms/<room_id>` to run it in the background. Contributed by @dklimpel. ([\#11223](matrix-org/synapse#11223)) - Allow the admin [Delete Room API](https://matrix-org.github.io/synapse/latest/admin_api/rooms.html#delete-room-api) to block a room without the need to join it. ([\#11228](matrix-org/synapse#11228)) - Add an admin API to un-shadow-ban a user. ([\#11347](matrix-org/synapse#11347)) - Add an admin API to run background database schema updates. ([\#11352](matrix-org/synapse#11352)) - Add an admin API for blocking a room. ([\#11324](matrix-org/synapse#11324)) - Update the JWT login type to support custom a `sub` claim. ([\#11361](matrix-org/synapse#11361)) - Store and allow querying of arbitrary event relations. ([\#11391](matrix-org/synapse#11391)) Bugfixes -------- - Fix a long-standing bug wherein display names or avatar URLs containing null bytes cause an internal server error when stored in the DB. ([\#11230](matrix-org/synapse#11230)) - Prevent [MSC2716](matrix-org/matrix-spec-proposals#2716) historical state events from being pushed to an application service via `/transactions`. ([\#11265](matrix-org/synapse#11265)) - Fix a long-standing bug where uploading extremely thin images (e.g. 1000x1) would fail. Contributed by @Neeeflix. ([\#11288](matrix-org/synapse#11288)) - Fix a bug, introduced in Synapse 1.46.0, which caused the `check_3pid_auth` and `on_logged_out` callbacks in legacy password authentication provider modules to not be registered. Modules using the generic module interface were not affected. ([\#11340](matrix-org/synapse#11340)) - Fix a bug introduced in 1.41.0 where space hierarchy responses would be incorrectly reused if multiple users were to make the same request at the same time. ([\#11355](matrix-org/synapse#11355)) - Fix a bug introduced in 1.45.0 where the `read_templates` method of the module API would error. ([\#11377](matrix-org/synapse#11377)) - Fix an issue introduced in 1.47.0 which prevented servers re-joining rooms they had previously left, if their signing keys were replaced. ([\#11379](matrix-org/synapse#11379)) - Fix a bug introduced in 1.13.0 where creating and publishing a room could cause errors if `room_list_publication_rules` is configured. ([\#11392](matrix-org/synapse#11392)) - Improve performance of various background database updates. ([\#11421](matrix-org/synapse#11421), [\#11422](matrix-org/synapse#11422)) Improved Documentation ---------------------- - Suggest users of the Debian packages add configuration to `/etc/matrix-synapse/conf.d/` to prevent, upon upgrade, being asked to choose between their configuration and the maintainer's. ([\#11281](matrix-org/synapse#11281)) - Fix typos in the documentation for the `username_available` admin API. Contributed by Stanislav Motylkov. ([\#11286](matrix-org/synapse#11286)) - Add Single Sign-On, SAML and CAS pages to the documentation. ([\#11298](matrix-org/synapse#11298)) - Change the word 'Home server' as one word 'homeserver' in documentation. ([\#11320](matrix-org/synapse#11320)) - Fix missing quotes for wildcard domains in `federation_certificate_verification_whitelist`. ([\#11381](matrix-org/synapse#11381)) Deprecations and Removals ------------------------- - Remove deprecated `trust_identity_server_for_password_resets` configuration flag. ([\#11333](matrix-org/synapse#11333), [\#11395](matrix-org/synapse#11395)) Internal Changes ---------------- - Add type annotations to `synapse.metrics`. ([\#10847](matrix-org/synapse#10847)) - Split out federated PDU retrieval function into a non-cached version. ([\#11242](matrix-org/synapse#11242)) - Clean up code relating to to-device messages and sending ephemeral events to application services. ([\#11247](matrix-org/synapse#11247)) - Fix a small typo in the error response when a relation type other than 'm.annotation' is passed to `GET /rooms/{room_id}/aggregations/{event_id}`. ([\#11278](matrix-org/synapse#11278)) - Drop unused database tables `room_stats_historical` and `user_stats_historical`. ([\#11280](matrix-org/synapse#11280)) - Require all files in synapse/ and tests/ to pass mypy unless specifically excluded. ([\#11282](matrix-org/synapse#11282), [\#11285](matrix-org/synapse#11285), [\#11359](matrix-org/synapse#11359)) - Add missing type hints to `synapse.app`. ([\#11287](matrix-org/synapse#11287)) - Remove unused parameters on `FederationEventHandler._check_event_auth`. ([\#11292](matrix-org/synapse#11292)) - Add type hints to `synapse._scripts`. ([\#11297](matrix-org/synapse#11297)) - Fix an issue which prevented the `remove_deleted_devices_from_device_inbox` background database schema update from running when updating from a recent Synapse version. ([\#11303](matrix-org/synapse#11303)) - Add type hints to storage classes. ([\#11307](matrix-org/synapse#11307), [\#11310](matrix-org/synapse#11310), [\#11311](matrix-org/synapse#11311), [\#11312](matrix-org/synapse#11312), [\#11313](matrix-org/synapse#11313), [\#11314](matrix-org/synapse#11314), [\#11316](matrix-org/synapse#11316), [\#11322](matrix-org/synapse#11322), [\#11332](matrix-org/synapse#11332), [\#11339](matrix-org/synapse#11339), [\#11342](matrix-org/synapse#11342)) - Add type hints to `synapse.util`. ([\#11321](matrix-org/synapse#11321), [\#11328](matrix-org/synapse#11328)) - Improve type annotations in Synapse's test suite. ([\#11323](matrix-org/synapse#11323), [\#11330](matrix-org/synapse#11330)) - Test that room alias deletion works as intended. ([\#11327](matrix-org/synapse#11327)) - Add type annotations for some methods and properties in the module API. ([\#11341](matrix-org/synapse#11341)) - Fix running `scripts-dev/complement.sh`, which was broken in v1.47.0rc1. ([\#11368](matrix-org/synapse#11368)) - Rename internal functions for token generation to better reflect what they do. ([\#11369](matrix-org/synapse#11369), [\#11370](matrix-org/synapse#11370)) - Add type hints to configuration classes. ([\#11377](matrix-org/synapse#11377)) - Publish a `develop` image to Docker Hub. ([\#11380](matrix-org/synapse#11380)) - Keep fallback key marked as used if it's re-uploaded. ([\#11382](matrix-org/synapse#11382)) - Use `auto_attribs` on the `attrs` class `RefreshTokenLookupResult`. ([\#11386](matrix-org/synapse#11386)) - Rename unstable `access_token_lifetime` configuration option to `refreshable_access_token_lifetime` to make it clear it only concerns refreshable access tokens. ([\#11388](matrix-org/synapse#11388)) - Do not run the broken MSC2716 tests when running `scripts-dev/complement.sh`. ([\#11389](matrix-org/synapse#11389)) - Remove dead code from supporting ACME. ([\#11393](matrix-org/synapse#11393)) - Refactor including the bundled relations when serializing an event. ([\#11408](matrix-org/synapse#11408))
After #10847, `looping_background_call` would print an error in the logs every time a non-async function was called. Since the error would be caught and ignored immediately, there were no other side effects.
After #10847, `looping_background_call` would print an error in the logs every time a non-async function was called. Since the error would be caught and ignored immediately, there were no other side effects.
Synapse 1.48.0 (2021-11-30) =========================== This release removes support for the long-deprecated `trust_identity_server_for_password_resets` configuration flag. This release also fixes some performance issues with some background database updates introduced in Synapse 1.47.0. No significant changes since 1.48.0rc1. Synapse 1.48.0rc1 (2021-11-25) ============================== Features -------- - Experimental support for the thread relation defined in [MSC3440](matrix-org/matrix-spec-proposals#3440). ([\matrix-org#11161](matrix-org#11161)) - Support filtering by relation senders & types per [MSC3440](matrix-org/matrix-spec-proposals#3440). ([\matrix-org#11236](matrix-org#11236)) - Add support for the `/_matrix/client/v3` and `/_matrix/media/v3` APIs from Matrix v1.1. ([\matrix-org#11318](matrix-org#11318), [\matrix-org#11371](matrix-org#11371)) - Support the stable version of [MSC2778](matrix-org/matrix-spec-proposals#2778): the `m.login.application_service` login type. Contributed by @tulir. ([\matrix-org#11335](matrix-org#11335)) - Add a new version of delete room admin API `DELETE /_synapse/admin/v2/rooms/<room_id>` to run it in the background. Contributed by @dklimpel. ([\matrix-org#11223](matrix-org#11223)) - Allow the admin [Delete Room API](https://matrix-org.github.io/synapse/latest/admin_api/rooms.html#delete-room-api) to block a room without the need to join it. ([\matrix-org#11228](matrix-org#11228)) - Add an admin API to un-shadow-ban a user. ([\matrix-org#11347](matrix-org#11347)) - Add an admin API to run background database schema updates. ([\matrix-org#11352](matrix-org#11352)) - Add an admin API for blocking a room. ([\matrix-org#11324](matrix-org#11324)) - Update the JWT login type to support custom a `sub` claim. ([\matrix-org#11361](matrix-org#11361)) - Store and allow querying of arbitrary event relations. ([\matrix-org#11391](matrix-org#11391)) Bugfixes -------- - Fix a long-standing bug wherein display names or avatar URLs containing null bytes cause an internal server error when stored in the DB. ([\matrix-org#11230](matrix-org#11230)) - Prevent [MSC2716](matrix-org/matrix-spec-proposals#2716) historical state events from being pushed to an application service via `/transactions`. ([\matrix-org#11265](matrix-org#11265)) - Fix a long-standing bug where uploading extremely thin images (e.g. 1000x1) would fail. Contributed by @Neeeflix. ([\matrix-org#11288](matrix-org#11288)) - Fix a bug, introduced in Synapse 1.46.0, which caused the `check_3pid_auth` and `on_logged_out` callbacks in legacy password authentication provider modules to not be registered. Modules using the generic module interface were not affected. ([\matrix-org#11340](matrix-org#11340)) - Fix a bug introduced in 1.41.0 where space hierarchy responses would be incorrectly reused if multiple users were to make the same request at the same time. ([\matrix-org#11355](matrix-org#11355)) - Fix a bug introduced in 1.45.0 where the `read_templates` method of the module API would error. ([\matrix-org#11377](matrix-org#11377)) - Fix an issue introduced in 1.47.0 which prevented servers re-joining rooms they had previously left, if their signing keys were replaced. ([\matrix-org#11379](matrix-org#11379)) - Fix a bug introduced in 1.13.0 where creating and publishing a room could cause errors if `room_list_publication_rules` is configured. ([\matrix-org#11392](matrix-org#11392)) - Improve performance of various background database updates. ([\matrix-org#11421](matrix-org#11421), [\matrix-org#11422](matrix-org#11422)) Improved Documentation ---------------------- - Suggest users of the Debian packages add configuration to `/etc/matrix-synapse/conf.d/` to prevent, upon upgrade, being asked to choose between their configuration and the maintainer's. ([\matrix-org#11281](matrix-org#11281)) - Fix typos in the documentation for the `username_available` admin API. Contributed by Stanislav Motylkov. ([\matrix-org#11286](matrix-org#11286)) - Add Single Sign-On, SAML and CAS pages to the documentation. ([\matrix-org#11298](matrix-org#11298)) - Change the word 'Home server' as one word 'homeserver' in documentation. ([\matrix-org#11320](matrix-org#11320)) - Fix missing quotes for wildcard domains in `federation_certificate_verification_whitelist`. ([\matrix-org#11381](matrix-org#11381)) Deprecations and Removals ------------------------- - Remove deprecated `trust_identity_server_for_password_resets` configuration flag. ([\matrix-org#11333](matrix-org#11333), [\matrix-org#11395](matrix-org#11395)) Internal Changes ---------------- - Add type annotations to `synapse.metrics`. ([\matrix-org#10847](matrix-org#10847)) - Split out federated PDU retrieval function into a non-cached version. ([\matrix-org#11242](matrix-org#11242)) - Clean up code relating to to-device messages and sending ephemeral events to application services. ([\matrix-org#11247](matrix-org#11247)) - Fix a small typo in the error response when a relation type other than 'm.annotation' is passed to `GET /rooms/{room_id}/aggregations/{event_id}`. ([\matrix-org#11278](matrix-org#11278)) - Drop unused database tables `room_stats_historical` and `user_stats_historical`. ([\matrix-org#11280](matrix-org#11280)) - Require all files in synapse/ and tests/ to pass mypy unless specifically excluded. ([\matrix-org#11282](matrix-org#11282), [\matrix-org#11285](matrix-org#11285), [\matrix-org#11359](matrix-org#11359)) - Add missing type hints to `synapse.app`. ([\matrix-org#11287](matrix-org#11287)) - Remove unused parameters on `FederationEventHandler._check_event_auth`. ([\matrix-org#11292](matrix-org#11292)) - Add type hints to `synapse._scripts`. ([\matrix-org#11297](matrix-org#11297)) - Fix an issue which prevented the `remove_deleted_devices_from_device_inbox` background database schema update from running when updating from a recent Synapse version. ([\matrix-org#11303](matrix-org#11303)) - Add type hints to storage classes. ([\matrix-org#11307](matrix-org#11307), [\matrix-org#11310](matrix-org#11310), [\matrix-org#11311](matrix-org#11311), [\matrix-org#11312](matrix-org#11312), [\matrix-org#11313](matrix-org#11313), [\matrix-org#11314](matrix-org#11314), [\matrix-org#11316](matrix-org#11316), [\matrix-org#11322](matrix-org#11322), [\matrix-org#11332](matrix-org#11332), [\matrix-org#11339](matrix-org#11339), [\matrix-org#11342](matrix-org#11342)) - Add type hints to `synapse.util`. ([\matrix-org#11321](matrix-org#11321), [\matrix-org#11328](matrix-org#11328)) - Improve type annotations in Synapse's test suite. ([\matrix-org#11323](matrix-org#11323), [\matrix-org#11330](matrix-org#11330)) - Test that room alias deletion works as intended. ([\matrix-org#11327](matrix-org#11327)) - Add type annotations for some methods and properties in the module API. ([\matrix-org#11341](matrix-org#11341)) - Fix running `scripts-dev/complement.sh`, which was broken in v1.47.0rc1. ([\matrix-org#11368](matrix-org#11368)) - Rename internal functions for token generation to better reflect what they do. ([\matrix-org#11369](matrix-org#11369), [\matrix-org#11370](matrix-org#11370)) - Add type hints to configuration classes. ([\matrix-org#11377](matrix-org#11377)) - Publish a `develop` image to Docker Hub. ([\matrix-org#11380](matrix-org#11380)) - Keep fallback key marked as used if it's re-uploaded. ([\matrix-org#11382](matrix-org#11382)) - Use `auto_attribs` on the `attrs` class `RefreshTokenLookupResult`. ([\matrix-org#11386](matrix-org#11386)) - Rename unstable `access_token_lifetime` configuration option to `refreshable_access_token_lifetime` to make it clear it only concerns refreshable access tokens. ([\matrix-org#11388](matrix-org#11388)) - Do not run the broken MSC2716 tests when running `scripts-dev/complement.sh`. ([\matrix-org#11389](matrix-org#11389)) - Remove dead code from supporting ACME. ([\matrix-org#11393](matrix-org#11393)) - Refactor including the bundled relations when serializing an event. ([\matrix-org#11408](matrix-org#11408))
No description provided.