Now that we run the `upload` method in different threads, we need to
synchronize writes to `STDOUT` which we can do so by using a `Logger`.
Follow-up to 49e8353959
The test was flaky and failing with the following errors:
```
Failure/Error:
klass
.connection
.select_raw(relation.arel) do |result, _|
result.type_map = DB.type_map
result.nfields == 1 ? result.column_values(0) : result.values
end
NoMethodError:
undefined method `select_raw' for nil
./lib/freedom_patches/fast_pluck.rb:60:in `pluck'
./vendor/bundle/ruby/3.3.0/gems/activerecord-7.2.2.1/lib/active_record/relation/calculations.rb:354:in `pick'
./app/models/web_crawler_request.rb:27:in `request_id'
./app/models/web_crawler_request.rb:31:in `rescue in request_id'
./app/models/web_crawler_request.rb:26:in `request_id'
./app/models/web_crawler_request.rb:19:in `write_cache!'
./app/models/concerns/cached_counting.rb:135:in `block (3 levels) in flush_to_db'
./vendor/bundle/ruby/3.3.0/gems/rails_multisite-6.1.0/lib/rails_multisite/connection_management/null_instance.rb:49:in `with_connection'
./vendor/bundle/ruby/3.3.0/gems/rails_multisite-6.1.0/lib/rails_multisite/connection_management.rb:21:in `with_connection'
./app/models/concerns/cached_counting.rb:134:in `block (2 levels) in flush_to_db'
./app/models/concerns/cached_counting.rb:124:in `each'
./app/models/concerns/cached_counting.rb:124:in `block in flush_to_db'
./lib/distributed_mutex.rb:53:in `block in synchronize'
./lib/distributed_mutex.rb:49:in `synchronize'
./lib/distributed_mutex.rb:49:in `synchronize'
./lib/distributed_mutex.rb:34:in `synchronize'
./app/models/concerns/cached_counting.rb:120:in `flush_to_db'
./app/models/concerns/cached_counting.rb:187:in `perform_increment!'
./app/models/web_crawler_request.rb:15:in `increment!'
./lib/middleware/request_tracker.rb:74:in `log_request'
./lib/middleware/request_tracker.rb:409:in `block in log_later'
./lib/scheduler/defer.rb:125:in `block in do_work'
./vendor/bundle/ruby/3.3.0/gems/rails_multisite-6.1.0/lib/rails_multisite/connection_management/null_instance.rb:49:in `with_connection'
./vendor/bundle/ruby/3.3.0/gems/rails_multisite-6.1.0/lib/rails_multisite/connection_management.rb:21:in `with_connection'
./lib/scheduler/defer.rb:119:in `do_work'
./lib/scheduler/defer.rb:105:in `block (2 levels) in start_thread'
```
This was due to running the defer thread in an async manner which is
actually no representative of the production environment. It also
revealed a spot in our code base where writes are happening in a GET
request which can cause requests to fail if ActiveRecord is in readonly
mode.
This reverts commit 766ff723f8.
Ensure that we create the sidekiq log file first before opening it for
logging. This avoids any issue of the log file not being present when we
initialize an instance of the `Logger`.
We identify and deny blocked crawlers here in anonymous_cache.
Separating the notion of the crawler identifier here lets plugins perform an
override if they perform more advanced detection.
We've seen in some communities abuse of user profile where bios and other fields are used in malicious ways, such as malware distribution. A common pattern between all the abuse cases we've seen is that the malicious actors tend to have 0 posts and have a low trust level.
To eliminate this abuse vector, or at least make it much less effective, we're making the following changes to user profiles:
1. Anonymous, TL0 and TL1 users cannot see any user profiles for users with 0 posts except for staff users
2. Anonymous and TL0 users can only see profiles of TL1 users and above
Users can always see their own profile, and they can still hide their profiles via the "Hide my public profile" preference. Staff can always see any user's profile.
Internal topic: t/142853.
Followup c7e471d35a
It is currently possible to add a bundle (which is a collection
of actions used for a dropdown on the client) for a reviewable
via actions.add_bundle and then never add any actions to it.
This causes the client to explode, as seen in the referenced
commit, because of the way our store expects to resolve objects
referenced by ID that are passed down by the serializer, which
then causes Ember to have an unrecoverable render error.
Fixing this on the serializer level is not really possible because
of all the ActiveModel::Serializer magic that serializes
objects by ID reference when doing things like has_many.
`Reviewable#actions_for` is a better place to do this anyway,
because this is the main location where the bundles and actions
are built for every action via the serializer.
Currently only system flags are translated. When we send message to the user that their post was deleted because of custom flag, we should default to custom flag name.
* DEV: unsilence deprecation warnings for old Font Awesome icon names
* update fa-user to user font awesome icon name
* update pencil-alt to pencil font awesome 6 icon name
In order to limit issues with duplicate inline CSS definitions, this will now deduplicate inline CSS styles with the "last-to-be-defined-wins" strategy.
Also removes unecessary whitespaces in inline styles.
Context - https://meta.discourse.org/t/resolve-final-styles-in-email-notifications/310219
Co-authored-by: Thomas Kalka <thomas.kalka@gmail.com>
When serializing the `body_changes` in the `PostRevisionSerializer`, we create two diffs: one for the `cooked` and another one for the `raw` version of the post.
Inside `DiscourseDiff`, we generate both `html` and `markdown` diffs when we only need the `html` diffs for the `cooked` version of the post and the `markdown` diff for the `raw` version of the post.
This solves the issue repored in https://meta.discourse.org/t/server-error-accessing-topic-revisions-on-a-specific-topic/339185 where some revisions would return 500 because of a `ArgumentError : Attributes per element limit exceeded` exception when trying to generate the `html` diff on a very large `raw`.
In some cases in CI env, it seems the AR connection isn’t available and
the `ensure` block is executed. It’s calling `#verify!` on the
connection, so it can fail sometimes. This is probably why
`#clear_active_connections!` was failing too sometimes.
Here, we just check the connection is present before clearing the
connections.
Spec was flaky cause work could still be in pipeline after the defer
length is 0. Our length denotes the backlog, not the in progress
count.
This adds a mechanism for gracefully stopping the queue and avoids
wait_for callse
We already add the "delete user" and "delete and block user" options to the drop-down for potential spam, but we should do this for potentially illegal posts as well.
This is entirely based on the implementation for the potential spam one, including caching the status on the reviewable record.
Also note that just as for potential spam, the user must be "deletable" for the option to appear.
I also took the liberty to move the options in the drop-down to what I think is a more intuitive place. (Between delete post and suspend/silence user.)
Sometimes changes to "What's new?" feed items are made or the feed items are
removed altogether, and the polling interval to check for new features is 1 day.
This is quite long, so this commit introduces a "Check for updates"
button for admins to click on the "What's new?" page which will bust
the cache for the feed and check again at the new features endpoint.
This is limited to 5 times per minute to avoid rapid sending of
requests.
This fix handles the case where an In-Reply-To mail header
can contain multiple Message-IDs. We use this header to
try look up an EmailLog record to find the post to reply
to in the group email inbox flow.
Since the case where multiple In-Reply-To Message-IDs is
rare (we've only seen a couple of instances of this causing
errors in the wild), we are just going to use the first one
in the array.
Also, Discourse does not support replying to multiple posts
at once, so it doesn't really make sense to use multiple
In-Reply-To Message-IDs anyway.
* DEV: Gracefully handle `regex_replace` violations of column length constraints
This is a follow-up to the `remap` [refactor](9b0cfa99c5).
Similar to `remap`, the entire `regex_replace` operation fails if the new content exceeds the column’s max length.
This change introduces an optional mode, controlled by the new `skip_max_length_violations` param
to skip records eligible for `regex_replace` where the new content violates the max column length constraint.
It also includes updates to the exception message raised when `regex_replace` fails to include more details
* DEV: Remove string escapes in heredoc text
Uploads that are linked to site settings shouldn't be flagged as secure in login-required sites that enable secure uploads. However, in order for site setting uploads to not be marked secured, the frontend uploader has to include 2 params in the upload request: `for_site_setting: true` and `type: "site_setting"`.
Since these 2 params are semantically identical, we want the `type: "site_setting"` param alone to make the upload correctly treated as a site setting upload. To achieve that, we need to include the `site_setting` type in the public types list because the `for_site_setting` param has the same effect — it marks the upload as a public type.
b138eaf9e5/lib/upload_security.rb (L128-L131)
Following a recent commit (cb4b8146a3),
the benchmark script wasn’t working anymore (and the related rake task).
This patch fixes it. It also adds some information about Ruby YJIT being
enabled or not.
This patch adds a new step to services named `try`.
It’s useful to rescue exceptions that some steps could raise. That way,
if an exception is caught, the service will stop its execution and can
be inspected like with any other steps.
Just wrap the steps that can raise with a `try` block:
```ruby
try do
step :step_that_can_raise
step :another_step_that_can_raise
end
```
By default, `try` will catch any exception inheriting from
`StandardError`, but we can specify what exceptions to catch:
```ruby
try(ArgumentError, RuntimeError) do
step :will_raise
end
```
An outcome matcher has been added: `on_exceptions`. By default it will
be executed for any exception caught by the `try` step.
Here also, we can specify what exceptions to catch:
```ruby
on_exceptions(ArgumentError, RuntimeError) do |exception|
…
end
```
Finally, an RSpec matcher has been added:
```ruby
it { is_expected.to fail_with_exception }
# or
it { is_expected.to fail_with_exception(ArgumentError) }
```
* DEV: Gracefully handle remaps which violate DB column constraints
This change implements length constraint enforcement to skip remaps
which exceed column max lengths
* DEV: Only perform skipped column stats lookup when verbose is true
* DEV: Tidy up specs
* DEV: Make skipping violating remap behaviour opt-in
This change introduces a new `skip_max_length_violations` param for
`remap`, set to `false` by default to ensure we still continue to fail
hard when max lenth constraints are violated.
To aid in quick resolution when remaps fail, this change also
adds more context to the exception message to include the offending table
and column information
* Apply suggestions from code review
Co-authored-by: Gerhard Schlager <gerhard.schlager@discourse.org>
* FIX: Various fixes
- Linter errors
- Remap status "logger" early return condition
---------
Co-authored-by: Gerhard Schlager <gerhard.schlager@discourse.org>
Firstly, we need to understand that ActiveRecord can be
connected to a role which prevent writes and this happens in Discourse when a
replica database has been setup for failover purposes. When a role
prevent writes from happening, ActiveRecord will raise the
`ActiveRecord::ReadOnlyError` if a write query is attempted.
Secondly, theme fields are baked at runtime within GET requests. The
baking process involves writing the baked value to the
`ThemeField#baked_value` column in the database.
If we combine the two points above, we can see how the writing of the
baked value to the database will trigger a `ActiveRecord::ReadOnlyError`
in a GET requests when the database is connected to a role preventing
writes. However, failing to bake a theme is not the end of the world and
should not cause GET requests to fail. Therefore, this commit adds a rescue
for `ActiveRecord::ReadOnlyError` in the `ThemeField#ensure_baked!`
method.
It splits the hide_profile_and_presence user option and the default_hide_profile_and_presence site setting for more granular control. It keeps the option to hide the profile under /u/username/preferences/interface and adds the presence toggle in the quick user menu.
Co-authored-by: Régis Hanol <regis@hanol.fr>
This reverts commit 5a00a041f1.
Implementation is currently not correct. Multiple uploads can share the
same etag but have different paths in the S3 bucket.
A "bad upload" in this context is a upload with a mismatched URL. This can happen when changing the S3 bucket used for uploads and the upload records in the database have not been remapped correctly.
When we added direct S3 uploads to Discourse, which use
presigned URLs, we never took into account the dualstack
endpoints for IPv6 on S3.
This commit fixes the issue by using the dualstack endpoints
for presigned URLs and requests, which are used in the
get-presigned-put and batch-presign-urls endpoints used when
directly uploading to S3.
It also makes regular S3 requests for `put` and so on use
dualstack URLs. It doesn't seem like there is a downside to
doing this, but a bunch of specs needed to be updated to reflect this.
This PR adds a small visual change to the new feature item on the `/admin/whats-new` page. When features are marked with an experimental site setting, they should show an indication on the feature item that it is "Experimental"
This commit changes the uploads:secure_upload_analyse_and_update
and uploads:disable_secure_uploads to no longer rebake affected
posts inline. This just took way too long, and if the task stalled
you couldn't be sure if the rest of it completed.
Instead, we can update the baked_version of affected posts and
utilize our PeriodicalUpdates job to gradually rebake them. I added
warnings about increasing the site setting rebake_old_posts_count and
the global setting max_old_rebakes_per_15_minutes before doing this
as well.
For good measure, the affected post IDs are written to a JSON file too.
This commit contains two changes to how our site setting
keyword system works:
1. Crowdin, our translation provider, does not support YAML lists,
so we are changing site setting keywords in server.en.yml to
be pipe-separated (|)
2. It's unclear to translators what they are supposed to do with
aliases of site settings where the name has changed, e.g.
min_trust_level_for_here_mention. Instead of getting these as
keywords from the yml file, we can discern these from
SiteSettings::DeprecatedSettings automatically, and still use
them for client-side search
These changes should help improve the situation for translators.
The `categories_only_optimized` category page style has been introduced
in commit d37a0d401c. This commit makes
sure that style is enforced for users who can see over 1000 categories
in order to keep `/categories` page functional.
This commit removes the feature flag for the new /about page, enabling it for all sites, and removes the code for old the /about page.
Internal topic: t/140413.
We decided to make contracts immutable once their validations have run.
Indeed, it doesn’t make a lot of sense to modify a contract value
outside the contract itself.
If processing is needed, then it should happen inside the contract
itself.
Followup 30fdd7738e
Adds a new site setting and corresponding user preference
to disable smart lists. By default they are enabled, because
this is a better experience for most users. A small number of
users would prefer to not have this enabled.
Smart lists automatically append new items to each
list started in the composer when enter is pressed. If
enter is pressed on an empty list item, it is cleared.
This setting will be removed when the new composer is complete.
This patch replaces the parameters provided to a service through
`params` by the contract object.
That way, it allows better consistency when accessing input params. For
example, if you have a service without a contract, to access a
parameter, you need to use `params[:my_parameter]`. But with a contract,
you do this through `contract.my_parameter`. Now, with this patch,
you’ll be able to access it through `params.my_parameter` or
`params[:my_parameter]`.
Some methods have been added to the contract object to better mimic a
Hash. That way, when accessing/using `params`, you don’t have to think
too much about it:
- `params.my_key` is also accessible through `params[:my_key]`.
- `params.my_key = value` can also be done through `params[:my_key] =
value`.
- `#slice` and `#merge` are available.
- `#to_hash` has been implemented, so the contract object will be
automatically cast as a hash by Ruby depending on the context. For
example, with an AR model, you can do this: `user.update(**params)`.
Currently in services, we don’t make a distinction between input
parameters, options and dependencies.
This can lead to user input modifying the service behavior, whereas it
was not the developer intention.
This patch addresses the issue by changing how data is provided to
services:
- `params` is now used to hold all data coming from outside (typically
user input from a controller) and a contract will take its values from
`params`.
- `options` is a new key to provide options to a service. This typically
allows changing a service behavior at runtime. It is, of course,
totally optional.
- `dependencies` is actually anything else provided to the service (like
`guardian`) and available directly from the context object.
The `service_params` helper in controllers has been updated to reflect
those changes, so most of the existing services didn’t need specific
changes.
The options block has the same DSL as contracts, as it’s also based on
`ActiveModel`. There aren’t any validations, though. Here’s an example:
```ruby
options do
attribute :allow_changing_hidden, :boolean, default: false
end
```
And here’s an example of how to call a service with the new keys:
```ruby
MyService.call(params: { key1: value1, … }, options: { my_option: true }, guardian:, …)
```
Bug introduced in this PR https://github.com/discourse/discourse/pull/29244
When the experiment toggle button was introduced, new features did not look right when the toggle button was not available.
In addition, the plugin name can be an empty string. In that case, information about new features should be displayed.
Currently, when calling a service with its block form, a `#result`
method is automatically created on the caller object. Even if it never
clashed so far, this could happen.
This patch removes that method, and instead use a more classical way of
doing things: the result object is now provided as an argument to the
main block. This means if we need to access the result object in an
outcome block, it will be done like this from now on:
```ruby
MyService.call(params) do |result|
on_success do
# do something with the result object
do_something(result)
end
end
```
In the same vein, this patch introduces the ability to match keys from
the result object in the outcome blocks, like we already do with step
definitions in a service. For example:
```ruby
on_success do |model:, contract:|
do_something(model, contract)
end
```
Instead of
```ruby
on_success do
do_something(result.model, result.contract)
end
```
Database dumps sometimes reference functions in the `discourse_functions` schema. It's possible that some of these functions have been dropped in a newer version of Discourse. In that case, restoring an older backup will fail with a `ERROR: function discourse_functions.something_something() does not exist` error. The restore functionality contains a workaround for that problem, but it didn't work with functions created in plugin migrations.
This commit adds support for temporarily creating missing `discourse_functions` from plugins. And it adds a simple check if the DB migration file even contains the required `DROPPED_TABLES` or `DROPPED_COLUMNS` constant. We don't need to create an instance of the DB migration class unless one of those constants is used. This makes the restore slightly faster and works around a problem with migrations that execute without `up` or `down` methods (e.g. `BackfillChatChannelAndThreadLastMessageIdsPostMigrate`).
Toggle the button to enable the experimental site setting from "What's new" announcement.
The toggle button is displayed when:
- site setting exists and is boolean;
- potentially required plugin is enabled.
* FIX: participating users statistics...
... was (mis-)counting
- bots
- anonymous users
- suspended users
There's now a "valid_users" function that holds the AR query for valid users and which is used in all "users", "active_users", and "participating_users" queries.
Internal ref - t/138435
- Add concurrency when running on multisite clusters (default 10, configurable via THEME_UPDATE_CONCURRENCY env)
- Add a version cache for the duration of the rake task. This avoids duplicating work when many sites in the cluster have the same theme installed, and it is already up-to-date
- Updates output to be more concurrent friendly (all `puts`, no `print`)
* FEATURE: Create rake for db migration in plugins
before the dev-xp was clunky, we had to create a migration file in core and
move it to the plugin.
Now we automated this process, we still create the migration file in core
but the rake task will move it to the plugin.
the usage is:
```
rake plugin:generate_migration[plugin_name,migration_name,migration_args]
rake plugin:generate_migration[discourse-automation,add_group_id_to_automation_rule,"group_id:integer"]
```
* DEV: change rake to be a generator for plugin migrations
* DEV: trying to add extra class option to migration generator
* DEV: revert to have only `plugin_migration_generator`
* DEV: remove rake task for plugin migration creation
* DEV: remove migration_generator.rb
* DEV: remove if because options with `plugin_name` will always be true
We want to allow lightboxing of smaller images, even if they are below the minimum size for image thumbnail generation.
This change sets a minimum threshold of 100 x 100 pixels for triggering the lightbox.
---------
Co-authored-by: Régis Hanol <regis@hanol.fr>
This patch improves the custom `array` type available in contracts.
It’s now able to split strings on `|` on top of `,`, and to be more
consistent, it also tries to cast the resulting items to integers.
Theme modifiers can now be defined as theme settings, this allows for
site operators to override behavior of theme modifiers.
New syntax is:
```
{
...
"modifiers": {
"modifier_name": {
"type": "setting",
"value": "setting_name"
}
}
}
```
This also introduces a new theme modifier for serialize_post_user_badges. Name of badge must match the name of the badge in the badges table. The client-side is updated to load this new data from the post-stream serializer.
Co-authored-by: David Taylor <david@taylorhq.com>
Constants should always be only assigned once. The logical OR assignment
of a constant is a relic of the past before we used zeitwerk for
autoloading and had bugs where a file could be loaded twice resulting in
constant redefinition warnings.
- limits security key deletes to second factor keys
- also deletes backup codes (lingering backup codes break login flow entirely)
* Add spec for rake task to disable 2FA for a user
Currently, when the MessageFormat compiler fails on some translations,
we just have the raw output from the compiler in the logs and that’s not
always very helpful.
Now, when there is an error, we iterate over the translation keys and
try to compile them one by one. When we detect one that is failing, it’s
added to a list that is now outputted in the logs. That way, it’s easier
to know which keys are not properly translated, and the problems can be
addressed quicker.
---
The previous implementation of this patch had a bug: it wasn’t handling
locales with country/region code properly. So instead of iterating over
the problematic keys, it was raising an error.
If a plugin's JS fails to load for some reason, most commonly
ad blockers, the entire admin interface would break. This is because
we are adding links to the admin routes for plugins that define
them in the sidebar.
We have a fix for this already in the plugin list which shows a warning
to the admin. This fix just prevents the broken link from rendering
in the sidebar if the route is not valid.
* Add migrations to ensure password hash is synced across users & user_passwords
* Persist password-related data in user_passwords instead of users
* Merge User#expire_old_email_tokens with User#expire_tokens_if_password_changed
* Add post deploy migration to mark password-related columns from users table as read-only
* Refactored UserPassword#confirm_password? and changes required to accommodate hashing the password after validations
In our production environment, we have been seeing Sidekiq processes
getting stuck randomly when a USR1 signal is sent to the Unicorn master
process. We have not been able to identify the root cause of why the
Sidekiq process gets stuck. We however noticed that when the Unicorn
master process receives a USR1 signal, it will reopen the logs for the
Unicorn master process first before sending a USR1 signal for the
Unicorn worker processes to reopen the logs. We figured that we should
do the same for the Sidekiq process as well when a USR1 signal.
In this commit, we introduce an arbitrary delay of 1 second before we
the Sidekiq process reopens its log files so as to allow enough time for the Unicorn
master to finish reopening it logs first.
We also do not send reopen logs for the Sidekiq process if the `DISCOURSE_LOG_SIDEKIQ`
env is not present because there is no need to reopen any logs.
There have been too many flaky tests as a result of leaking state in
Redis so it is easier to resolve them by ensuring we flush Redis'
database.
Locally on my machine, calling `Discourse.redis.flushdb` takes around
0.1ms which means this change will have very little impact on test
runtimes.
While using `OpenStruct` is nice, it’s generally not a very good idea as
it usually leads to performance problems.
The `OpenStruct` source code even says basically to avoid it.
Since the context object is crucial in our services, this patch replaces
`OpenStruct` with a custom implementation instead.
When a post has some replies, and the user click on the button to show them, we would load ALL the replies. This could lead to DoS if there were a very large number of replies.
This adds support for pagination to these post replies.
Internal ref t/129773
FIX: Duplicated parent posts
DEV: Query refactor
XHR requests are handled differently by the application and the
responses do not have any preloaded data so the cache key needs to
differntiate between those requests.
Remove emoji.clear cache calls as data.js.es6.erb hasn't existed in a while.
Emoji data is now compiled separately via javascript rake tasks.
Skip db and redis precompilation when no db is present
Currently, when the MessageFormat compiler fails on some translations,
we just have the raw output from the compiler in the logs and that’s not
always very helpful.
Now, when there is an error, we iterate over the translation keys and
try to compile them one by one. When we detect one that is failing, it’s
added to a list that is now outputted in the logs. That way, it’s easier
to know which keys are not properly translated, and the problems can be
addressed quicker.
My theory is that there were nil entries (that we were filtering out) that then changed and we weren't resetting them properly.
(the failure no longer repro'd in 30 CI runs in this PR)
