Through internal discussion, it has become clear that
we need a conceptual Guardian user that bridges the
gap between anon users and a logged in forum user with
an absolute baseline level of access to public topics,
which can be used in cases where:
1. Automated systems are running which shouldn't see any
private data
1. A baseline level of user access is needed
In this case we are fixing the latter; when oneboxing a local
topic, and we are linking to a topic in another category from
the current one, we need to operate off a baseline level of
access, since not all users have access to the same categories,
and we don't want e.g. editing a post with an internal link to
expose sensitive internal information.
We add `Access-Control-Allow-Origin: *` to all asset requests which are requested via a configured CDN. This is particularly important now that we're using browser-native `import()` to load the highlightjs bundle. Unfortunately, user-configurable 'cors_origins' site setting was overriding the wldcard value on CDN assets and causing CORS errors.
This commit updates the logic to give the `*` value precedence, and adds a spec for the situation. It also invalidates the cache of hljs assets (because CDNs will have cached the bad Access-Control-Allow-Origin header).
The rack-cors middleware is also slightly tweaked so that it is always inserted. This makes things easier to test and more consistent.
Why was the problem?
ActiveRecord's query cache for the connection pool wasn't disabled after the
`with a fake provider runs 'other_phase' for enabled auth methods` test
in `omniauth_callbacks_controller_spec.rb` was run. This was because the
Rack response body in `FakeAuthenticator::Strategy::other_phase` did not
adhere to the expected Rack body format which is "typically an Array of
String instances". Because this expectation was broken, it cascaded the
problem down where it resulted in the ActiveRecord's query cache for the
connection pool not being disabled as it normally should when the
response body is closed.
When the query cache is left enabled, common assertions pattern in RSpec
like `expect { something }.to change { Group.count }` will fail since
the query cache is enabled and the call first call to `Group.count` will
cache the result to be reused later on.
To see the bug in action, one can run the following command:
`bundle exec rspec --seed 44747
spec/requests/omniauth_callbacks_controller_spec.rb:1150
spec/models/group_spec.rb:283`
Followup e37fb3042d
* Automatically remove the prefix `Discourse ` from all the plugin titles to avoid repetition
* Remove the :discourse_dev: icon from the author. Consider a "By Discourse" with no labels as official
* We add a `label` metadata to plugin.rb
* Only plugins made by us in `discourse` and `discourse-org` GitHub organizations will show these in the list
* Make the plugin author font size a little smaller
* Make the commit sha look like a link so it's more obvious it goes to the code
Also I added some validation and truncation for plugin metadata
parsing since currently you can put absolutely anything in there
and it will show on the plugin list.
The category drop was rerendered after every category async change
because it updated the categories list. This is not necessary and
categories can be referenced indirectly by ID instead.
This change refactors the check `user.groups.any?` and instead uses
`user.staged?` to check if the user is staged or not.
Also fixes several tests to ensure the users have their auto trust level
groups created.
Follow up to:
- 8a45f84277
- 447d9b2105
- c89edd9e86
Why this change?
Asserting against records of the database in system tests can be flaky
because those assertions can run against the database before the server
has actually saved the necessary changes to the database.
What does this change do?
While the assertion is not ideal, we are working around this as a
temporary fix by using `try_until_success` which will retry the
assertion up till the default capybara timeout.
This value is included when generating static asset URLs. Updating the value will allow site operators to invalidate all asset urls to recover from configuration issues which may have been cached by CDNs/browsers.
When sending SMTP for group SMTP functionality, we
are running into timeouts for both read and open
when sending mail occassionally, which can cause issues
like the email only being sent to _some_ of the recipients
or to fail altogether.
The defaults of 5s are too low, so bumping them up to
the defaults of the `net-smtp` gem.
When we check upload security, one of the checks is to
run `access_control_post.with_secure_uploads?`. The problem
here is that the `topic` for the post could be deleted,
which would make the check return `nil` sometimes instead
of false because of safe navigation. We just need to be
more explicit.
Admin can add tag description up to 1000 characters.
Full description is displayed on tag page, however on topic list it is truncated to 80 characters.
Why this change?
In the `invites_controller_spec.rb` file, we had several tests that were
checking for assets path in the response's body to determine which
layout has been rendered. However, those test fails if `bin/ember-cli
--build` has been run locally.
What does this change do?
Instead of checking for asset paths to determine the layout that has
been rendered, this change relies on the fact that the `no_ember` layout
has a `no-ember` class on the `body` element. This is more deterministic
as compared to relying on the different asset paths that are rendered in
the response.
Followup to 2443446e62
We introduced video placeholders which prevent preloading
metadata for videos in posts. The structure looks like this
in HTML when the post is cooked:
```
<div class="video-placeholder-container" data-video-src="http://some-url.com/video.mp4" dir="ltr" style="cursor: pointer;">
<div class="video-placeholder-wrapper">
<div class="video-placeholder-overlay">
<svg class="fa d-icon d-icon-play svg-icon svg-string" xmlns="http://www.w3.org/2000/svg">
<use href="#play"></use>
</svg>
</div>
</div>
</div>
```
However, we did not update the code that links post uploads
to the post via UploadReference, so any videos uploaded since
this change are essentially dangling and liable to be deleted.
This also causes some uploads to be marked secure when they
shouldn't be, because they are not picked up and analysed in the
CookedPostProcessor flow.
Followup to e37fb3042d,
in some cases we cannot get git information for the
plugin folder (e.g. permission issues), so we need
to only try and get information about it if
commit_hash is present.
Reverts
- DEV: maxmind license checking failing tests #24534
- UX: Show if MaxMind key is missing on IP lookup #18993
These changes are leading to surprising results, our logs are now filling up with warnings on dev environments
We need the change to be redone
This improves the implementation of #18993
1. Error message displayed to user is clearer
2. open_db will also be called, even if license key is blank, as it was previously
3. This in turn means no need to keep stubbing 'maxmind_license_key'
The parent category needs to be serialized before the child category
because they are parsed in order. Otherwise the client will not build
the parent-child relationship correctly.
This change converts the `email_in_min_trust` site setting to
`email_in_allowed_groups`.
See: https://meta.discourse.org/t/283408
- Hides the old setting
- Adds the new site setting
- Add a deprecation warning
- Updates to use the new setting
- Adds a migration to fill in the new setting if the old setting was
changed
- Adds an entry to the site_setting.keywords section
- Updates tests to account for the new change
After a couple of months we will remove the
`email_in_min_trust` setting entirely.
Internal ref: /t/115696
Followup to e37fb3042d
Some plugins like discourse-ai and discourse-saml do not
nicely change from kebab-case to Title Case (e.g. Ai, Saml),
and anyway this method of getting the plugin name is not
translated either.
Better to use the plugin setting category if it exists,
since that is written by a human and is translated.
* DEV: Convert approve_new_topics_unless_trust_level to groups
This change converts the `approve_new_topics_unless_trust_level` site
setting to `approve_new_topics_unless_allowed_groups`.
See: https://meta.discourse.org/t/283408
- Hides the old setting
- Adds the new site setting
- Add a deprecation warning
- Updates to use the new setting
- Adds a migration to fill in the new setting if the old setting was
changed
- Adds an entry to the site_setting.keywords section
- Updates tests to account for the new change
After a couple of months we will remove the
`approve_new_topics_unless_trust_level` setting entirely.
Internal ref: /t/115696
* add missing translation
* Add keyword entry
* Add migration
This commit extracts the storage part of the route-scroll-manager into a dedicated service. This provides a key/value store which will reset for each navigation, and restore previous values when the user uses the back/forward buttons in their browser.
This gives us a reliable replacement for the old `DiscourseRoute.isPoppedState` function, which would not work under all situations.
Previously reverted in e6370decfd. This version has been significantly refactored, and includes an additional system spec for the issue we identified.
We were throwing ArgumentError in UrlHelper.normalised_encode,
but it was incorrect -- we were passing ArgumentError.new
2 arguments which is not supported. Fix this and have a hint
of which URL is causing the issue for debugging.
This change converts the `approve_unless_trust_level` site setting to
`approve_unless_allowed_groups`.
See: https://meta.discourse.org/t/283408
- Adds the new site setting
- Adds a deprecation warning
- Updates core to use the new settings.
- Adds a migration to fill in the new setting of the old setting was
changed
- Adds an entry to the site_setting.keywords section
- Updates many tests to account for the new change
After a couple of months we will remove the `approve_unless_trust_level`
setting entirely.
Internal ref: /t/115696
Why this change?
The test was randomly failing in
https://github.com/discourse/discourse/actions/runs/6936264158/job/18868087113
with the following failure:
```
expect do user.update_ip_address!("127.0.0.1") end.to change {
UserIpAddressHistory.where(user_id: user.id).count
}.by(1)
expected `UserIpAddressHistory.where(user_id: user.id).count` to have changed by 1, but was changed by 0
```
This is due to the fact that ActiveRecord will actually cache the result
of `UserIpAddressHistory.where(user_id: user.id).count`. However,
`User.update_ip_address!` relies on mini_sql and does not go through
ActiveRecord. As a result, the query cache is not cleared and hence the
flakiness.
What does this change do?
This change uses the `uncached` method provided by ActiveRecord when
we are fetching the count.
* Remove checkmark for official plugins
* Add author for plugin, which is By Discourse for all discourse
and discourse-org github plugins
* Link to meta topic instead of github repo
* Add experimental flag for plugin metadata and show this as a
badge on the plugin list if present
---------
Co-authored-by: chapoi <101828855+chapoi@users.noreply.github.com>
In the long term we should aim to modernize these places, but for now this change will make them compatible with Ember 5.x (while maintaining compatibility with Ember 3.28)
This commit adds a new `search_default_sort_order` site setting,
set to "relevance" by default, that controls the default sort order
for the full page /search route.
If the user changes the order in the dropdown on that page, we remember
their preference automatically, and it takes precedence over the site
setting as a default from then on. This way people who prefer e.g.
Latest Post as their default can make it so.
Currently to use a limit in the notifications index, you have to also pass recent: true as a param.
This PR:
Adds optional limit param to be used in the notifications query, regardless of the presence of recent
Raises the max limit of the response with recent present from 50 -> 60. It is super weird we have a hard-limit of 50 before with recent param, and 60 without the param.
config.after(:suite) which stops minio server is called every time one
of the groups of parallel tests complete. This works fine most of the
time with parallel spec runs, but sometimes one of these
MinioRunner.stop calls happens while a spec is running in another
process that expects the minio server to be running.
Skipping these tests to avoid flakys for now.
Why this change?
The test became flaky due to d208396c5c.
In that commit, we introduced `page.has_no_css?("div.menu-panel.animating")` to `PageObjects::Components::NavigationMenu::Sidebar#open_on_mobile` but
it did not work as intended because `page.has_no_css?("div.menu-panel.animating")` can return `true` immediately as the `animating` class has not been added
to the element.
What does this change do?
Switch to the `wait_for_animation` system helper to ensure that all
animations have ended on the element.
When we started using NumberField for integer site settings
in e113eff663, we did not end up
passing down a min/max value for the integer to the field, which
meant that for some fields where negative numbers were allowed
we were not accepting that as valid input.
This commit passes down the min/max options from the server for
integer settings then in turn passes them down to NumberField.
c.f. https://meta.discourse.org/t/delete-user-self-max-post-count-not-accepting-1-to-disable/285162
This PR refactors the following:
* leaving all the CSS applied to the old `modal-body` classes in their respective files
* made new clean styling for `.d-modal` and refactored the template to use the new BEM classes
* `inner-`, `middle-`, `outer-` container classes are gone and replaced with simplified `wrapper` and `container` classes
* use standardised max-sizes with modifiers `-large` and `-max`
* lighter backdrop,
* min-width to prevent puny modals
* other styling changes regarding padding, close button,…
* pulled out all modal overrides into a general `modal-overrides` file + cleanup of outdated CSS
* pulled out login and create account modal styling into their own file, cause it's such a big override
* removed old general login.scss file for mobile & desktop
* only kept some remainders I don't want to touch in `app/assets/stylesheets/common/base/login.scss`
Previously we would only recompile a theme locale when its own data changes. However, the output also includes fallback data from other locales, so we need to invalidate all locales when fallback locale data is changed. Building a list of dependent locales is tricky, so let's just invalidate them all.
We ask users to confirm their session if they are making a sensitive
action, such as adding/updating second factors or passkeys. This
commit adds the ability to confirm sessions with passkeys as an option
to the password confirmation.
The `src` of js files is now dependent on the ember-cli/webpack build, so it's not a good thing to check in specs. In CI it passes because the ember-cli build is not run. But locally it would fail if you had a build in `app/assets/javascripts/discourse/dist`.
This commit updates the specs to check for the presence of a stable data attribute instead.
- Remove vendored copy
- Update Rails implementation to look for language definitions in node_modules
- Use webpack-based dynamic import for hljs core
- Use browser-native dynamic import for site-specific language bundle (and fallback to webpack-based dynamic import in tests)
- Simplify markdown implementation to allow all languages into the `lang-{blah}` className
- Now that all languages are passed through, resolve aliases at runtime to avoid the need for the pre-built `highlightjs-aliases` index
Previously, the app HTML served by the Ember-CLI proxy was generated based on a 'bootstrap json' payload generated by Rails. This inevitably leads to differences between the Rails HTML and the Ember-CLI HTML.
This commit overhauls our proxying strategy. Now, we totally ignore the ember-cli `index.html` file. Instead, we take the full HTML from Rails and surgically replace script URLs based on a `data-discourse-entrypoint` attribute. This should be faster (only one request to Rails), more robust, and less confusing for developers.
The most common thing that we do with fab! is:
fab!(:thing) { Fabricate(:thing) }
This commit adds a shorthand for this which is just simply:
fab!(:thing)
i.e. If you omit the block, then, by default, you'll get a `Fabricate`d object using the fabricator of the same name.
This adds the ability to collect stats without exposing them
among other stats via API.
The most important thing I wanted to achieve is to provide
an API where stats are not exposed by default, and a developer
has to explicitly specify that they should be
exposed (`expose_via_api: true`). Implementing an opposite
solution would be simpler, but that's less safe in terms of
potential security issues.
When working on this, I had to refactor the current solution.
I would go even further with the refactoring, but the next steps
seem to be going too far in changing the solution we have,
and that would also take more time. Two things that can be
improved in the future:
1. Data structures for holding stats can be further improved
2. Core stats are hard-coded in the About template (it's hard
to fix it without correcting data structures first, see point 1):
63a0700d45/app/views/about/index.html.erb (L61-L101)
The most significant refactorings are:
1. Introducing the `Stat` model
2. Aligning the way the core and the plugin stats' are registered
There was a registry for preloaded site categories and a new one has
been introduced recently for categories serialized through a
CategoryList.
Having two registries created a lot of friction for developers and this
commit merges them into a single one, providing a unified API.
There is an edge case where the following occurs:
1. The user sets a bookmark reminder on a post/topic
2. The post/topic is changed to a PM before or after the reminder
fires, and the notification remains unread by the user
3. The user opens their bookmark reminder notification list
and they can still see the notification even though they cannot
access the topic anymore
There is a very low chance for information leaking here, since
the only thing that could be exposed is the topic title if it
changes to something sensitive.
This commit filters the bookmark unread notifications by using
the bookmarkable can_see? methods and also prevents sending
reminder notifications for bookmarks the user can no longer see.
When quoting a chat message in a post, if that message contains a mention,
that mention should be ignored. But we've been detecting them and sending
notifications to users. This PR fixes the problem. Since this fix is for
the chat plugin, I had to introduce a new API for plugins:
# We strip posts before detecting mentions, oneboxes, attachments etc.
# We strip those elements that shouldn't be detected. For example,
# a mention inside a quote should be ignored, so we strip it off.
# Using this API plugins can register their own post strippers.
def register_post_stripper(&block)
end
This commit adds an /admin/customize/theme-components route,
that opens the theme page with the components tab pre-selected,
so people can navigate to that directly.
Switches to using a dialog to confirm a session (i.e. sudo mode for
account changes where we want to be extra sure the current user is who
they say they are) to match what we do with passkeys.
Previously, we were parsing webpack JS chunk filenames from the HTML files which ember-cli generates. This worked ok for simple entrypoints, but falls apart once we start using async imports(), which are not included in the HTML.
This commit uses the stats plugin to generate an assets.json file, and updates Rails to parse it instead of the HTML. Caching on the Rails side is also improved to avoid reading from the filesystem multiple times per request in develoment.
Co-authored-by: Godfrey Chan <godfreykfc@gmail.com>
Daniel Waterworth
Joffrey JAFFEUX
Martin Brennan
Keegan George
David Taylor
Roman Rizzi
Jarek Radosz
Ted Johansson
Penar Musaraj
Alan Guo Xiang Tan
Bianca Nenciu
Sam
Blake Erickson
Natalie Tay
Krzysztof Kotlarek
MichaIng
Gerhard Schlager
Mark VanLandingham
chapoi
Kris
David Battersby
Andrei Prigorshnev
Régis Hanol
Jean
Kelv