This timeout strategy using asynchronous exceptions has a number of
safety caveats (read the docstring!!) and should only be used in very
specific circumstances.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
This is preparatory work towards adding a Topic model.
We plan to use the local variable name as 'topic' for
the Topic model objects.
Currently, we use *topic as the local variable name for
topic names.
We rename local variables of the form *topic to *topic_name
so that we don't need to think about type collisions in
individual code paths where we might want to talk about both
Topic objects and strings for the topic name.
This is preparatory work towards adding a Topic model.
We plan to use the local variable name as 'topic' for
the Topic model objects.
Currently, we use *topic as the local variable name for
topic names.
We rename local variables of the form *topic to *topic_name
so that we don't need to think about type collisions in
individual code paths where we might want to talk about both
Topic objects and strings for the topic name.
This creates a valid registration, for two reasons:
1. Avoid the need to run "manage.py register_server" in dev env to
register, when wanting to to test stuff with
`PUSH_NOTIFICATION_BOUNCER_URL = "http://localhost:9991"`.
2. Avoid breaking RemoteRealm syncing, due to duplicate registrations
(first set of registrations that gets set up with the dummy
RemoteZulipServer in populate_db, and the second that gets set up via
the regular syncing mechanism with the new RemoteZulipServer created
during register_server).
This default setup will be more realistic, matching the ordinary
conditions for a modern server.
Especially needed as we add bouncer code that will expect to have
RemoteRealm entries for realm_uuid values for which it receives
requests.
We now pass bogus data for inaccessible users when sending
the users data in "realm_users" field of "register" response
or when using endpoints like "GET /users" to get data of
all the users in realm.
We would add a client capability field in future commits
such that new clients would receive data only for accessible
users and they can form the bogus data by themselves.
The event for stream typing notifications is no longer sent
to the long_term_idle subscribers of the stream.
This helps to reduce the tornado's work of parsing super-long
JSON-encoded lists of user IDs in large streams. Now the lists
are shorter.
This commit renames the two unused and historical bits of the
'fields' bitfield of the 'UserMessage' and 'ArchivedUserMessage'
tables.
* 'summarize_in_home' to 'topic_wildcard_mentioned'
* 'summarize_in_stream' to 'group_mentioned'
The 'group_mentioned' flag doesn't affect the feature,
but completing the work here helps to save future migration
and indexing efforts on the UserMessage table, as we plan to
use this flag in the future for group mentions.
The unused bits may have old data; we'll clear that in
a separate commit.
It creates the 'zerver_usermessage_any_mentioned_message_id'
index concurrently.
This commit moves constants for system group names to a new
"SystemGroups" class so that we can use these group names
in multiple classes in models.py without worrying about the
order of defining them.
Previous behavior-
- Guest did not receive stream creation events for new
web-public streams.
- Guest did not receive peer_add and peer_remove events
for web-public and subscribed public streams.
This commit fixes the behavior to be -
- Guests now receive stream creation events for new
web-public streams.
- Guest now receive peer_add and peer_remove events for
web-public and subscribed public streams.
In this commit, we add a new dropdown 'Organization language' on
the `/new` and `/realm/register` pages. This dropdown allows setting
the language of the organization during its creation. This allows
messages from Welcome Bot and introductory messages in streams to be
internationalized.
Fixes a part of #25729.
Add an optional `automatic_new_visibility_policy` enum field
in the success response to indicate the new visibility policy
value due to the `automatically_follow_topics_policy` and
`automatically_unmute_topics_in_muted_streams_policy` user settings
during the send message action.
Only present if there is a change in the visibility policy.
The `expected` flag was incredibly confusing, as you
couldn't tell from the calling code what you were
actually expecting to happen.
I avoid the context manager idiom in order to force
the callers to create simple helper functions, and
I de-duplicate some code in some places.
I also force the caller to explicitly soft-deactivate
the user with one simple line of code, so that the
person reading the test doesn't have to research
the side effects of the helper. (And I make it
very easy for new authors to follow the practice
going forward.)
This is also somewhat of a prep commit to avoid
the obfuscated use of refresh_from_db.
The get_user function is poorly named, but I don't want to
sweep the entire codebase yet.
It's also nice to have a test wrapper for little experiments
like profiling tests or hunting down calls to refresh_from_db.
It's possible that we would also just change the new wrapper
to more directly call Django. The `get_user` function isn't
used in a ton of real-world places, so we might want the test
code to just bypass the cache.
I add a bunch of cute helper methods to make
the test a bit more readable.
And then I make sure to get clean objects,
which precludes the need for our callback
functions to refresh the user objects.
And finally I make sure that our validation
functions don't cause any round trips (assuming
we have fetched objects using a standard
Zulip helper, which example_user ensures.)
_default_manager is the same as objects on most of our models. But
when a model class is stored in a variable, the type system doesn’t
know which model the variable is referring to, so it can’t know that
objects even exists (Django doesn’t add it if the user added a custom
manager of a different name). django-stubs used to incorrectly assume
it exists unconditionally, but it no longer does.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
There was a bug in apply_event code where only a stream which
is not private is added to the "never_subscribed" data after
a stream creation event. Instead, it should be added to the
"never_subscribed" data irrespective of permission policy of
the stream as we already send stream creation events only to
those users who can access the stream. Due to the current
bug, private streams were not being added to "never_subscribed"
data in apply_event for admins as well. This commit fixes it
and also makes sure the "never_subscribed" list is sorted
which was not done before and was also a bug.
The bugs mentioned above were unnoticed as the tests did not
cover these cases and this commit also adds tests for those
cases.
**Background**
User groups are expected to comply with the DAG constraint for the
many-to-many inter-group membership. The check for this constraint has
to be performed recursively so that we can find all direct and indirect
subgroups of the user group to be added.
This kind of check is vulnerable to phantom reads which is possible at
the default read committed isolation level because we cannot guarantee
that the check is still valid when we are adding the subgroups to the
user group.
**Solution**
To avoid having another transaction concurrently update one of the
to-be-subgroup after the recursive check is done, and before the subgroup
is added, we use SELECT FOR UPDATE to lock the user group rows.
The lock needs to be acquired before a group membership change is about
to occur before any check has been conducted.
Suppose that we are adding subgroup B to supergroup A, the locking protocol
is specified as follows:
1. Acquire a lock for B and all its direct and indirect subgroups.
2. Acquire a lock for A.
For the removal of user groups, we acquire a lock for the user group to
be removed with all its direct and indirect subgroups. This is the special
case A=B, which is still complaint with the protocol.
**Error handling**
We currently rely on Postgres' deadlock detection to abort transactions
and show an error for the users. In the future, we might need some
recovery mechanism or at least better error handling.
**Notes**
An important note is that we need to reuse the recursive CTE query that
finds the direct and indirect subgroups when applying the lock on the
rows. And the lock needs to be acquired the same way for the addition and
removal of direct subgroups.
User membership change (as opposed to user group membership) is not
affected. Read-only queries aren't either. The locks only protect
critical regions where the user group dependency graph might violate
the DAG constraint, where users are not participating.
**Testing**
We implement a transaction test case targeting some typical scenarios
when an internal server error is expected to happen (this means that the
user group view makes the correct decision to abort the transaction when
something goes wrong with locks).
To achieve this, we add a development view intended only for unit tests.
It has a global BARRIER that can be shared across threads, so that we
can synchronize them to consistently reproduce certain potential race
conditions prevented by the database locks.
The transaction test case lanuches pairs of threads initiating possibly
conflicting requests at the same time. The tests are set up such that exactly N
of them are expected to succeed with a certain error message (while we don't
know each one).
**Security notes**
get_recursive_subgroups_for_groups will no longer fetch user groups from
other realms. As a result, trying to add/remove a subgroup from another
realm results in a UserGroup not found error response.
We also implement subgroup-specific checks in has_user_group_access to
keep permission managing in a single place. Do note that the API
currently don't have a way to violate that check because we are only
checking the realm ID now.
We have historically cached two types of values
on a per-request basis inside of memory:
* linkifiers
* display recipients
Both of these caches were hand-written, and they
both actually cache values that are also in memcached,
so the per-request cache essentially only saves us
from a few memcached hits.
I think the linkifier per-request cache is a necessary
evil. It's an important part of message rendering, and
it's not super easy to structure the code to just get
a single value up front and pass it down the stack.
I'm not so sure we even need the display recipient
per-request cache any more, as we are generally pretty
smart now about hydrating recipient data in terms of
how the code is organized. But I haven't done thorough
research on that hypotheseis.
Fortunately, it's not rocket science to just write
a glorified memoize decorator and tie it into key
places in the code:
* middleware
* tests (e.g. asserting db counts)
* queue processors
That's what I did in this commit.
This commit definitely reduces the amount of code
to maintain. I think it also gets us closer to
possibly phasing out this whole technique, but that
effort is beyond the scope of this PR. We could
add some instrumentation to the decorator to see
how often we get a non-trivial number of saved
round trips to memcached.
Note that when we flush linkifiers, we just use
a big hammer and flush the entire per-request
cache for linkifiers, since there is only ever
one realm in the cache.
There's no need for the complexity and extra round
trips to call get_display_recipient in a testing
context.
We also eliminate the unnecessary call to check_string.
This function is poorly named, but that's a sweep
for another day.
Mateusz Mandera
Sahil Batra
Anders Kaseorg
Aman Agrawal
Alex Vandiver
Prakhar Pratyush
Tim Abbott
Hemant Umre
Steve Howell
Zixuan James Li