When the server gets overloaded and lots of requests start queuing server
will attempt to shed load by returning 429 errors on background requests.
The client can flag a request as background by setting the header:
`Discourse-Background` to `true`
Out-of-the-box we shed load when the queue time goes above 0.5 seconds.
The only request we shed at the moment is the request to load up a new post
when someone posts to a topic.
We can extend this as we go with a more general pattern on the client.
Previous to this change, rate limiting would "break" the post stream which
would make suggested topics vanish and users would have to scroll the page
to see more posts in the topic.
Server needs this protection for cases where tons of clients are navigated
to a topic and a new post is made. This can lead to a self inflicted denial
of service if enough clients are viewing the topic.
Due to the internal security design of Discourse it is hard for a large
number of clients to share a channel where we would pass the full post body
via the message bus.
It also renames (and deprecates) triggerNewPostInStream to triggerNewPostsInStream
This allows us to load a batch of new posts cleanly, so the controller can
keep track of a backlog
Co-authored-by: Joffrey JAFFEUX <j.jaffeux@gmail.com>
A follow-up to a follow-up. (6932a373a3 and 572da7a57b)
Our `discourse_test` Docker image uses git 2.20.1 released on Dec 15, 2018. It does not support `git branch --show-current`. (it was added in 2.22.0)
Previously, any errors in those files would e.g. blow up the update process in docker_manager.
Now it prints out an error and proceeds as if there was no compatibility file.
Includes:
* DEV: Extract setup_git_repo
* DEV: Use `Dir.mktmpdir`
* DEV: Default to `main` branch (The latest versions of git already do this, so to avoid problems do this by default)
We were trying to observe a non-ember object which is undefined
behavior and was leaking to odd bugs. This replaces the `filter` object
with an Ember Object and things seem to work.
I also took the opportunity with this commit to move some test specific
stuff out of `discourse-loader` which is loaded on the front end of the
application. The test module building now happens in the `test_helper`
bundle.
* FIX: second factor cannot be enabled if SSO is enabled
If `enable_sso` setting is enabled then admin should not be able to
enable `enforce_second_factor` setting as that will lock users out.
Co-authored-by: Robin Ward <robin.ward@gmail.com>
DEV: Replace instances of Discourse.base_uri with Discourse.base_path
This is clearer because the base_uri is actually just a path prefix. This continues the work started in 555f467.
This misses a test because Favcount doesn't exposes a get to the counter.
Also, since this code deals with all possible notifications configs we support:
- favicon notification
- favicon new content
- title notification
- title new content
the code is a bit complicated to follow. We may look into refactoring it when a
good opportunity arises, like if https://w3c.github.io/badging/ setClientBadge() method
gives us a cleaner way to notify users.
We can't use erb in Ember CLI (since it does not have Ruby) so this has
been ported to use our `javascript:update_constants` rake test instead.
Note we don't have to run this every time a notification type as it's
only used by fixtures to fill in some specific types we test against.
This is long overdue. We had a lot of (not linted) code to initialize
our test suite as part of the Ruby `test_helper.js` bundle.
This refactor moves that out to a `setup-tests` module, which imports
all the modules properly, rather than using `require`.
It also removes the global `server` variable which some tests were using
for pretender. Those tests are fixed, and in the case of widget tests,
support for a `pretend()` was added, which mimics our acceptance tests.
One problematic test was removed, which overwrites `/posts` - this could
break tons of other tests depending on order.
The prefixing logic is moved into a `prefixProtocol` function in lib:url.
This commit also renames an incorrectly named test and uses https as default instead of http, in 2020 it's reasonable to think we most likely want https and not http. User can still specify http if required.
This commit is also moving one test to a component test.
A followup to this commit would be to ensure every dropdowns are using a regex instead of the normalize/lowercase system we have now.
See https://meta.discourse.org/t/email-address-change-confirmation-email-not-sent-but-every-other-notification-emails-are/165358
In short: with disable emails set to non-staff, email address change confirmation emails (those sent to the new address) are not sent for staff or admin members.
This was happening because we were looking up the staff user with the to_address of the email, but the to address was the new email address because we are sending a confirm email change email, and thus the user could not be found. We didn't need to do this anyway because we are passing the user into the Email::Sender class anyway.
This code is a no-op on all sites, even though it looks rather dangerous
this migration has long run prior to people trying exploit it.
That said ... hygiene here ... is not good.
Remove this legacy, we do not want it, even in historical migrations.
* PERF: avoid lookbehinds when indexing search
Previously we used a `EmailCook.url_regexp` this regex used lookbehinds
Unfortunately certain strings could lead to pathological behavior causing
CPU to skyrocket and regex replace to take a very very long time.
EmailCook still needs a fix, but it is less urgent cause it already splits
to single lines. That said we will correct that as well in a seperate PR.
New implementation is far more naive and relies on the extra spaces search
indexer inserts.
The dark-mode-friendly SVG mask for the wizard's background image
introduced in 8fcfb9586c does not work with
CDNs, because CORS restrictions apply to SVG masks.
It would be complicated to modify CDN access origin rules for this one
specific assets, so instead, this PR moves the contents of the SVG file
inside the stylesheet.
When that site setting is enabled, the category counts (new/unread)
include the subcategory definition topics, but the topics aren't included
in the list. This fixes that discrepancy.
These are tricky because `module.exports` is used by nodejs files as a
global, which is OK. But we don't want to allow `module` in JS tests
for qunit without importing it first.
We used many global functions to handle tests when they should be
imported like other libraries in our application. This also gets us
closer to the way Ember CLI prefers our tests to be laid out.
Previously, Jobs::EnqueueDigestEmails would enqueue a digest job for every user, even if there are no topics to send. The digest job would exit, no email would send, and last_emailed_at would not change. 30 minutes later, Jobs::EnqueueDigestEmails would run again and re-enqueue jobs for the same users.
120fa8ad introduced a temporary mitigation for this issue, by randomly selecting a subset of those users each time.
This commit adds a new `digest_attempted_at` column to the `user_stats` table. This column is updated every time a digest job completes for a user. Using this, we can avoid scheduling digest jobs for the same user every 30 minutes. This also removes the random user selection in 120fa8ad, and instead prioritizes users who had digests attempted the longest time ago.
This plugin is only useful for developers, however, making it core allows us to centralize any component modification in one commit.
This integration also adds a new site_setting: `styleguide_admin_only` which allows to enable a styleguide on a live site while restricting visibility to admins only.
By default, styleguide is disabled.
To avoid blocking the sidekiq queue a limit of 10,000 digests per 30 minutes
is introduced.
This acts as a safety measure that makes sure we don't keep pouring oil on
a fire.
On multisites it is recommended to set the number way lower so sites do not
dominate the backlog. A reasonable default for multisites may be 100-500.
This can be controlled with the environment var
DISCOURSE_MAX_DIGESTS_ENQUEUED_PER_30_MINS_PER_SITE
See https://meta.discourse.org/t/changing-a-users-email/164512 for additional context.
Previously when an admin user changed a user's email we assumed that they would need a password reset too because they likely did not have access to their account. This proved to be incorrect, as there are other reasons a user needs admin to change their email. This PR:
* Changes the admin change email for user flow so the user is sent an email to confirm the change
* We now record who the email change request was requested by
* If the requested by user is admin and not the user we note this in the email sent to the user
* We also make the confirm change email route open to anonymous users, so it can be clicked by the user even if they do not have access to their account. If there is a logged in user we make sure the confirmation matches the current user.
