Empirically, the retry in `_on_connection_closed` didn't actually work
-- if a reconnect failed, that was it, and the exception handler
didn't get run. A traceback would get logged, but all its frames were
in Tornado or Pika, not our own code; presumably something magic and
async was happening to the exception.
Moreover, though we would make one attempt to reconnect if we had a
connection that got closed, we didn't have any form of retry if the
original attempt at connecting failed in the first place.
Happily, upstream offers a perfectly reasonable bit of API that avoids
both of these problems: the on-open-error callback. So use that.
This method was new in Tornado 4.0. It saves us from having to get
the time ourselves and do the arithmetic -- which not only makes the
code a bit shorter, but also easier to get right. Tornado docs (see
http://www.tornadoweb.org/en/stable/ioloop.html) say we should have
been getting the time from `ioloop.time()` rather than hardcoding
`time.time()`, because the loop could e.g. be running on the
`time.monotonic()` clock.
Adding it afterward is inherently racy, and upstream's API is quite
reasonable for avoiding that -- just like we can pass an on-open
callback up front, we can do the same with the on-close callback.
This is a more thorough version of 4adf2d5c2 from back in 2013-04.
The default value of this parameter is already False upstream.
(It was already False in pika version 0.9.6, which we were
supposedly using when we introduced this in 4baeaaa52; not sure
what the story was there.)
Previously, we weren't doing a proper left join in
user_groups_in_realm_serialized, resulting in empty user groups being
excluded from the query. We want to leave decisions about excluding
empty user groups to the UI layer, so we include these here.
Because we use access_stream_by_id here, and that checks for an active
subscription to interact with a private stream, this didn't work.
The correct fix to add an option to active_stream_by_id to accept an
argument indicating whether we need an active subscription; for this
use case, we definitely do not.
[Modified by greg to (1) keep `USERNAME_FIELD = 'email'`,
(2) silence the corresponding system check, and (3) ban
reusing a system bot's email address, just like we do in
realm creation.]
As we migrate to allow reuse of the same email with multiple realms,
we need to replace the old "no email reuse" validators. Because
stealing the email for a system bot would be problematic, we still ban
doing so.
This commit only affects the realm creation logic, not registering an
account in an existing realm.
Originally this used signals, namely SIGRTMIN. But in prod, the
signal handler never fired; I debugged fruitlessly for a while, and
suspect uwsgi was foiling it in a mysterious way (which is kind of
the only way uwsgi does anything.)
So, we listen on a socket. Bit more code, and a bit trickier to
invoke, but it works.
This was developed for the investigation of memory-bloating on
chat.zulip.org that led to a331b4f64 "Optimize query_all_subs_by_stream()".
For usage instructions, see docstring.
We would allow a user with a valid invitation for one realm to use it
on a different realm instead. On a server with multiple realms, an
authorized user of one realm could use this (by sending invites to
other email addresses they control) to create accounts on other
realms. (CVE-2017-0910)
With this commit, when sending an invitation, we record the inviting
user's realm on the PreregistrationUser row; and when registering a
user, we check that the PregistrationUser realm matches the realm the
user is trying to register on. This resolves CVE-2017-0910 for
newly-sent invitations; the next commit completes the fix.
[greg: rewrote commit message]
This fixes some subtle JavaScript exceptions we've been getting in
zulipchat.com, caused by the system bot realm there not being "zulip"
interacting with get_cross_realm_users.
This should help protect us from future issues with the way that
`bulk_get_users` does caching.
It's likely that we'll want to further restructure `bulk_get_users` to
not have this base_query code path altogether (since it's kinda
buggy), but I'm going to defer that for a time when we have another
user.
The previous implementation had a subtle caching bug: because it was
sharing its cache with the `get_user_profile_by_email` cache, if a
user happened to have an email in that cache, we'd return it, even
though that user didn't match `base_query`.
This causes `get_cross_realm_users` to no longer have a problematic
caching bug.
Hides URL if the message content == image url so that sending gifs or
images feels less cluttered. Uses the url_to_a() function to generate
the expected url string for matching.
Fixes#7324.
We include ERROR_BOT in this set, even though it's not technically
cross-realm (it just lives in the admin realm).
This code path does not correctly handle emails that correspond to
multiple accounts (because `get_system_bot` does not). Since it's
intended to only be used by system bots, we add an appropriate
assertion to ensure it is only used for system bots.
This was causing problems, because internal_send_message assumes that
there is a unique user (across all realms) with the given email
address (which is sorta required to support cross-realm bot messages
the way it does).
With this change, it now, in practice, only sends cross-realm bot
messages.
Previously, this was a ValidationError, but that doesn't really make
sense, since this condition reflects an actual bug in the code.
Because this happened to be our only test coverage the ValidationError
catch on line 84 of registration.py, we add nocoverage there for now.
This buggy logic from e1686f427c had
broken do-destroy-rebuild-test-database.
Now that we're not just trying to add the Recipient objects for every
user on the system here to profiles_by_id, we also shouldn't be
processing every Recipeint object on the system. The fix is simple:
because of the patch we got merged into Django upstream,
recipients_to_create actually has the object IDs added to the
Recipient objects passed into Recipient.objects.bulk_create.
This was missed in manual testing, since it only broke `populate_db
--test-suite`.
An Integration object doesn't need access to the context dict used
to render its doc.md, since the context dict is just passed directly to
render_markdown_path.
This fixes a bug where, when a user is unsubscribed from a stream,
they might have unread messages on that stream leak. While it might
seem to be a minor problem, it can cause significant problems for
computing the `unread_msgs` data structures, since it means we need to
add an extra filter for whether the user is still subscribed, either
in the backend or in the UI.
Fixes#7095.
Rishi Gupta
Greg Price
Tim Abbott
rht
derAnfaenger
Vishnu Ks
Rohitt Vashishtha
Vaibhav Sagar
Andy Perez
Harshit Bansal
Eeshan Garg