Running this spec locally I was getting an error:
```
1) translate accelerator plugins loads plural rules from plugins
Failure/Error: DiscoursePluginRegistry.unregister_locale("foo")
NoMethodError:
undefined method `unregister_locale' for class DiscoursePluginRegistry
# ./spec/lib/freedom_patches/translate_accelerator_spec.rb:113:in `block (3 levels) in <main>'
Finished in 0.17998 seconds (files took 1.88 seconds to load)
1 example, 1 failure
Failed examples:
rspec ./spec/lib/freedom_patches/translate_accelerator_spec.rb:117 # translate accelerator plugins loads plural rules from plugins
```
On top of this, this spec was flakey, Im not sure this is going to fix
flakyness, but this seems like a good first step.
Previously when changing back to the same email (ie. change to new
email, then change back again) we can easily end up showing the
incorrect old email if changing back to the same email more than once.
Then passing the incorrect old email causes an error as it can't be found.
This happens because we only searched for a combination of `user_id` and
`new_email` which can load an older change from the database. By using
the current email address when finding / initializing the
`EmailChangeRequest` we can prevent this issue as we will be creating a
new entry rather than loading an outdated one.
There are instances of posts being deleted by system_user where the context is left blank in the staff action logs, leading to confusion about why exactly they have been deleted.
This change deprecates using the PostDestroyer as system_user without providing a context, and adds a context to all call sites currently missing it in core. Plugins to be done after this is merged.
Previously we were compiling core and theme CSS into two targets:
Desktop and Mobile. The majority of both files was the 'common' css.
This commit splits those common styles into their own targets so that
there is less duplication. This should improve compilation times + cache
reuse, as well as opening the door for experiments with
media-query-based mobile-modes.
The only functional change is that we can no longer use `@extend` to
copy 'common' rules in core to mobile/desktop. This is probably for the
best. Duplication and/or mixins are a more native-css pattern for this.
Plugins already have a common / mobile / desktop pattern, so are
unchanged by this commit.
- Move some data transforming into contracts.
- Add some missing specs.
- Use the `try` step.
- Improve the `model` step a bit by allowing to catch any exception,
and not only `ArgumentError`. We already had the mechanism to inspect
which exception was caught.
We are no longer using any of the transpilation/bundling features of
Sprockets. We only use it to serve assets in development, and then
collect & fingerprint them in production. This commit switches us to use
the more modern "Propshaft" gem for that functionality.
Propshaft is much simpler than Sprockets. Instead of taking a
combination of paths + "precompile" list, Propshaft simply assumes all
files in the configured directory are required in production. Previously
we had some base paths configured quite high in the directory structure,
and then only precompiled selected assets within the directory. That's
no longer possible, so this commit refactors those places (mostly
plugin-related) to use dedicated directories under
`app/assets/generated/`.
Another difference is that Propshaft applies asset digests in
development as well as production. This is great for caching & dev/prod
consistency, but does mean some small changes were required in tests.
We previously had some freedom-patches applied to Sprockets. Some of
those had to be ported across to Propshaft. We now have three patches:
1. Skip adding digest hashes to webpack-generated chunks (which are
already digested, and referred to from other js files)
2. Avoid raising errors for missing assets in test mode. We don't always
compile assets before running basic RSpec tests.
3. Maintain relative paths for sourcemap URLs, so that files don't need
to be recompiled depending on their CDN path
Significant refactors are made to the `assets.rake` and `s3.rake` tasks,
which rely on implementation details of Sprockets/Propshaft.
Previously all locale bundles would be built & compressed during
assets:precompile. For most sites, only one of these languages was
actually used, so this is fairly wasteful.
This commit moves the main locale bundle into the
ExtraLocalesController, which has recently undergone many improvements
to make it more efficient. This allows locale files to be bundled "just
in time" when they're first accessed.
Now that brotli level=6 is enabled for these assets in our nginx config,
this change should have no impact on the locale bundle size.
This commit adds
- `topic_localization` containing its topic, a locale, title, and
fancy_title
- `post_localization` containing its post, a locale, raw, cooked, the
associated post's version
- and also APIs to add them
Reviewer note: We may ask ourselves "why create separate models instead
of one that is generic?" but the different localizable models may be
vastly different. For example in posts we only have raw that we need to
translate, and topics we have only title, but for categories we have
name and description. Then, we may ask ourselves "why not create a
polymorphic one that takes in model and column name?" and then we end up
with the same thing as what we have currently which is custom fields
(which is a mess in itself). Also, when replacing the untranslated
content to the translated one, we may find it easier to just `join` +
`coalesce` on the dedicated table - it would be a much simpler query
than polymorphism.
- Load moment and moment-timezone-with-data via webpack import instead
of including copies in every locale file
- Fetch static files from node_modules instead of `vendor/`
This cuts the size of locale-specific JS files in half, since they no
longer include moment itself.
Currently, trust level is calculated with this formula:
`[granted_trust_level, previous_trust_level,
SiteSetting.default_trust_level].max`
When a user is invited, SiteSetting.default_invitee_trust_level should
be respected in that calculation.
This commit adds a `:sidekiq_job_error` event which is
triggered when an error is encountered while running a Sidekiq job. The
intent of this change is to complement the `:sidekq_job_ran` event which
can be used with the `:sidekiq_job_error` event to calculate the error
rate for each job.
Allow admins to edit user custom flags. Because changing
name/description will update name/description for old reviewables,
warning has to be displayed.
Still, system flags can never be edited or deleted (only disabled).
This makes two tweaks to the setup wizard:
1. Move the branding (site logo) step to after styling (look & feel).
2. Remove the corporate (your organization) step.
Add new advanced search filters for post types
- `in:bot` or `in:bots`: Filters for posts made by bot users (user_id <
0)
- `in:human` or `in:humans`: Filters for posts made by human users
(user_id >= 0)
- `in:whisper` or `in:whispers`: Filters for whisper posts (respects
permissions)
- `in:regular`: Filters for regular posts only
Currently when using `register_topic_preloader_associations`, we are
not able to specify a condition that is evaluated at runtime.
This commit allows specifying a condition and also keeps backward
compatibility.
```
register_topic_preloader_associations(:linked_topic) { true }
register_topic_preloader_associations({
first_post: [uploads]
})
```
When bulk removing tags, the tag topic count isn't updated. You have to wait for the daily "ensure consistency" job to run for it to update properly. This is causing downstream problems, like the inability to use the "remove unused tags" feature.
The counter updating is handled on destroy by the join table model TopicTag. But we are calling #delete_all to perform the bulk tag removal, which bypasses callbacks.
This changes from #delete_all to #destroy_all, which invokes callbacks and updates the counters.
On 1st April 2025, we start showing no icons if the icon name used is a
deprecated one and therefore no longer part of the svg set.
We'll continue showing the messages with the correct icon name to aid
correction of these names.
Console logging will now be done at an error level for such icons.
We retain the behaviour of raising an error for such icons in plugins
from svg_sprite.rb in test environments, but removed this from
icon-library.js as it's harder to test the actual expected behaviour of
returning the original icon now that it's not part of the deprecation
workflow. (sinon.stub doesn't work well here for `isTesting` - the
alternative would be to override the environment.js module with
`proxyquire`) In any case, once we remove the mapping logic, we won't be
raising errors in test environment either for this scenario.
These aren't technically needed in the output, since we inline all
styles. However, there's no harm in including them, and having them
visible will improve the developer experience when restyling emails.
Currently when using `register_topic_preloader_associations`, we are not
able to specify a condition that is evaluated at runtime.
This commit allows specifying a condition and also keeps backward
compatibility.
```
register_topic_preloader_associations({
association: :linked_topic, condition: ->(topic) { topic.custom_field.present? }
})
```
When a site has the `must_approve_users` setting enabled, new user data is stored on the Reviewable model, including username, email, and any other data that is entered during signup. If the user is rejected, that data is retained, without a clear path to deleting it.
In order to allow data that could be PII to be removed, without breaking Discourse's audit and logging trails, this change scrubs the PII from the relevant `ReviewableUser` and `UserHistory` objects, replacing that data with who scrubbed it, and why.
7d587937 introduced css property deduplication in emails to workaround
bugs in some email clients. However, this didn't account for
admin-supplied customizations which were overriding core rules using
`!important`. This commit ensures that the deduplicator will always
prioritise `!important` rules, just like a browser.
This is a hidden site setting which has never been publicized, and is
not recommended for use. If we decide to add a feature like this in
future as a visible site setting, it would need many more safeguards to
prevent misuse.
b1924c35 switched our compiler to use `@import` internally for scss
entrypoints. This logic also applied to `.css` files, but unfortunately
sass doesn't do anything with `@import` of CSS files, so they'll be left
intact all the way to the browser. Continue using the old concatenation
approach for them in the compiler.
Followup to b1924c352487ab2c85ae50af45c5b3e098589014
Previously we would prepend extra content to developer-authored files,
which means adding `@use` in some files would throw an error because
`@use` must be at the top of any compiled file.
Instead, we can ensure any developer-authored files are on the load
path, and then `@import` them into the synthetic entrypoint.
Plugin color_definitions stylesheets are an edge case here, and will
need to be handled separately (or... wait until we move to native css
relative-color syntax, then we can drop color-definition stylesheets
altogether)
The URLs returned by DiscourseConnect for the user's avatar, profile and
card backgrounds were not always correctly handling CDN.
This make use of the `GlobalPath.full_cdn_url` helper method which has
been battle-tested.
Ref - https://meta.discourse.org/t/-/356599
When multiple admins are working in the review queue, it's quite easy for two people to try and handle the same reviewable at the same time. This change addresses the two major situations where this can occur.
The `ReviewableClaimedTopic` model has been extended to allow the system to mark a reviewable as claimed as soon as the first moderator starts handling the reviewable, even when the `reviewable_claiming` setting is disabled. This ensures that reviewable actions with client-site activity (for example, `agree_and_suspend`) will lock the reviewable before another moderator starts working on it.
When someone handles handles a reviewable, we now use `MessageBus` to inform other moderators that it's changed. If any of the other moderator have that reviewable open (either individually, or on the list screen), it will automatically refresh that data.
Currently, the light version of mobile logo falls back to the desktop
version if the mobile version isn't set. It makes sense to have the same
fallback rule for the dark version as well, i.e. if there's no dark
mobile logo, use the dark desktop logo.
Internal topic: t/150316.
This PR renames a couple of settings related to anonymous mode:
1. `allow_anonymous_posting` → `allow_anonymous_mode`. This setting is
used as a switch for the entire anonymous mode feature, so it makes
sense to give it a generic name that better reflects what the setting
does.
2. `allow_anonymous_likes` → `allow_likes_in_anonymous_mode`. The new
name is clearer and will match a new setting that we'll add to allow
anonymous users to post in chat.
Internal topic: t/148088.
When hard deleting a first post by passing force_destroy: true as an option to PostDestroyer, the post/topic is correctly deleted, and the staff record is created, but the app then errors out.
This only happens on sites with a topic_destroyed webhook setup.
After deleting the record, we pass the topic's ID to TopicView, which then raises an error because it can not load it from the DB.
TopicView supports being initialized with either an ID or an already instantiated record. Since we still have the record in memory after deleting, we can pass that to TopicView.
Currently we allow for 2 theme screenshots to be specified,
with a lightweight spec to allow both a light and dark
version of the screenshot. However, we were not storing this
screenshot name anywhere, so we would not be able to use it
for light/dark switching.
This commit fixes that issue, and also does some general refactoring
around theme screenshots, and adds more tests.
This is a follow up to 6820622467ab3613e824f0cb6219def2a575bc1d.
This commit addresses migrations that uses `remove_index` with the
`if_exists: true` option to drop an existing index before creating an
index using the `concurrently` option.
This commit also reruns two migration which may have caused indexes to
be left in an `invalid` state.
### Reviewers Note
Plugin tests are failing due to
https://github.com/discourse/discourse-translator/pull/251
This commit updates `Migration::SafeMigrate` to protect against unsafe
ways of adding a Postgres index concurrently.
Per postgres documentation:
If a problem arises while scanning the table, such as a deadlock or a
uniqueness violation in a unique index,
the CREATE INDEX command will fail but leave behind an "invalid" index.
This index will be ignored for querying
purposes because it might be incomplete; however it will still consume
update overhead. The recommended recovery
method in such cases is to drop the index and try again to perform
CREATE INDEX CONCURRENTLY .
Therefore, the simplest way for us to ensure that migrations that create
indexes concurrently are idempotent is to follow postgres'
recommendation of dropping the index first before trying to create the
index concurrently.
Followup https://github.com/discourse/discourse/pull/31505/
When sending notification emails for system user responses for PMs,
we removed the part of the CTA where it says "to respond to xyz" in a
previous commit.
This commit takes it slightly further -- we now only show a "Visit
Topic"
or "Visit Message" button if the PM notification is from a system user,
it's a bit cleaner.
This commit also adds more in-depth tests, and refactors the message
builder a little.
