* FIX: bots could generate errors when slug generation method is encoded
When slug generation method is encoded (non default) then bots could
cause errors in the logs for urls containing special chars.
ó for example in a URL can be requested in a valid ASCII-8BIT string, and
later when joined to UTF-8 would result in encoding issues.
Fix here ensures we force encoding correctly for outlier cases.
Browser tend to always encode these chars, hence we did not notice this.
Co-authored-by: Jarek Radosz <jradosz@gmail.com>
A click event is trigger on a link click as well which is not what we
want. This caused the links to trigger a full reload instead of an Ember
transition for some reason.
There was a minor issue where the bookmark loading
spinner would not show correctly because of how
the route was handling the setting of loading,
this fixes the issue.
* Use `Set` instead of `Array` for `this.newIncoming`
* Remove `isUnseen()`
* Use array spread instead of `Array.from()`
* Don't use `@on()`
* Fix typos
* Make sure `this.incomingCount` is always a Number
Relative URLs will work just fine for Web Workers, which were the original reason for introducing the `theme_uploads_local` feature
Making them relative will mean that `loadScript()` automatically uses the CDN (when enabled), which is doubly important because our CSP doesn't allow loading theme-javascripts from the host domain when the CDN is enabled.
Some errors (e.g. InvalidAccess) are rendered with `include_ember: true`. Booting the ember app requires that the 'preload' data is rendered in the HTML.
If a particular route was configured to `skip_before_action :preload_json`, and then went on to raise an InvalidAccess error, then we'd attempt to render the Ember app without the preload json. This led to a blank screen and a client-side error.
This commit ensures that error pages will fallback to the no_ember view if there is no preload data. It also adds a sanity check in `discourse-bootstrap` so that it's easier for us to identify similar errors in future.
Occasionally some code (e.g. live-reload) would try to clear a timer that was set up before fake timers were installed. That would lead to issues and warnings. Enabling `shouldClearNativeTimers` option fixes it.
This will allow consumers to inject it using `messageBus: service()` in preparation for the removal of implicit injections in Ember 4.0. `message-bus:main` is still available and will print a deprecation notice.
The MessageBus library is not en ember object, and doesn't need access to any of our injections. Therefore, we can set up a simple class which defines itself as a 'Service Factory', and returns the MessageBus library in the `create` method.
This will allow consumers to inject it using `pmTopicTrackingState: service()` in preparation for the removal of implicit injections in Ember 4.0. `pm-topic-tracking-state:main` is still available and will print a deprecation notice.
This will allow consumers to inject it using `keyValueStore: service()` in preparation for the removal of implicit injections in Ember 4.0. `key-value-store:main` is still available and will print a deprecation notice.
To make this conversion possible, we have to bypass the `app.inject` logic which blocks injecting services into services. This is not ideal, but there is no other way for us to do this in a way that is backwards-compatible, and will still print a useful deprecation message when we eventually turn on the implicit-injections deprecation notice.
Not only is the current code not needed but it's also creating invalid values on safari if the header goes out of viewport with negative values.
This commit also adds a missing test for `--header-offset` property.
When viewing a topic, we execute two queries to fetch the topic's
public topic timer and slow mode timer. The former query happens to be
able to use a unique index but the latter has to do a seq scan which is
slow. The query itself is not expensive but since viewing a topic is a
hot path, the little cuts add up overtime and the query itself
contributes significantly to the load of the database.
This commit removes the ability to enable/disable the Sidebar on a per
user basis and introduces a site wide setting. For testing purposes, sidebar can be enabled/disabled via the `enable_sidebar=1` or `enable_sidebar=0` query param.
* FEATURE: revamped wizard
* UX: Wizard redesign (#17381)
* UX: Step 1-2
* swap out images
* UX: Finalize all steps
* UX: mobile
* UX: Fix test
* more test
* DEV: remove unneeded wizard components
* DEV: fix wizard tests
* DEV: update rails tests for new wizard
* Remove empty hbs files that were created because of rebase
* Fixes for rebase
* Fix wizard image link
* More rebase fixes
* Fix rails tests
* FIX: Update preview for new color schemes: (#17481)
* UX: make layout more responsive, update images
* fix typo
* DEV: move discourse logo svg to template only component
* DEV: formatting improvements
* Remove unneeded files
* Add tests for privacy step
* Fix banner image height for step "ready"
Co-authored-by: Jordan Vidrine <30537603+jordanvidrine@users.noreply.github.com>
Co-authored-by: awesomerobot <kris.aubuchon@discourse.org>
Previously, non-staff users could only quote if they had an open composer.
This change shows the quote control when selecting text in closed topics
at all times and if the composer isn't already open, it will default to
creating a linked topic.
Previously, if a non-admin controller did not have a template defined, then the resolver would return an admin template with the same name. This is not the desired behavior, and regressed in fc36ac6cde. However, we *do* want this behavior for components defined in the admin bundle (because admin components are not namespaced).
This was noticed because the non-admin `badges` route was using the `admin/badges` template
This commit fixes the behavior, and adds a tests for these cases.
The previous method for reused the PrettyText logic which applied the
watched word logic, but had the unwanted effect of cooking the text too.
This meant that regular text values were converted to HTML.
Follow up to commit 5a4c35f627.
Sometimes the user list was incomplete when multiple requests were
created to fetch next pages. If the responses did not arrive in the
same order as the requests then only the last response was parsed.
This is a follow up commit to a0f4c7fe88.
The main landing page for admins/mods should be the admin Dashboard, not the site settings. Having a "Settings" link can be confused with user settings.
This commit also displays the button for moderators, who are also allowed access to parts of the admin dashboard)
This allows consumers to pass in, and receive, timestamps for a different timezone. Previously, attempting this would lead to very strange behavior which would become worse the further the input timestamp's timezone was from the browser's timezone.
The default behavior is unchanged - the browser's timezone will be assumed.
Follow up to: #17619
Context: https://meta.discourse.org/t/introducing-discourse-splash-a-visual-preloader-displayed-while-site-assets-load/232003/17
We previously relied on the user's browser when deciding when to show the splash in light/dark mode. This worked well but can fail if the user manually selects a theme with a default "dark" scheme.
This PR will now factor that in. If the user selects a theme with a default dark scheme, use that. If a user selects a theme with a "light" default scheme and also picks a secondary "dark" scheme, use the media detection we had before.
This PR also removes the dark mode theme-color that was added in the previous PR. That will now go in a separate PR
Tags mixin is already filtering a lot of data from the user submitted filter in `createContentFromInput()` which can lead to sk receiving an empty filter while the input actually has a value.
Before this change, undocking the sidebar would just hide the sidebar
from the screen which led people to complain that they "lost" their
sidebar and had to "find" it. With this change, we automatically display
the sidebar hamburger dropdown when you undock the sidebar. Like wise
when the sidebar is docked, the sidebar hamburger dropdown is
automatically collapsed.
Context: https://meta.discourse.org/t/introducing-discourse-splash-a-visual-preloader-displayed-while-site-assets-load/232003/17
We currently set the theme secondary color as the background for the splash, and this works and respects light/dark modes.
The issue is that we set it on the #d-splash div. That div doesn't have a specified height and only gets its height when the splash image loads.
This can cause a flicker effect where the <HTML> background shows for a fraction of a second while the splash image loads.
This PR sets the theme color on the <HTML> tag to alleviate this. This allows us to set the theme color a little bit sooner and should hopefully prevent the flicker effect from happening.
This PR also adds the theme-color <meta> tag for dark mode. Browsers that don't support multiple theme-color tags will ignore the second tag and fall back to the first one.
Our theme system is very complex and it can take a while to figure out how to invalidate the various types of caches that are used throughout the theme system. So, having a single helper method that invalidates everything can be useful in emergency situations where there is no time to read through the code and figure out how to clear the various caches.
Internal ticket: t64732.
When the experimental Sidebar is enabled, the hamburger drop down is replaced by a sidebar drop down. A user is given the ability to dock and undock the sidebar depending on their personal preference.
Do also note that the experimental sidebar is well, considered experimental at this point so I do not intend for the features here to be perfect. What I aim to do here is to ship the changes fast so that it can be used internally by the team to provide feedback. Custom links added by plugins and dark mode toggle has not been implemented as part of this commit as I aim to tackle it in another commit.
Co-authored-by: awesomerobot <kris.aubuchon@discourse.org>
* DEV: Improve documentation for custom sidebar section interfaces
* DEV: Improvements to sidebar custom section plugin API tests
* DEV: Support models argument for custom sidebar section link API
This allows an arbitrary Glimmer template to be rendered inside a Widget. That template can include any kind content, including Classic Ember components and Glimmer components. This leans on Ember's official `{{#in-element}}` helper which means that all component lifecycle hooks are called correctly.
This is a modern replacement for our existing `ComponentConnector` implementation. We'll deprecate `ComponentConnector` in the near future.
Example usage:
```javascript
// (inside an existing widget)
html(){
return [
new RenderGlimmer(
this,
"div.my-wrapper-class",
hbs`<MyComponent @arg1={{@data.arg1}} />`,
{
arg1: "some argument value"
}
),
]
}
```
See `widgets/render-glimmer.js` for documentation, and `render-glimmer-test` for more example uses.
`ember-modifier` provides APIs to write our own modifiers. `@ember/render-modifiers` provides some simple `did-insert`, `did-update` and `will-destroy` modifiers.
Upcoming `ember-modifier` deprecations have been set to 'throw' in our deprecation-workflow config to ensure we don't accidently start using deprecated behaviour.
Previously it would randomize the order only when running tests:
1 .through ember-exam
2. in browser, with no params
Running just core tests, or just plugins, or a single plugin, or with filter, etc. disabled randomization.
Now all those cases are covered.
When a topic was published from a shared draft and it had tags, the
users watching the tags were not notified. The problem was that the
topics are usually created in a secret category and publishing it just
moves an existent topic to the target category, without making any
changes to the tags.
This commit improves several parts of the group members bulk operation.
It fixes the bug that did not show the menu button when the Select all
button was clicked. The other changes make the behavior more consistent
with topic list bulk operations.
The invites should be redeemed during the signup process. This was a
problem because when user tried to redeem an admin invite it tried to
authenticate the user using information from the session that was not
available.
Plugins/themes use injections in their raw view class. On recent versions of Ember these were failing with 'Attempting to lookup an injected property on an object without a container, ensure that the object was instantiated via a container.'
This prevents a storm of deprecation messages in the developer console. We'll be working through and enabling these one-by-one over the coming weeks/months.
A dummy `discourse-ensure-deprecation-order` package is introduced to ensure that deprecation-workflow is loaded before `@ember/jquery`. This ensures that the `@ember/jquery`-triggered deprecation warnings can be silenced correctly
This also introduces a system for silencing CLI warnings.
No need for us to specify our custom teardown hook when Sidebar
component is destroyed when each custom section link is expected to be
its own component and can implement its own `willDestory` hook.
This is so we can join the Notification table onto the
Bookmark table. A slight refactor was needed to ensure
that the required values are always included and the
consumer does not need to think about this.
The discourse-chat and discourse-data-explorer plugins
will be updated to take advantage of this commit.
1. Injecting `appEvents` service into `screen-track` was unnecessary as it's already injected into all services (and was causing an assertion error)
2. Return a promise from `sendNextConsolidatedTiming()` (no need for `await settled()` then)
Also, the change in insert-hyperlink (from `this.linkUrl.indexOf("http") === -1` to `!this.linkUrl.startsWith("http")`) was intentional fix: we don't want to prevent users from looking up topics with http in their titles.
This commit introduces a new plugin API to register
a group of stats that will be included in about.json
and also conditionally in the site about UI at /about.
The usage is like this:
```ruby
register_about_stat_group("chat_messages", show_in_ui: true) do
{
last_day: 1,
"7_days" => 10,
"30_days" => 100,
count: 1000,
previous_30_days: 120
}
end
```
In reality the stats will be generated any way the implementer
chooses within the plugin. The `last_day`, `7_days`, `30_days,` and `count`
keys must be present but apart from that additional stats may be added.
Only those core 4 stat keys will be shown in the UI, but everything will be shown
in about.json.
The stat group name is used to prefix the stats in about.json like so:
```json
"chat_messages_last_day": 2322,
"chat_messages_7_days": 2322,
"chat_messages_30_days": 2322,
"chat_messages_count": 2322,
```
The `show_in_ui` option (default false) is used to determine whether the
group of stats is shown on the site About page in the Site Statistics
table. Some stats may be needed purely for reporting purposes and thus
do not need to be shown in the UI to admins/users. An extension to the Site
serializer, `displayed_about_plugin_stat_groups`, has been added so this
can be inspected on the client-side.
* FIX: properly validate multiselect user fields on user creation
* Add test cases
* FIX: don't check multiselect user fields for watched words
* Clarifiy/simplify tests
* Roll back apply_watched_words changes
Since this method no longer needs to deal with arrays for now. If/when
we add new user fields which uses them, we can deal with it then.
All `DistributedCache` instances in Discourse are automatically keyed on the `Discourse.git_version`. Normally the theme compiler version is updated via a commit, and everything is fine. However, in some situations, it's possible for the BASE_COMPILER_VERSION to change without a change to the git_version (e.g. when applying patches directly to the codebase).
This commit adds the `BASE_COMPILER_VERSION` to the DistributedCache key to ensure that content from different compiler versions does not leak into other processes.
This logic should no longer be triggered. The EMBER_GLIMMER_SET_COMPONENT_TEMPLATE gets removed in recent versions of Ember, which can cause it to accidently trigger and cause layout issues with some plugins/themes.
The `unread_not_too_old` attribute is a little odd because there should never be a case where
the user's first_unread_at column is less than the `Topic#updated_at`
column of an unread topic. The `unread_not_too_old` attribute is causing
a bug where topic states synced into `TopicTrackingState` do not appear
as unread because the attribute does not exsist on a normal `Topic`
object and hence never set.
It makes more sense to use user_ids for the UserCommScreener
introduced in fa5f3e228c since
in most cases the ID will be available, not the username. This
was discovered while starting work on a plugin that will
use this. In the cases where only usernames are available
the extra query is negligble.
Context: https://meta.discourse.org/t/pixel-jump-whenever-page-refreshes-mobile-desktop/231053
We currently add 1 extra pixel when we try to restore the last scroll position on a few routes.
This is causing a bit of jumpiness, as described in the linked topic above.
<img height="300" src="ae9e69f7fd.gif">
Notice how the content shifts by 1px while stuff loads.
I believe this 1px that we add is an artifact from the days when we used to set the header to `fixed` but I'm not sure. Either way, the header now uses `position: sticky;` so we shouldn't need that 1px adjustment.
This PR introduced no visual changes except that it fixes the jitter mentioned above.
The idea behind this refactor is to centralise all of the user ignoring / muting / disallow PM checks in a single place, so they can be used consistently in core as well as for plugins like chat, while improving the main bulk of the checks to run in a single fast non-AR query.
Also fixed up the invite error when someone is muting/ignoring the user that is trying to invite them to the topic.
- following c3fd91670e `paste` has been typoed into `pase`
- adds two tests for pasting in `multi-select` and `email-group-user-chooser`
- selectKitOptions would not be following the right overriding order
- `category-selector` was using `selectKitOptions` directly which shouldn't be the case as it's not using computed values
- apparently since a recent ember upgrade, paste event is not providing `originalEvent` anymore and `clipboardData` should be retrieved directly on the event
Adds BEM-style classes to the input-group elements, and a dynamic class to the create-account-form element that reflects which auth provider is being used (if any).
Similar to e15c6302, overriding an auto-injected value like this triggers errors in more recent versions of Ember. Instead, we can use the registry to inject the value we need.
Co-authored-by: Peter Wagenet <peter.wagenet@gmail.com>
All our development-mode assets serve a `Cache-Control: no-cache` header, so a query parameter shouldn't be needed. Ember CLI does not include cache-busting parameters, so this change will move the development rails app to the same behaviour.
This will fix adding persistent breakpoints in the dev tools. Previously, the browser would think that the assets have been replaced and throw away the breakpoints.
Prior to v89, Firefox has bugs with document.execCommand("insertText"): https://bugzil.la/1220696
This commit introduces some variables to browser-detect, and therefore wraps the entire logic in an IIFE to avoid state leaking. (`let`/`const` are not supported on older browsers)
This manual assignment was added before the keyValueStore was refactored into a service. Now that it's a service, it gets all our standard auto-injections, including the keyValueStore.
Overwriting an automatic injection like this raises an error in future Ember versions.
This is useful when debugging exceptions - when the box is checked, exceptions will be thrown and handled by the browser rather than being absorbed by qunit. This allows developers to intercept the exception and inspect the stack.
This update covers commits:
* e309b6d [BREAKING Make JS client throw if lastId not number](e309b6d533)
* f0bae69 [DEV: removes dead code](f0bae695b0)
* a72b930 [FIX: force a poll more consistently when visibility changes](a72b9308b4)
* 5c01715 [Permit CORS preflight caching](5c01715432)
* 1789784 [DEV: lint files](17897843b4)
* b9cfb90 [FIX: do not leak visibility event subscriptions on stop/start](b9cfb90dd6)
* Use QUnit `module` instead of `discourseModule`
* Use QUnit `test` instead of `componentTest`
* Use angle-bracket syntax
* Remove jQuery usage
* Improve assertions (and actually fix some of them)
Currently we only apply watched words of the `Block` type to custom user
fields and user profile fields.
This patch enables all rules to be applied such as `Censor` or
`Replace`.
Previously we would only expand the main `<noscript>` element for older browsers. This commit ensures that we expand all noscript elements on the page, including the one used by the splashscreen to hide itself on the no-js view.
It ends up just triggering an error:
```
[Intervention] Blocked call to navigator.vibrate because user hasn't tapped on the frame or any embedded frame yet: https://www.chromestatus.com/feature/5644273861001216.
```
We don't count quote characters as part of the reply length.
We don't save drafts if the reply length is less than the min_post_length site setting.
If you start a reply that only contains a bunch of quotes with the intent to continue later, you get no draft.
This PR fixes that.
Note that we still don't save drafts if the composer is completely empty or if you're composing a new topic. This only affects replies.
This PR only changes the behavior if the reply composer contains something regardless of whether that something is a quote or not and ignores the min_post_length site setting.
The dots in the splash were previously hard-coded (v1). This PR makes progress towards making them be based on current theme colors.
Note that this is an improvement and not the "final" version. We're going to dynamically generate the splash file and the base64 URL later on.
All files under `app/assets/javascripts/scripts` will be run through babel, given sourcemaps, and output under `/assets/{name}.js`. This is another step towards removing our sprockets dependence.
The default Ember resolver implementation allows this for components. We need the same for connectors (which are essentially components behind-the-scenes)
This switches us to use the modern ember resolver package, and re-implements a number of our custom resolution rules within it. The legacy resolver remains for now, and is used as a fallback if the modern resolver is unable to resolve a package. When this happens, a warning will be printed to the console.
Co-authored-by: Peter Wagenet <peter.wagenet@gmail.com>
