Several of our queues are capable of doing work that includes
rendering markdown (outgoing_webhook, embedded_bots, embed_links, and
email_mirror). As a result, it's essential that these don't cache
per-request data (specifically, realm filters) longer than they
should, making editing/deleting linkifiers potentially use old
settings until the relevant process was restarted.
Flushing these caches is extremely cheap (just clearing two
dictionaries) and thus is reasonable to do after every queue event,
rather than trying to do it only the ~1/3 of queues that specifically
do markdown processing. We do the same in our middleware for
reset_queries.
It's not worth writing a test for this because it's very difficult to
create the test setup situation for this bug with a single test worker
process; one needs to edit the linkifier configuration in a different
process than the one sending the message in order to see the bug.
This was a much larger visible bug on Zulip 2.1.x, where the presence
of the message_sender queue meant that this would apply to messages
sent via a browser.
Fixes#14095.
Previously, the input:
====================
- One
- Two
Two continued
====================
Would produce the same output as:
====================
- One
- Two
```
Two continued
```
====================
This was because our CodeBlockProcessor had a higher priority than
the ListIndentProcessor. This issue was discussed here:
https://chat.zulip.org/#narrow/stream/9-issues/topic/continuation.20paragraphs.20in.20list.20items.
/delete_topic endpoint could be used to request the deletion of a topic,
that would cause do_delete_messages to be called with an empty set in
these cases:
1. Requesting deletion of an empty stream.
2. Requesting deletion of a topic in a private stream with history not
public to subscribers, if the requesting admin doesn't have access to
any of the messages in that topic.
This function slims down the data that we get
from the database in order to create the
streams part of our client payload.
We also fix a typo.
We also clearly distinguish between queries
and lists here.
This new method prevents us from getting fat
objects from the database.
Instead, now we just get ids from the database
to build our subqueries.
Note that we could also technically eliminate
the `set(...)` wrappers in this code to have
Django make a subquery and save a round trip.
I am postponing that for another commit (since
it's still somewhat coupled to some other
complexity in `do_get_streams` that I am trying
to cut through, plus it's not the main point
of this commit.)
BEFORE:
# old, still in use for other codepaths
def get_stream_subscriptions_for_user(user_profile: UserProfile) -> QuerySet:
# TODO: Change return type to QuerySet[Subscription]
return Subscription.objects.filter(
user_profile=user_profile,
recipient__type=Recipient.STREAM,
)
user_subs = get_stream_subscriptions_for_user(user_profile).filter(
active=True,
).select_related('recipient')
recipient_check = Q(id__in=[sub.recipient.type_id for sub in user_subs])
AFTER:
# newly added
def get_subscribed_stream_ids_for_user(user_profile: UserProfile) -> QuerySet:
return Subscription.objects.filter(
user_profile_id=user_profile,
recipient__type=Recipient.STREAM,
active=True,
).values_list('recipient__type_id', flat=True)
subscribed_stream_ids = get_subscribed_stream_ids_for_user(user_profile)
recipient_check = Q(id__in=set(subscribed_stream_ids))
We calculate `max_message_id` for the mobile client.
Our query now no longer joins to the Message table
and just grabs one value instead of fat objects.
We were only checking error handling before, not
the happy path. The structure of the code
made it so that we effectively tested most of the
logic for this use case (since all the other flags
are sort of just filters on top of this), but
obviously we want explicit coverage here. Also,
we weren't testing the is-admin-but-not-api-super-user
error checking until this commit.
For historical reasons we were creating Recipient
objects at some point in the typing-notifications
codepath. Now we just work with UserProfiles.
This removes some queries, as indicated by
the change to `len(queries)` in a couple of the
tests.
The one subtle thing that changes here is huddles.
If user 10 sends a typing notification that they
are talking to users 20 and 30, there might not
actually be a huddle for users 10/20/30, but
we were actually creating huddles on the fly!
There is no need to create huddles just for
typing notifications, since we don't even
share huddle ids with our clients. The clients
just infer the huddles.
Some of the code that gets killed off here as
somewhat "collateral damage" is some
defensive code related to formerly supporting streams
in typing indicators. The support for streams
was killed off almost as soon as we released
the feature, and the codepath is pretty clearly
user-centric at this point.
I actually like this pattern:
def check_send_typing_notification(...):
typing_notification = check_typing_notification(...)
do_send_typing_notification(...)
It can help divide responsibilities nicely and make it easy
to write detailed unit tests against each of the two helpers.
Unfortunately, the good things didn't really happen here, and
instead we got the worst aspects of the pattern:
- The responsibilities for validation leaked into
the second function.
- Both functions were doing sane things individually
that became not-so-sane in the big picture (namely,
we ended up making Recipient objects for no reason,
but if you read each of the helpers, it was just one
step that seemed reasonable).
- Passing around dictionaries for results can be annoying.
Also, the pattern made a lot more sense when the validation
for typing was a lot more complicated. My prior commit makes
it so that we only ever deal with a list of user_ids.
Anyway, now I'm inlining it. :)
Subsequent commits will clean up the more substantive issue
here, which is that we are building Recipients for no reason.
The only clients that should use the typing
indicators endpoint are our internal clients,
and they should send a JSON-formatted list
of user_ids.
Unfortunately, we still have some older versions
of mobile that still send emails.
In this commit we fix non-user-facing things
like docs and tests to promote the user_ids
interface that has existed since about version
2.0 of the server.
One annoyance is that we documented the
typing endpoint with emails, instead of the
more modern user_ids, which may have delayed
mobile converting to user_ids (and which
certainly caused confusion). It's trivial
to update the docs, but we need to short
circuit one assertion in the openapi tests.
We also clean up the test structure for the
typing tests:
TypingHappyPathTest.test_start_to_another_user
TypingHappyPathTest.test_start_to_multiple_recipients
TypingHappyPathTest.test_start_to_self
TypingHappyPathTest.test_start_to_single_recipient
TypingHappyPathTest.test_stop_to_another_user
TypingHappyPathTest.test_stop_to_self
TypingValidateOperatorTest.test_invalid_parameter
TypingValidateOperatorTest.test_missing_parameter
TypingValidateUsersTest.test_argument_to_is_not_valid_json
TypingValidateUsersTest.test_bogus_user_id
TypingValidateUsersTest.test_empty_array
TypingValidateUsersTest.test_missing_recipient
TypingValidationHelpersTest.test_recipient_for_user_ids
TypingValidationHelpersTest.test_recipient_for_user_ids_non_existent_id
TypingLegacyMobileSupportTest.test_legacy_email_interface
Users who are using ZulipDesktop or haven't managed to auto-update to
ZulipElectron should be strongly encouraged to upgrade.
We'll likely want to move to something even stricter that blocks
loading the app at all, but this is a good start.
Before the Django 2.x upgrade, the DatabaseCreation
argument took an integer value. To deal with running
mulitple test instances, we created a random start
range that could count up 100 workers until the next
random id. Arbitrarily limiting the number of workers
to 100.
Post upgrade, we can now use string values. Enabling
the database + worker numbers to be more readable, as
well as removing the cap on the worker count.
This field wasn't accessed by any clients and was a less robust
version of the user_id field. Any client hoping to be interested in
who did message edits should be able to handle working with user IDs
rather than email addresses.
This is preparation for supporting moving messages between streams in
some cases.
It doesn't actually have any functional effect, since flush_message
clears the message unconditionally anyway.
We should not need so many queries here,
although a couple of the queries are just
standard things that apply to all requests.
I will reduce the number of queries in a
later commit.
This is mostly refactoring, but we also prevent a new
type of value error (list of non-int-or-string). The
new test code helps enforce that.
Cleanup includes:
- Use early-exit for email case.
- Rename helpers to get_validate_*.
- Avoid clumsy rebuilding of lists in helpers.
- Avoid the confusing `recipient` name (which
can be confused with the model by the same
name).
- Just delegate duplicate-id/email-removal to
the helpers.
The cleaner structure allows us to elminate a couple
mypy workarounds.
Credits to @xpac1985 for reporting, debugging and proposing fix to the
issue. The proposed fix was modified slightly by @hackerkid to set the
correct value for max_invites and upload_quota_gb. Tests added by
@hackerkid.
Fixes#13974
This gives them cache-compatible URLs, and also avoids some extra
copies of the sprite sheet images.
Comments on the Octopus emoji added by tabbott.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This fixes a confusing aspect of how our automated tests worked
previously, where we'd almost all HTTP requests in the unlikely
configuration with no User-Agent string specified.
We need to adjust query counts in a few tests that now are a bit
cheaper because they now can take advantage of a Client object created
in server_initialization.py in `process_client`.
So far the conversion was in a very random place -
register_remote_user(). All other codepaths that use
login_or_register_remote_user() call it with the user's email address.
Making remote_user_sso convert remote_username to the email address
before calling login_or_register_remote_user makes this usage consistent
across the board.
finish_desktop_flow is called with the assumption that the request
successfully proved control over the user_profile and generates a
special link to log into the user_profile account. There's no reason to
pass the realm param, as user_profile.realm can be assumed.
In `auth.py` there are three `if` blocks for different backends
to redirect to config error page with similar code. It is better
handled with common code using `get_attr()` function on
constructed setting names.
There was some duplicated code to test config error pages for
different auths which could be handled with less duplicated code
by adding those functions to `SocialAuthBase`.
Also moving the other tests makes it easier to access tests related
to a backend auth when they are in the same file.
Original idea was that KeyError was only going to happen there in case
of user passing bad input params to the endpoint, so logging a generic
message seemed sufficient. But this can also happen in case of
misconfiguration, so it's worth logging more info as it may help in
debugging the configuration.
Create a new page for desktop auth flow, in which
users can select one from going to the app or
continue the flow in the browser.
Co-authored-by: Mateusz Mandera <mateusz.mandera@protonmail.com>
To avoid some hidden bugs in tests caused by every ldap user having the
same password, we give each user a different password, generated based
on their uids (to avoid some ugly hard-coding in a bunch of places).
This avoids using `.save()` directly for editing stream properties,
and also uses the API in _send_and_verify_message to avoid confusing
logic around which user is doing what request.
Fixes part of #13823
This adds update_user to python_examples.py in zerver/openapi.
This also adds update-user.md to templates/zerver/api and adds
update-user to rest-endpoints.md (templates/zerver/help/include).
This adds deactivate_user to python_examples.py in zerver/openapi.
This also adds delete-user.md to templates/zerver/api and adds
delete-user to rest-endpoints.md (templates/zerver/help/include).
This adds get_single_user to python_examples.py in zerver/openapi.
This also adds get-single-user.md to templates/zerver/api and adds
get-single-user to rest-endpoints.md (templates/zerver/help/include).
This adds the OpenAPI format data for /users/{user_id} endpoint
and also removes 'users/{user_id}' from 'pending_endpoints' in
zerver/tests/test_openapi.py .
‘req_var in request.GET’ was previously believed to be slow from
profiling results. However, the real explanation for those profiling
results is that WSGIRequest.GET is a lazy cached property, so there’s
no reason to avoid it if we’re accessing request.GET anyway.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
In https://github.com/zulip/zulip/pull/12823 some changes to the realms
structure have been made, so now both in production and development
cross-realm bots live in the realm with string_id "zulipinternal".
There was a TODO in retention code to eliminate a conditional in a query
that became redundant with this change, and also the zulipinternal realm
should be omitted from the archiving process in archive_messages().
Profiling suggests this saves about 600us in the runtime of every GET
/events request attempting to resolve URLs to determine whether we
need to do the APPEND_SLASH behavior.
It's possible that we end up doing the same URL resolution work later
and we're just moving around some runtime, but I think even if we do,
Django probably doesn't do any fancy caching that would mean doing
this query twice doesn't just do twice the work.
In any case, we probably want to extend this behavior to our whole API
because the APPEND_SLASH redirect behavior is essentially a bug there.
That is a more involved refactor, however.
We use this single regular expression for processing essentially every
request, so it's definitely worth hinting to Python that we're going
to do so by compiling it. Saves about 40us per request.
A sloppy implementation of the main has_request_variables wrapper
function meant that it did two very inefficient things:
* To combine together the GET and POST parameters, it would make a
copy of the request.GET QueryDict object, which combined with the
fact that these objects are slow to access, consumed about 90us per
argument.
* Doing this in a loop (one time per argument), rather than once,
which resulted in us doing this 11 times for a `GET /events` query.
Fixing this to just make a dictionary and combine things with some
small loops saved about 1 millisecond from the total runtime of GET
/events (for comparison, the total actual work of that view function
is about 700ms).
We need to fix at least one test that used a bad mock HttpRequest
object that didn't have a .GET property.
The comment explains this issue, but effectively, the upgrade to
Django 2.x means that Django's built-in django.request logger was
writing to our errors logs WARNING-level data for every 404 and 400
error. We don't consider user errors to be a problem worth
highlighting in that log file.
Django 2.2.x is the next LTS release after Django 1.11.x; I expect
we'll be on it for a while, as Django 3.x won't have an LTS release
series out for a while.
Because of upstream API changes in Django, this commit includes
several changes beyond requirements and:
* urls: django.urls.resolvers.RegexURLPattern has been replaced by
django.urls.resolvers.URLPattern; affects OpenAPI code and related
features which re-parse Django's internals.
https://code.djangoproject.com/ticket/28593
* test_runner: Change number to suffix. Django changed the name in this
ticket: https://code.djangoproject.com/ticket/28578
* Delete now-unnecessary SameSite cookie code (it's now the default).
* forms: urlsafe_base64_encode returns string in Django 2.2.
https://docs.djangoproject.com/en/2.2/ref/utils/#django.utils.http.urlsafe_base64_encode
* upload: Django's File.size property replaces _get_size().
https://docs.djangoproject.com/en/2.2/_modules/django/core/files/base/
* process_queue: Migrate to new autoreload API.
* test_messages: Add an extra query caused by .refresh_from_db() losing
the .select_related() on the Realm object.
* session: Sync SessionHostDomainMiddleware with Django 2.2.
There's a lot more we can do to take advantage of the new release;
this is tracked in #11341.
Many changes by Tim Abbott, Umair Waheed, and Mateusz Mandera squashed
are squashed into this commit.
Fixes#10835.
Since essentially the first use of Tornado in Zulip, we've been
maintaining our Tornado+Django system, AsyncDjangoHandler, with
several hundred lines of Django code copied into it.
The goal for that code was simple: We wanted a way to use our Django
middleware (for code sharing reasons) inside a Tornado process (since
we wanted to use Tornado for our async events system).
As part of the Django 2.2.x upgrade, I looked at upgrading this
implementation to be based off modern Django, and it's definitely
possible to do that:
* Continue forking load_middleware to save response middleware.
* Continue manually running the Django response middleware.
* Continue working out a hack involving copying all of _get_response
to change a couple lines allowing us our Tornado code to not
actually return the Django HttpResponse so we can long-poll. The
previous hack of returning None stopped being viable with the Django 2.2
MiddlewareMixin.__call__ implementation.
But I decided to take this opportunity to look at trying to avoid
copying material Django code, and there is a way to do it:
* Replace RespondAsynchronously with a response.asynchronous attribute
on the HttpResponse; this allows Django to run its normal plumbing
happily in a way that should be stable over time, and then we
proceed to discard the response inside the Tornado `get()` method to
implement long-polling. (Better yet might be raising an
exception?). This lets us eliminate maintaining a patched copy of
_get_response.
* Removing the @asynchronous decorator, which didn't add anything now
that we only have one API endpoint backend (with two frontend call
points) that could call into this. Combined with the last bullet,
this lets us remove a significant hack from our
never_cache_responses function.
* Calling the normal Django `get_response` method from zulip_finish
after creating a duplicate request to process, rather than writing
totally custom code to do that. This lets us eliminate maintaining
a patched copy of Django's load_middleware.
* Adding detailed comments explaining how this is supposed to work,
what problems we encounter, and how we solve various problems, which
is critical to being able to modify this code in the future.
A key advantage of these changes is that the exact same code should
work on Django 1.11, Django 2.2, and Django 3.x, because we're no
longer copying large blocks of core Django code and thus should be
much less vulnerable to refactors.
There may be a modest performance downside, in that we now run both
request and response middleware twice when longpolling (once for the
request we discard). We may be able to avoid the expensive part of
it, Zulip's own request/response middleware, with a bit of additional
custom code to save work for requests where we're planning to discard
the response. Profiling will be important to understanding what's
worth doing here.
Hellosign now posts their callback as form/multipart, which Django only
permits to be read once. Attempts to access request.body after the
initial read throw "django.http.request.RawPostDataException: You
cannot access body after reading from request's data stream".
Fixes#13847.
This makes it possible to create a Zulip account from the mobile or
desktop apps and have the end result be that the user is logged in on
their mobile device.
We may need small changes in the desktop and/or mobile apps to support
this.
Closes#10859.
1) Created a new class `DatabaseType` and access its objects inside
`template_database_status()` instead of sending five arguments with
default values.
2) Made `check_files` and `setting_name` local variables instead of
function parameters since they had same value(None) for every call.
Fixes#13845.
webpack optimizes JSON modules using JSON.parse("{…}"), which is
faster than the normal JavaScript parser.
Update the backend to use emoji_codes.json too instead of the three
separate JSON files.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This code was very useful when first implemented to help catch errors
where our backend templates didn't render, but has been superceded by
the success of our URL coverage testing (which ensures every URL
supported by Zulip's urls.py is accessed by our tests, with a few
exceptions) and other tests covering all of the emails Zulip sends.
It has a significant maintenance cost because it's a bit hacky and
involves generating fake context, so it makes sense to remove these.
Any future coverage issues with templates should be addressed with a
direct test that just accessing the relevant URL or sends the relevant
email.
In python 3.5-3.6, generic types had an __origin__ attribute which
indicated which generic they originated from; the code was reflecting on
that value to check types against the openapi spec. In python3.7, this
changed, and there's no longer an immediately simple way to get this
information in all cases. __origin__ appears to be the implementing
class now, returning `list` or `collections.abc.Iterator` rather than
`typing.List` and `typing.Iterator`. This adds a sloppy-but-effective
mechanism for inferring if a type maps to the List/Dict/Iterator/Mapping
types and gets the test suite passing again.
We now use realm_id for querying UserPresence
instead of building a big WHERE clause from the
list of user_ids.
This commit may be a bit hard to measure, since
we still get the list of user_ids for the PushToken
query in the same method.
It adds this index:
"zerver_userpresence_realm_id_timestamp_25f410da_idx" btree (realm_id, "timestamp")
We expect this index to provide a major performance improvement when
fetching presence data for the whole realm from the database on
servers like zulipchat.com hosting several realms.
We now validate streams with a separate
function from PM recipients.
It's confusing enough all the ways you can
encode a stream or encode the PM recipients,
but trying to do it all in one function was
hard to reason about and led to at least one
bug.
In particular, there was a bug where streams
with commas in them would get split. Now
we just don't ever split on commas inside
of `extract_stream_indicator`.
Fixes#13836
After removing internal_send_message() in a recent
commit, we now have only two callers for
extract_recipients, and they are both related
to our REQ mechanism that always passes strings
to converters. (If there are default values,
REQ does not call the converters.)
We therefore make two changes:
- use the more strict annotation of "str"
for the `s` parameter
- don't bother with the isinstance check
Note that while the test mocks the actual message
send, we now have a `get_stream` call in the queue
worker, so we have to set up a real stream for
testing (or we could have mocked that as well, but
it didn't seem necessary). The setup queries add
to the amount of queries reported by the test,
plus the `get_stream` call. I just made the
query count a digits regex, which is a little bit
lame, but I don't think it's worth risking test
flakes for this.
This index is intended to optimize the performance of the very
frequently run query of "what is the presence status of all users in a
realm?".
Main changes:
- add realm_id to UserPresence
- add index for realm_id
- backfill realm_id for old rows
- change all writes to UserPresence to include
realm_id
The index is of this form:
"zerver_userpresence_realm_id_5c4ef5a9" btree (realm_id)
We will create an index on (realm_id, timestamp) in a
future commit, but I think it's a bit faster if you do
the backfill before the index.
There's also a minor tweak to the populate_db script.
This is just a refactoring to the more modern API
for sending internal messages.
To make this work we now plumb the email_gateway
flag through `internal_send_stream_message` instead
of `internal_send_message`.
We also change `send_zulip` to have its callers
pass in a full UserProfile object (which one of
them already had).
We prefer this to internal_send_message().
We are trying to deprecate `internal_send_message`,
which has extra moving parts related to
`extract_recipients` and `Addressee.legacy_build`.
There are two chunks of code that I touch here
that look pretty similar, but I'm not quite
sure they're worth de-duplicating, since they
use different topics and different message
content.
Instead of having `notify_new_user` delegate
all the heavy lifting to `send_signup_message`,
we just rename `send_signup_message` to be
`notify_new_user` and remove the one-line
wrapper.
We remove a lot of obsolete complexity:
- `internal` was no longer ever set to True
by real code, so we kill it off as well
as well as killing off the internal_blurb code
and the now-obsolete test
- the `sender` parameter was actually an
email, not a UserProfile, but I think
that got past mypy due to the caller
passing in something from settings.py
- we were only passing in NOTIFICATION_BOT
for the sender, so we just hard code
that now
- we eliminate the verbose
`admin_realm_signup_notifications_stream`
parameter and just hard code it to
"signups"
- we weren't using the optional realm
parameter
There's also a long ugly comment in
`get_recipient_info` related to this code
that I amended for now.
We should try to take action in a subsequent
commit.
This avoids an unnecessary join to UserProfile.
To verify this, you can do `print(queries)` in the
`test_get_custom_profile_fields_from_api` test. It's
kinda noisy, so I excerpted them below...
Before:
SELECT ...
FROM "zerver_customprofilefieldvalue"
INNER JOIN "zerver_userprofile" ON ("zerver_customprofilefieldvalue"."user_profile_id" = "zerver_userprofile"."id")
INNER JOIN "zerver_customprofilefield" ON ("zerver_customprofilefieldvalue"."field_id" = "zerver_customprofilefield"."id")
WHERE "zerver_userprofile"."realm_id" = 2
After:
SELECT ...
FROM "zerver_customprofilefieldvalue"
INNER JOIN "zerver_customprofilefield" ON ("zerver_customprofilefieldvalue"."field_id" = "zerver_customprofilefield"."id")
WHERE "zerver_customprofilefield"."realm_id" = 2'
I don't have any way to measure the two queries with
realistic data, but I would assume the second
query is significantly faster on most of our instances,
since CustomProfileField should be tiny.
I am trying to optimize a query in this endpoint.
I don't think I'll actually reduce the number of
queries, but I wanted to capture the query and
this was the easiest way to do it, so might as
well check in the code! :)
The line removed here is a noop, as both sides of the
immediately following conditional reassign the
same variable.
This harmless cruft was the result of the recent commit
1ae5964ab8, which added
support for single-user GETs.
This fixes a bug where our asynchronous requests were only copying the
Content-Type header (i.e. the one case where we're noticed) from the
Django HttpResponse. I'm not sure what the impact of this would be;
the rate-limiting headers rarely come up when breaking a long-polled
request. But it seems clearly an improvement to do this in a
consistent fashion.
Only the headers piece is a change; in Tornado
self.finish(x)
is equivalent to:
self.write(x)
self.finish()
Apparently, the arguments passed to template_database_status were
incorrect for the manual testing development database, in that we
didn't pass a status_dir when calling into that code from provision.
The result was that provisioning before running `test-backend` would
ignore changes to the list of check_files (etc.) made after rebasing,
and vice versa.
The cleanest fix is to compute status_dir from other values passed in;
I'm also going to open a follow-up issue for creating a better overall
interface here.
This adds a new API endpoint for querying basic data on a single other
user in the organization, reusing the existing infrastructure (and
view function!) for getting data on all users in an organization.
Fixes#12277.
This code is a bit flatter and just preps the data
for a single user. There is never any interaction
between the data for user A and user B, so we can
mostly avoid complicated nested data structures
and do most of the data-crunching on a per-user basis.
We also do an explicit sort of the data before
running it through groupby. The explicit sort
simplifies how we calculate `most_recent_info`
and also avoids needing to add `dt` to an intermediate
data structure.
Finally, when it comes to the individual client data,
the code has relied on the assumption that there is
only one row per client, which I believe to be true,
but now the code is more explicit about that.
The word "status" is vague, and this isn't
actually returning a list, so we now name it
get_presence_response.
I originally was gonna rename this to
get_presence_dict, but there's a function
called get_status_dict that returns a subset
of the response, so I think it's a bit more
clear that this is the bigger dict that
actually gets sent back.
We want to err on the side of server_timestamp being
old, since we may eventually use this to make responses
just include incremental changes, and we don't want a
time window (however small) when we miss presence rows.
The clients will be able to deal with duplicate data
to the extent that the time windows are overlapping.
Also, extracting the other local var here
(for `presences`) will set up a subsequent commit
where we re-format the data for clients with
slim_presence=True.
In e3ad9baf1d, we introduced yet another
bug where we incorrectly shared event dictionaries between multiple
queues.
Fortunately, the logging that reports on "event was not in the queue"
issues worked and detected this on chat.zulip.org, but this is a clear
indication that the comments we have around this system were not
sufficient to produce correct behavior.
We fix this by changing event_queue.push, the code that mutates the
event dictionaries, to do the shallow copies itself. The only
downside here is process_message_event, a relatively low-traffic code
path, does an extra per-queue dictionary copy. Given that presence,
heartbeat, and message reading events are likely more traffic and
dealing with HTTP is likely much more expensive than a dictionary
copy, this probably doesn't matter performance-wise.
(And if profiling later finds it is, there are potential workarounds
like passing a skip_copy argument we can do).
django-phonenumber-field 2.4.0 adds tighter phone number validation
that rejects +12223334444 for having an invalid area code. This was
reverted in 4.0.0, but django-two-factor-auth still requires <3.99.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This commit includes a new `stream_post_policy` setting,
by replacing the `is_announcement_only` field from the Stream model,
which is done by mirroring the structure of the existing
`create_stream_policy`.
It includes the necessary schema and database migrations to migrate
the is_announcement_only boolean field to stream_post_policy,
a smallPositiveInteger field similar to many other settings.
This change is done to allow organization administrators to restrict
new members from creating and posting to a stream. However, this does
not affect admins who are new members.
With many tweaks by tabbott to documentation under /help, etc.
Fixes#13616.
This flag affects page_params and the
payload you get back from POSTs to this
url:
users/me/presence
The flag does not yet affect the
presence events that get sent to a
client.
This should ensure that folks rebasing past this commit from an older
database model get their database rebuilt in the way that will
match the test_subs.py query count of 40.
Add a simple compatibility function for AWX 9.x.x. Before AWX 9.x.x
a "friendly_name" key was sent by default. Afterwards it was removed
from being a default key but we can still more or less determine if
the triggering event was a job from the REST-style URL.
Note: It is also technically possible to add the key back by defining
a custom notification template in AWX/Tower.
Resolves#13295.
This applies rate limiting (through a decorator) of authenticate()
functions in the Email and LDAP backends - because those are the ones
where we check user's password.
The limiting is based on the username that the authentication is
attempted for - more than X attempts in Y minutes to a username is not
permitted.
If the limit is exceeded, RateLimited exception will be raised - this
can be either handled in a custom way by the code that calls
authenticate(), or it will be handled by RateLimitMiddleware and return
a json_error as the response.
We will want to raise RateLimited in authenticate() in rate limiting
code - Django's authenticate() mechanism catches PermissionDenied, which
we don't want for RateLimited. We want RateLimited to propagate to our
code that called the authenticate() function.
As more types of rate limiting of requests are added, one request may
end up having various limits applied to it - and the middleware needs to
be able to handle that. We implement that through a set_response_headers
function, which sets the X-RateLimit-* headers in a sensible way based
on all the limits that were applied to the request.
validate_otp_params needs to be moved to backends.py, because as of this
commit it'll be used both there and in views.auth - and import from
views.auth to backends.py causes circular import issue.
This makes get_raw_user_data, which was being imported indirectly
from zerver.lib.events inside zerver/views/users.py, get imported
from zerver.lib.users where it actually is.
While the result of this change doesn't completely do what we need, it
does remove a huge amount of duplicated lists of fields. With a bit
more similar work, we should be able to eliminate a broad category of
potential bugs involving Stream and Subscription objects being
represented inconsistently in the API.
Work towards #13787.
This has the side of effect of making new fields we add to Stream be
automatically included, which will help maintain this code as we
upgrade it.
This commit adds is_web_public, history_public_to_subscribers, and
email_notifications fields to the dictionary.
Tests require adjusting, because the class-based view has an additional
redirect - through /uid/set-password/ and the token is read from the
session. See Django code of PasswordResetConfirmView.
The `notification_settings_null` field of the `client_capabilities`
parameter is, apparently unintentionally, required.
This is mostly harmless. However, if any _future_ fields are made
required, all existing clients using this parameter will break, and it
will be needlessly difficult for new clients to specify new
capabilities in a backwards-compatible way.
Attempt to stave that possibility off with warnings.
(No functional changes.)
This modifies get_cross_realm_dicts in zerver.lib.users to call
format_user_row. This is done to remove current and prevent future
inconsistencies between in the dictionary formats for get_raw_user_data
and get_cross_realm_dicts.
Implementation substantially rewritten by tabbott.
Fixes#13638.
This moves get_cross_realm_dicts (from zerver.lib.actions),
get_raw_user_data and get_custom_profile_field_values (from
zerver.lib.events) to zerver.lib.users.
Only the getter of the is_new_member property is added,
to the UserProfile class. This is done to deduplicate
action of checking whether a user is a new member or not.
This extracts the user_data inner function from get_raw_user_data as a
reusable function. We intend to reuse it for cross-realm user dicts.
A few changes were made while extracting it:
* Renaming the UserProfile argument to acting_user, so we can do loops
over a local user_profile variable.
* Moved it to zerver.lib.users, as that's a more appropriate home for
this function formatting data on users.
* Simplified the calling convention for passing custom profile fields
to reflect the fact that this function processes a single user (and
is expected to be called in a loop).
"Zulip Voyager" was a name invented during the Hack Week to open
source Zulip for what a single-system Zulip server might be called, as
a Star Trek pun on the code it was based on, "Zulip Enterprise".
At the time, we just needed a name quickly, but it was never a good
name, just a placeholder. This removes that placeholder name from
much of the codebase. A bit more work will be required to transition
the `zulip::voyager` Puppet class, as that has some migration work
involved.
These docstrings hadn't been properly updated in years, and bad an
awkward mix of a bad version of the user-facing documentation and
details that are no longer true (e.g. references to "Voyager").
(One important detail is that we have real documentation for this
system now).
Closes#13736.
zerver.lib.server_initialization.create_internal has precisely the same
code (you can copy-and-paste swap them, with one level of indentation
adjustment, without generating any diff) so they can be trivially
deduplicated.
zerver.lib.server_initialization.create_users has precisely the same
code (you can copy-and-paste swap them without generating any diff) so
they can be trivially deduplicated.
This doesn't change any behavior, the purpose of this is to make the
function identical to what we have in server_initialization.py so that
it can be deduplicated in follow-up commits.
In addition to making our schema check stricter, it also makes it
possible for us to extend check_events_dict to do additional
validation that's only expected for the full event object.
Now that we have the type situation of having anchor support passing a
string, this is a much more natural way to implement
use_first_unread_anchor.
We still support the old interface to avoid breaking compatibility
with legacy versions of the mobile apps.
A wart that has long been present inin Zulip's get_messages API is how
to request "the latest messages" in the API. Previously, the
recommendation was basically to pass anchor=10000000000000000 (for an
appropriately huge number). An accident of the server's implementation
meant that specific number of 0s was actually important to avoid a
buggy (or at least wasteful) value of found_newest=False if the query
had specified num_after=0 (since we didn't check).
This was the cause of the mobile issue
https://github.com/zulip/zulip-mobile/issues/3654.
The solution is to allow passing a special value of anchor='newest',
basically a special string-type value that the server can interpret as
meaning the user precisely just wants the most recent messages. We
also add an analogous anchor='oldest' or similar to avoid folks
needing to write a somewhat ugly anchor=0 for fetching the very first
messages.
We may want to also replace the use_first_unread_anchor argument to be
a "first_unread" value for the anchor parameter.
While it's not always ideal to make a value have a variable type like
this, in this case it seems like a really clean way to express the
idea of what the user is asking for in the API.
This fixes a bug where that clients using the legacy approach of a
"very large anchor" value with the intent to only get the most recent
messages would only get found_newest=True if they used the specific
value LARGER_THAN_MAX_MESSAGE_ID. Now any value at least that large
will work.
In upcoming commits, we plan to replace this with passing the string
"last", but it seems worth removing the buggy "special value" behavior
while we're touching this code.
This is required for the upcoming type behavior of the "anchor"
parameter.
This change is the minimal work required to have our OpenAPI code not
fail when checking a union-type value of this form. We'll likely want
to, in the future, do something nicer, but it'd require more extensive
infrastructure for parsing of OpenAPI data that it's worth with our
current approach (we may want to switch to using a library).
The proximal issue here is that in upcoming commits, we're going to
change the type of the `anchor` field in `get_messages_backend` to
support passing either an integer or a string.
Many of our tests using POSTRequestMock currently define a query
object that uses integer values for the integer fields we're going to
pass into it, e.g. {'num_after': 0}. That is the correct type for
that field in the Zulip API, before HTTP encoding turns it into a
string. However, because POSTRequestMock didn't use HTTP encoding at
all (which will convert the 0 into a '0'), it ended up passing an
integer to a function that can't possible receive one as an argument.
Ideally, we'd just get rid of POSTRequestMock, since it's a hack, and
just do real HTTP requests instead.
But since it's used in a lot of places making doing so somewhat
impractical, we can get past this issue by just making POSTRequestMock
convert integers to strings.
Using logging.info() rather than logger.info() meant that our
zulip.soft_deactivation logger configuration (which, in particular,
included not logging to the console) was not active on this log line,
resulting in the `manage.py soft_deactivate_users` cron job sending
emails every time it ran.
Fixes#13750.
In Django 2.0, request.user.is_authenticated stops supporting
`.is_authenticated()` and becomes just a property. In 1.11, it's a
CallableProperty (i.e. can be used either way), and we already use it
as a property in several other places, so we should just switch to
using it consistently now to get it off of our Django 2.x migration
checklist.
Previously, we didn't track opening and closing fences separately,
with led to bugs like not parsing a list that was immediately after
a quoted fence; we treated each ``` as a new fence.
This commit rewrites the function to maintain a stack of currently
open fences. If any of the parent fences is a code fence, we do not
insert a new line before a list.
We also add some test cases specifically to test this behavior with
complexly nested lists.
Fixes#13745.
Adding invited users to the notifications stream unconditionally isn't
a correct behaviour for guest users, where the previous behavior of
including the notifications stream no longer makes sense. Therefore,
while inviting a new user, the notifications stream is listed along
with other streams with a message "recieves notifications for new
streams" in order to distinguish it from other streams.
Fixes#13645.
The desktop otp flow (to be added in next commits) will want to generate
one-time tokens for the app that will allow it to obtain an
authenticated session. log_into_subdomain will be the endpoint to pass
the one-time token to. Currently it uses signed data as its input
"tokens", which is not compatible with the otp flow, which requires
simpler (and fixed-length) token. Thus the correct scheme to use is to
store the authenticated data in redis and return a token tied to the
data, which should be passed to the log_into_subdomain endpoint.
In this commit, we replace the "pass signed data around" scheme with the
redis scheme, because there's no point having both.
This extracts a function for computing show_invites and
show_add_streams, for better readability and testability.
This commit was substantially cleaned up by tabbott.
This legacy cross-realm bot hasn't been used in several years, as far
as I know. If we wanted to re-introduce it, I'd want to implement it
as an embedded bot using those common APIs, rather than the totally
custom hacky code used for it that involves unnecessary queue workers
and similar details.
Fixes#13533.
authenticate_remote_user already takes care of calling the authenticate
with the dummy backend. Also, return_data is not used and catching
DoesNotExist exception is not needed, as the dummy backend just returns
None if user isn't found.
In other places where we set request._email, we set it to the
delivery_email, as that's more informative in orgs with hidden email
settings, where user.email will be useless.
This updates our error handling of invalid Slack API tokens (and other
networking error handling) to mostly make sense:
* A token that doesn't start with `xoxp-` gives an extended error early.
* An AssertionError for the codebase is correctly declared as such.
* We check for token shape errors before querying the Slack API.
We could still do useful work to raise custom exception classes here.
Thanks to @stavrospat for raising this issue.
responses is an module analogous to httpretty for mocking external
URLs, with a very similar interface (potentially cleaner in that it
makes use of context managers).
The most important (in the moment) problem with httpretty is that it
breaks the ability to use redis in parts of code where httpretty is
enabled. From more research, the module in general has tendency to
have various troublesome bugs with breaking URLs that it shouldn't be
affecting, caused by it working at the socket interface layer. While
those issues could be fixed, responses seems to be less buggy (based
on both third-party reports like ckan/ckan#4755 and our own experience
in removing workarounds for bugs in httpretty) and is more actively
maintained.
If an email is sent with the .prefer-html option, but it has no html
body, it's better to fall back to plaintext content instead of treating
it as a user error.
Closes#13484.
These options tell zulip whether to prefer the plaintext or html version
of the email message. prefer-text is the default behavior, so including
the option doesn't change anything as of now, but we're adding it to
prepare to potentially change the default behavior in the future.
As we add more address options, which will have different behavior than
simply setting option_name=True, we need to migrate this subsystem to
something that better supports more complex logic and will allow
encapsulating it, instead of needing to be put all over the
decode_email_address function.
This essentially unused legacy variable was causing Zulip to query the
database at import time, which is generally not something we aim to
do.
Combined with the issue fixed in the previous commit, this variable
resulted in test-backend providing an unhelpful crash when provision
hadn't updated the unit testing database.
Since the intent of our testing code was clearly to clear this cache
for every test, there's no reason for it to be a module-level global.
This allows us to remove an unnecessary import from test_runner.py,
which in combination with DEFAULT_REALM's definition was causing us to
run models code before running migrations inside test-backend.
(That bug, in turn, caused test-backend's check for whether migrations
needs to be run to happen sadly after trying to access a Realm,
trigger a test-backend crash if the Realm model had changed since the
last provision).
Due to a known but unfixed bug in the Python standard library’s
urllib.parse module (CVE-2015-2104), a crafted URL could bypass the
validation in the previous patch and still achieve an open redirect.
https://bugs.python.org/issue23505
Switch to using django.utils.http.is_safe_url, which already contains
a workaround for this bug.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
It's theoretically possible to have configured a Zulip server where
the system bots live in the same realm as normal users (and may have
in fact been the default in early Zulip releases? Unclear.). We
should handle these without the migration intended to clean up naming
for the system bot realm crashing.
Fixes#13660.
Zulip has had a small use of WebSockets (specifically, for the code
path of sending messages, via the webapp only) since ~2013. We
originally added this use of WebSockets in the hope that the latency
benefits of doing so would allow us to avoid implementing a markdown
local echo; they were not. Further, HTTP/2 may have eliminated the
latency difference we hoped to exploit by using WebSockets in any
case.
While we’d originally imagined using WebSockets for other endpoints,
there was never a good justification for moving more components to the
WebSockets system.
This WebSockets code path had a lot of downsides/complexity,
including:
* The messy hack involving constructing an emulated request object to
hook into doing Django requests.
* The `message_senders` queue processor system, which increases RAM
needs and must be provisioned independently from the rest of the
server).
* A duplicate check_send_receive_time Nagios test specific to
WebSockets.
* The requirement for users to have their firewalls/NATs allow
WebSocket connections, and a setting to disable them for networks
where WebSockets don’t work.
* Dependencies on the SockJS family of libraries, which has at times
been poorly maintained, and periodically throws random JavaScript
exceptions in our production environments without a deep enough
traceback to effectively investigate.
* A total of about 1600 lines of our code related to the feature.
* Increased load on the Tornado system, especially around a Zulip
server restart, and especially for large installations like
zulipchat.com, resulting in extra delay before messages can be sent
again.
As detailed in
https://github.com/zulip/zulip/pull/12862#issuecomment-536152397, it
appears that removing WebSockets moderately increases the time it
takes for the `send_message` API query to return from the server, but
does not significantly change the time between when a message is sent
and when it is received by clients. We don’t understand the reason
for that change (suggesting the possibility of a measurement error),
and even if it is a real change, we consider that potential small
latency regression to be acceptable.
If we later want WebSockets, we’ll likely want to just use Django
Channels.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
orientor
Mateusz Mandera
Tim Abbott
Rohitt Vashishtha
Steve Howell
Chris Bobbe
Wyatt Hoodes
Vishnu KS
Anders Kaseorg
Dinesh
harshavardhanpb
vsvipul
Ryan Rehman
akashaviator
Chris Heald
rht
Hashir Sarwar
Hemanth V. Alluri
Ray Kraesig
Greg Price
Abhishek-Balaji
Jonathan Cobb