Previously, get_recent_private_messages could take 100ms-1s to run,
contributing a substantial portion of the total runtime of `/`.
We fix this by taking advantage of the recent denormalization of
personal_recipient into the UserProfile model, allowing us to avoid
the complex join with Recipient that was previously required.
The change that requires additional commentary is the change to the
main, big SQL query:
1. We eliminate UserMessage table from the query, because the condition
m.recipient_id=%(my_recipient_id)d
implies m is a personal message to the user being processed - so joining
with usermessage to check for user_profile_id and flags&2048 (which
checks the message is private) is redundant.
2. We only need to join the Message table with UserProfile
(on sender_id) and get the sender's personal_recipient_id from their
UserProfile row.
Fixes#13437.
This is adds foreign keys to the corresponding Recipient object in the
UserProfile on Stream tables, a denormalization intended to improve
performance as this is a common query.
In the migration for setting the field correctly for existing users,
we do a direct SQL query (because Django 1.11 doesn't provide any good
method for doing it properly in bulk using the ORM.).
A consequence of this change to the model is that a bit of code needs
to be added to the functions responsible for creating new users (to
set the field after the Recipient object gets created). Fortunately,
there's only a few code paths for doing that.
Also an adjustment is needed in the import system - this introduces a
circular relation between Recipient and UserProfile. The field cannot be
set until the Recipient objects have been created, but UserProfiles need
to be created before their corresponding Recipients. We deal with this
by first importing UserProfiles same way as before, but we leave the
personal_recipient field uninitialized. After creating the Recipient
objects, we call a function to set the field for all the imported users
in bulk.
A similar change is made for managing Stream objects.
Fixes#13452.
The migration from UserProfile.is_realm_admin/UserProfile.is_guest in
e10361a832 broke our LDAP-based support
for setting a user's role via LDAP properties, which relied on setting
those fields. Because the django-auth-ldap feature powering that only
supports booleans (and in any case, we don't want to expose constants
like `ROLE_REALM_ADMINISTRATOR` to the LDAP configuration interface),
it makes sense to provide setters for these legacy fields for
backwards-compatibility.
We lint against using these setters directly in Zulip's codebase
directly. The issue with using these is that when changing user's
.role we want to create appropriate RealmAuditLog entries and send
events. This isn't possible when using these setters - the log entries
and events should be created if the role change in the UserProfile is
actually save()-ed to the database - and on the level of the setter
function, it's not known whether the change will indeed be saved.
It would have to be somehow figured out on the level of post_save
signal handlers, but it doesn't seem like a good design to have such
complexity there, for the sake of setters that generally shouldn't be
used anyway - because we prefer the do_change_is_* functions.
The purpose of this change is narrowly to handle use cases like the
setattr on these boolean properties.
We used to specify the securityScheme for each REST operation seperately.
This is unecessary as the securityScheme can be specified in root level
and would be automatically applied to all operations. This also prevents
us accidentally not specifying the securityScheme for some operations and
was the case for /users/me/subscriptions PATCH endpoint. The root level
securityScheme can be also overriden in the operational level when
necessary.
swagger.io/docs/specification/authentication/#security
We use the plumbing introduced in a previous commit, to now raise
PushNotificationBouncerRetryLaterError in send_to_push_bouncer in case
of issues with talking to the bouncer server. That's a better way of
dealing with the errors than the previous approach of returning a
"failed" boolean, which generally wasn't checked in the code anyway and
did nothing.
The PushNotificationBouncerRetryLaterError exception will be nicely
handled by queue processors to retry sending again, and due to being a
JsonableError, it will also communicate the error to API users.
We add PushNotificationBouncerRetryLaterError as an exception to signal
an error occurred when trying to communicate with the bouncer and it
should be retried. We use JsonableError as the base class, because this
signal will need to work in two roles:
1. When the push notification was being issued by the queue worker
PushNotificationsWorker, it will signal to the worker to requeue the
event and try again later.
2. The exception will also possibly be raised (this will be added in the
next commit) on codepaths coming from a request to an API endpoint (for
example to add a token, to users/me/apns_device_token). In that case,
it'll be needed to provide a good error to the API user - and basing
this exception on JsonableError will allow that.
If a message begins with /me, we do not have any cases where the
rendered content would not begin with `<p>/me`. Thus, we can safely
remove the redundant checks both on the backend and frontend.
It appears we forgot to make identical changes to the backend
in #11089 while adding support for multiline /me messages,
resulting in any messages that didn't end in a paragraph getting
rendered as a regular message instead.
Fixes#13454.
In configurations with LDAP_APPEND_DOMAIN, we don't want people creating
non-ldap accounts with emails matching the ldap domain.
So in the registration flow, if the email isn't found in LDAP, but
matches LDAP_APPEND_DOMAIN, we stop, rather than proceeding with account
creation. In case of emails not matching LDAP_APPEND_DOMAIN, we will
still continue to make a normal, non-ldap account.
The problem was that, for example, given a configuration of social
backend + LDAPPopulator, if a user that's not in ldap was being
registered, the Full Name field in the registration form would be
empty instead of getting prefilled with the name provided by the
social backend.
This fixes it - first we try to get the name from ldap. If that
succeeds, a form is created pre-filled with that name. Otherwise, we
proceed to attempt to pre-fill with other means.
This also has a nice side effect of reorganizing most of the logic to
be more parallel between LDAP and other sources of name data.
This is a performance optimization, since we can avoid doing work
related to wildcard mentions in the common case that the message can't
have any. We also add a unit test for adding wildcard mentions in a
message edit.
We also switch the underlying exctact_mention_text method to use
a regular for loop, as well as make the related methods return
tuples of (names, is_wildcard). This abstraction is hidden from the
MentionData callers behind mention_data.message_has_wildcards().
Concerns #13430.
Previously, the LDAP code for syncing user data was not
multiple-realm-aware, resulting in errors trying to sync data for an
LDAP user present in multiple realms.
Tweaked by tabbott to add some extended comments.
Fixes#11520.
For a long time, we've been only doing the zxcvbn password strength
checks on the browser, which is helpful, but means users could through
hackery (or a bug in the frontend validation code) manage to set a
too-weak password. We fix this by running our password strength
validation on the backend as well, using python-zxcvbn.
In theory, a bug in python-zxcvbn could result in it producing a
different opinion than the frontend version; if so, it'd be a pretty
bad bug in the library, and hopefully we'd hear about it from users,
report upstream, and get it fixed that way. Alternatively, we can
switch to shelling out to node like we do for KaTeX.
Fixes#6880.
A bug in Zulip's new user signup process meant that users who
registered their account using social authentication (e.g. GitHub or
Google SSO) in an organization that also allows password
authentication could have their personal API key stolen by an
unprivileged attacker, allowing nearly full access to the user's
account.
Zulip versions between 1.7.0 and 2.0.6 were affected.
This commit fixes the original bug and also contains a database
migration to fix any users with corrupt `password` fields in the
database as a result of the bug.
Out of an abundance of caution (and to protect the users of any
installations that delay applying this commit), the migration also
resets the API keys of any users where Zulip's logs cannot prove the
user's API key was not previously stolen via this bug. Resetting
those API keys will be inconvenient for users:
* Users of the Zulip mobile and terminal apps whose API keys are reset
will be logged out and need to login again.
* Users using their personal API keys for any other reason will need
to re-fetch their personal API key.
We discovered this bug internally and don't believe it was disclosed
prior to our publishing it through this commit. Because the algorithm
for determining which users might have been affected is very
conservative, many users who were never at risk will have their API
keys reset by this migration.
To avoid this on self-hosted installations that have always used
e.g. LDAP authentication, we skip resetting API keys on installations
that don't have password authentication enabled. System
administrators on installations that used to have email authentication
enabled, but no longer do, should temporarily enable EmailAuthBackend
before applying this migration.
The migration also records which users had their passwords or API keys
reset in the usual RealmAuditLog table.
This change makes it possible for users to control the notification
settings for wildcard mentions as a separate control from PMs and
direct @-mentions.
This includes adding a new endpoint to the push notification bouncer
interface, and code to call it appropriately after resetting a user's
personal API key.
When we add support for a user having multiple API keys, we may need
to add an additional key here to support removing keys associated with
just one client.
Since years ago, this field hasn't been used for anything other than
some logging that would be better off logging the user ID anyway.
It existed in the first place simply because we weren't passing the
user_profile_id to Tornado at all.
Then, find and fix a predictable number of previous misuses.
With a small change by tabbott to preserve backwards compatibility for
sending `yes` for the `forged` field.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
The original/legacy emoji reactions endpoints made use of HTTP PUT and
didn't have an API that could correctly handle situations where the
emoji names change over time. We stopped using the legacy endpoints
some time ago, so we can remove them now.
This requires straightforward updates to older tests that were still
written against the legacy API.
Fixes#12940.
The function only used the user's realm anyway, so this is a cleaner
API.
This should also make it more convenient to permanently delete
messages manually, since one doesn't have to fetch a random user in
the realm in order to delete a message using the management shell.
No functional change.
When creating realm with the ldap backend, the registration flow didn't
properly handle some things - the user wouldn't be set as realm admin,
initial subscriptions and messages weren't created, and the redirect
wasn't happening properly in the case of subdomains.
The state of the FAKELDAP setup for the dev env has fallen behind the
backend changes and updates to fakeldap (which implemented
SCOPE_ONELEVEL searches), as well as having some other minor issues.
This commit restore it to a working state and now all three config modes
work properly.
django_to_ldap_username is now able to find the correct ldap username in
every supported type of configuration, so we can remove these
conditionals and use django_to_ldap_username in a straight-forward
manner.
Having to account everywhere for both cases of having and not
having email search configured makes things needlessly complicated.
It's better to make the setting obligatory in configurations other than
LDAP_APPEND_DOMAIN.
Previously, we were using user_profile.email rather than
user_profile.delivery_email in all calculations involving Gravatar
URLs, which meant that all organizations with the new
EMAIL_ADDRESS_VISIBILITY_ADMINS setting enabled had useless gravatars
not based on the `user15@host.domain` type fake email addresses we
generate for the API to refer to users.
The fix is to convert these calculations to use the user's
delivery_email. Some refactoring is required to ensure the data is
passed through to the parts of the codebase that do the check;
fortunately, our automated tests of schemas are effective in verifying
that the new `sender_delivery_email` field isn't visible to the API.
Fixes#13369.
Previously, we weren't properly passing through the value of the
client_gravatar flag from the caller, resulting in buggy results if
the caller passed client_gravatar=False to do_test().
We happened to not have any uses of this before, but we're about to
add one.
As discussed in the comment, ideally these checks should be added
completely automatically, rather than needing to be manually added
every time we add a new setting. But hopefully the example code for
all of the similar enums that this provides will at least provide some
help.
By adding some additional plumbing (through PreregistrationUser) of the
full_name and an additional full_name_validated option, we
pre-populate the Full Name field in the registration form when coming
through a social backend (google/github/saml/etc.) and potentially skip
the registration form (if the user would have nothing to do there other
than clicking the Confirm button) and just create the account and log
the user in.
The main purpose of this is to make that name change happen in
/server_settings. external_authentication_methods is a much better, more
descriptive name than social_backends from API perspective.
These are returned through the API, at the /server_settings
endpoint. It's better to just return the list of dicts with a guarantee
of being sorted in the correct order, than to clutter things with the
sort_order field.
This legacy endpoint was designed for the original native Zulip mobile
apps, which were deprecated years ago in favor of the React Native
app.
It was replaced by /server_settings for active use years ago, so it's
safe to remove it now.
The url scheme is now /accounts/login/social/saml/{idp_name} to initiate
login using the IdP configured under "idp_name" name.
display_name and display_logo (the name and icon to show on the "Log in
with" button) can be customized by adding the apprioprate settings in
the configured IdP dictionaries.
This changes the way django_to_ldap_username works to make sure the ldap
username it returns actually has a corresponding ldap entry and raise an
exception if that's not possible. It seems to be a more sound approach
than just having it return its best guess - which was the case so far.
Now there is a guarantee that what it returns is the username of an
actual ldap user.
This allows communicating to the registration flow when the email being
registered doesn't belong to ldap, which then will proceed to register
it via the normal email backend flow - finally fixing the bug where you
couldn't register a non-ldap email even with the email backend enabled.
These changes to the behavior of django_to_ldap_username require small
refactorings in a couple of other functions that call it, as well as
adapting some tests to these changes. Finally, additional tests are
added for the above-mentioned registration flow behavior and some
related corner-cases.
Instead of mocking the _LDAPUser class, these tests can now take
advantage of the test directory that other ldap are using. After these
changes, test_query_email_attr also verifies that query_ldap can
successfully be used to query by user email, if email search is
configured.
Fixes#11878
Instead of a confusing mix of django_auth_backed applying
ldap_to_django_username in its internals for one part of the
translation, and then custom logic for grabbing it from the email
attribute of the ldapuser in ZulipLDAPAuthBackend.get_or_build_user
for the second part of the translation,
we put all the logic in a single function user_email_from_ldapuser
which will be used by get_or_build of both ZulipLDAPUserPopulator and
ZulipLDAPAuthBackend.
This, building on the previous commits with the email search feature,
fixes the ldap sync bug from issue #11878.
If we can get upstream django-auth-ldap to merge
https://github.com/django-auth-ldap/django-auth-ldap/pull/154, we'll
be able to go back to using the version of ldap_to_django_username
that accepts a _LDAPUser object.
With this, django_to_ldap_username can take an email and find the ldap
username of the ldap user who has this email - if email search is
configured.
This allows successful authenticate() with ldap email and ldap password,
instead of ldap username. This is especially useful because when
a user wants to fetch their api key, the server attempts authenticate
with user_profile.email - and this used to fail if the user was an ldap
user (because the ldap username was required to authenticate
succesfully). See issue #9277.
This fixes a collection of bugs surrounding LDAP configurations A and
C (i.e. LDAP_APPEND_DOMAIN=None) with EmailAuthBackend also enabled.
The core problem was that our desired security model in that setting
of requiring LDAP authentication for accounts managed by LDAP was not
implementable without a way to
Now admins can configure an LDAPSearch query that will find if there
are users in LDAP that have the email address and
email_belongs_to_ldap() will take advantage of that - no longer
returning True in response to all requests and thus blocking email
backend authentication.
In the documentation, we describe this as mandatory configuration for
users (and likely will make it so soon in the code) because the
failure modes for this not being configured are confusing.
But making that change is pending work to improve the relevant error
messages.
Fixes#11715.
The value of realm attribute in confirmation object used to be empty
before. We are not currently using the realm attribute of reactivation
links anywhere. The value of realm stored in content_object is currently
used.
We currently have code to calculate the value of realm_icon_url,
admin_emails and default_discount in two diffrent places. With
the addition of showing confirmation links it would become three.
The easiest way to deduplicate the code and make the view cleaner
is by doing the calculations in template. Alternatively one can
write a function that takes users, realms and confirmations as
arguments and sets the value of realm_icon_url, admin_emails and
default_discount appropriately in realm object according to the
type of the confirmation. But that seems more messy than passing
the functions directly to template approach.
Most of the failures were due to parameters that are not intended to
be used by third-party code, so the correct fix for those was the set
intentionally_undocumented=True.
Fixes#12969.
MigrationsTestCase is intentionally omitted from this, since migrations
tests are different in their nature and so whatever setUp()
ZulipTestCase may do in the future, MigrationsTestCase may not
necessarily want to replicate.
Previously, the logic for determining whether to provide an LDAP
password prompt on the registration page was incorrectly including it
if any LDAP authentication was backend enabled, even if LDAP was
configured with the populate-only backend that is not responsible for
authentication (just for filling in name and custom profile fields).
We fix this by correcting the conditional, and add a test.
There's still follow-up work to do here: We may still end up
presenting a registration form in situations where it's useless
because we got all the data from SAML + LDAP. But that's for a future
issue.
This fixes a bug reported in #13275.
This is a follow-up to b69213808a.
We now actually send messages from the notification_bot, which
is the real usecase for this code.
Also, this cleans up the code and removes needless asserts like
`assertNotEqual(zulip_realm, lear_realm)` making the test easier
to read.
A confirmation object is already created when
do_send_confirmation_email is called just above.
Tweaked by tabbott to remove an unnecessary somewhat hacky database
query.
There are a few outstanding issues that we expect to resolve beforce
including this in a release, but this is good checkpoint to merge.
This PR is a collaboration with Tim Abbott.
Fixes#716.
Priviously, we rendered the topic links using the msg.sender.realm.
This resulted in issues with Zulip's internal bots not having access
to the realm_filters of the destination stream's realm. For example,
sending a message via the email gateway or notification would not
linkify any realm filters that a user would expect them to.
Even though required attribute of stream and stream_id params is marked
false in openapi specification, the API expects atleast one of the
params to be set. There is no way to specify relationships like this
openapi and they dont seem to have any plan to implement this in future.
https://github.com/OAI/OpenAPI-Specification/issues/256
One small change in behavior is that this creates an array with all the
row_objects at once, rather than creating them 1000 at a time.
That should be fine, given that the client batches these in units of
10000 anyway, and so we're just creating 10K rows of a relatively
small data structure in Python code here.
Fixes#1727.
With the server down, apply migrations 0245 and 0246. 0246 will remove
the pub_date column, so it's essential that the previous migrations
ran correctly to copy data before running this.
Apparently, our change in b8a1050fc4 to
stop caching responses on API endpoints accidentally ended up
affecting uploaded files as well.
Fix this by explicitly setting a Cache-Control header in our Sendfile
responses, as well as changing our outer API caching code to only set
the never cache headers if the view function didn't explicitly specify
them itself.
This is not directly related to #13088, as that is a similar issue
with the S3 backend.
Thanks to Gert Burger for the report.
This change adds the OpenAPI data needed to document the POST and
DELETE methods associated with this endpoint.
Descriptions edited slightly by tabbott.
Apparently, the Zulip notifications (and resulting emails) were
correct, but the download links inside the Zulip UI were incorrectly
not including S3 prefix on the URL, making them not work.
While we're at this, we rewrite the somewhat convoluted previous
system for formatting the data export output.
When using our EMAIL_ADDRESS_VISIBILITY_ADMINS feature, we were
apparently creating bot users with different email and delivery_email
properties, due to effectively an oversight in how the code was
written (the initial migration handled bots correctly, but not bots
created after the transition).
Following the refactor in the last commit, the fix for this is just
adding the missing conditional, a test, and a database migration to
fix any incorrectly created bots leaked previously.
This should dramatically improve the queue processor's performance in
cases where there's a very high volume of requests on a given endpoint
by a given user, as described in the new docstring.
Until we test this more broadly in production, we won't know if this
is a full solution to the problem, but I think it's likely. We've
never seen the UserActivityInterval worker end up backlogged without a
total queue processor outage, and it should have a similar workload.
Fixes#13180.
With the way these tests are, it's unnecessary to have 3 separate
classes, and it makes it confusing to decide where to add a potential
additional mm email test.
In a gigantic realm where we send several MB of `page_params`, it’s
slightly better to have the rest of the `<body>` available to the
browser earlier, so it can show the “Loading…” spinner and start
fetching subresources.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This sidesteps tricky escaping issues, and will make it easier to
build a strict Content-Security-Policy.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This brings us in line, and also allows us to style these more like
unordered lists, which is visually more appealing.
On the backend, we now use the default list blockprocessor + sane list
extension of python-markdown to get proper list markup; on the
frontend, we mostly return to upstream's code as they have followed
CommonMark on this issue.
Using <ol> here necessarily removes the behaviour of not renumbering
on lists written like 3, 4, 7; hopefully users will be OK with the
change.
Fixes#12822.
Also cleans up the interface between the management command and the
LDAP backends code to not guess/recompute under what circumstances
what should be logged.
Co-authored-by: mateuszmandera <mateusz.mandera@protonmail.com>
The order of operations for our LDAP synchronization code wasn't
correct: We would run the code to sync avatars (etc.) even for
deactivated users.
Thanks to niels for the report.
Co-authored-by: mateuszmandera <mateusz.mandera@protonmail.com>
Fixes#13130.
django_auth_ldap doesn't give any other way of detecting that LDAPError
happened other than catching the signal it emits - so we have to
register a receiver. In the receiver we just raise our own Exception
which will properly propagate without being silenced by
django_auth_ldap. This will stop execution before the user gets
deactivated.
So the reason 38f8cf612c seems
to be flaking is because the value of harry id switches between
1 and 2 in Xenial while in Bionic it would be fixed at 2. The
reason behind this is that Bionic ships with Python3.6 which
preserves dict insert order while Python3.5 that ships with Xenial
dont preserve the order. In initialize_stream_membership_dicts
we iterate user_data_map dict and the order in which the iteration
happens affects the ID of the users.
We have a very useful piece of code, _RateLimitFilter, which is
designed to avoid sending us a billion error emails in the event that
a Zulip production server is down in a way that throws the same
exception a lot. The code uses memcached to ensure we send each
traceback roughly once per Zulip server per 10 minutes (or if
memcached is unavailable, at most 1/process/10 minutes, since we use
memcached to coordinate between processes)
However, if memcached is down, there is a logging.error call internal
to the Django/memcached setup that happens inside the cache.set() call,
and those aren't caught by the `except Exception` block around it.
This ends up resulting in infinite recursion, eventually leading to
Fatal Python error: Cannot recover from stack overflow., since this
handler is configured to run for logging.error in addition to
logging.exception.
We fix this using a thread-local variable to detect whether we are
being called recursively.
This change should prevent some nasty failure modes we've had in the
past where memcached being down resulted in infinite recursion
(resulting in extra resources being consumed by our error
notifications code, and most importantly, the error notifications not
being sent).
Fixes#12595.
There's no reason for this to be a category of error that emails the
server administrator, since there's a good chance that fixing it will
need to be done in the Zulip codebase, not administrator action.
Fixes#9401.
This adds a FAKE_EMAIL_DOMAIN setting, which should be used if
EXTERNAL_HOST is not a valid domain, and something else is needed to
form bot and dummy user emails (if email visibility is turned off).
It defaults to EXTERNAL_HOST.
get_fake_email_domain() should be used to get this value. It validates
that it's correctly set - that it can be used to form valid emails.
If it's not set correctly, an exception is raised. This is the right
approach, because it's undesirable to have the server seemingly
peacefully operating with that setting misconfigured, as that could
mask some hidden sneaky bugs due to UserProfiles with invalid emails,
which would blow up the moment some code that does validate the emails
is called.
We’re about to start using PostgreSQL-specific syntax that can’t be
stringified without a specified dialect.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
Previously, several of our URL patterns accidentally did not end with
`$`, and thus ended up controlling just the stated URL, but actually a
much broader set of URLs starting with it.
I did an audit and fixed what I believe are all instances of this URL
pattern behavior. In the process, I fixed a few tests that were
unintentionally relying on the behavior.
Fixes#13082.
Historically, Zulip's implementation of wildcard mentions never
triggered either email or push notifications, instead being limited to
desktop notifications and the "mentions" counter.
We fix this just by plumbing the "wildcard_mentioned" flag through our
system.
Implements much of
https://github.com/zulip/zulip/issues/6040#issuecomment-510157264.
We're also now ready to seriously work on #3750.
Mateusz Mandera
Vishnu KS
Rohitt Vashishtha
Anders Kaseorg
Tim Abbott
Matheus Melo
Gloria Elston
Tlazypanda
Rafid Aslam
Hemanth V. Alluri
Pragati Agrawal
Rishi Gupta
Laurel Michaels
Wyatt Hoodes
Thomas Ip
Yashashvi Dave