Instead of proxying through the magic getter of the RootConfig
object. This should be more performant (and is more explicit).

This fixes mypy errors with jaeger-client >= 4.7.0 and should be a no-op
for versions before that.

By copying the `room_type` field of the create event and
migrating any non-empty `m.space.child` events to the
new room that is created.

No functional changes here. This came out as I was working to tackle #5677 

We want to blacklist only known, consistent failures. We should deflake tests rather than ignoring them.

Fixes #10732: consent flow skipped during SSO user registration if username is left at default

Signed-off-by: Andrew Ferrazzutti <fair@miscworks.net>

We've already batched up the events previously, and assume in other
places in the events.py file that we have. Removing this makes it easier
to adjust the batch sizes in one place.

Signed-off-by: Sean Quah <seanq@element.io>

Hint to clients via the room capabilities API (MSC3244) that
room version 9 should be preferred for creating a room with
restricted join rules (instead of room version 8).

This is only called in two places, and the code seems much clearer without it.

* Split up the documentation in several files rather than one huge one
* Add examples for each callback category
* Other niceties like fixing https://github.com/matrix-org/synapse/issues/10632
* Add titles to callbacks so they're easier to find in the navigation panels and link to

When releasing 1.42.0 with @Azrenbeth and talking with @clokep yesterday I realised doing the dch incantations related to releasing Synapse wasn't trivial on eg a macOS system, so this is a script to run in a Debian container to make things a bit easier.

Also refactors some of the registration of endpoints on workers.

Outlier events don't ever have push actions associated with them, so we
can skip some expensive queries during event persistence.

This used to be a comma and got accidentally changed to a period in #9654, but a pipe character is more easier to parse visually.

This adds the format to the request arguments / URL to
ensure that JSON data is returned (which is all that
Synapse supports).

This also adds additional error checking / filtering to the
configuration file to ignore XML-only providers.

I think I have finally teased apart the codepaths which handle outliers, and those that handle non-outliers. 
Let's add some assertions to demonstrate my newfound knowledge.

If we're persisting an event E which has auth_events A1, A2, then we ought to make sure that we correctly auth
and persist A1 and A2, before we blindly accept E.

This PR does part of that - it persists the auth events first - but it does not fully solve the problem, because we
still don't check that the auth events weren't rejected.

The full event content cannot be trusted from this API (as no auth
chain, etc.) is processed over federation. Returning the full event
content was a bug as MSC2946 specifies that only the stripped
state should be returned.

This also avoids calculating aggregations / annotations which go
unused.

Synapse 1.42.0rc2 (2021-09-06)
==============================

This version of Synapse removes deprecated room-management admin APIs, removes out-of-date
email pushers, and improves error handling for fallback templates for user-interactive
authentication. For more information on these points, server administrators are
encouraged to read [the upgrade notes](docs/upgrade.md#upgrading-to-v1420).

Features
--------

- Support room version 9 from [MSC3375](https://github.com/matrix-org/matrix-doc/pull/3375). ([\#10747](https://github.com/matrix-org/synapse/issues/10747))

Internal Changes
----------------

- Print a warning when using one of the deprecated `template_dir` settings. ([\#10768](https://github.com/matrix-org/synapse/issues/10768))

The deprecation itself happened in #10596 which shipped with Synapse v1.41.0. However, it doesn't seem fair to suddenly drop support for these settings in ~4-6w without being more vocal about said deprecation.

It's now only used in a couple of places, so we can drop it altogether.