The behaviour of #TERM in search has been amended
1. We try category or subcategory slugs
2. We try tags
3. We try tag-groups
The term `hello #my-group` will search for all posts tagged with any of
the tags in the tag group `My Group`
Future work may be introducing a slug cache here or caching it in the table
but the assumption is that the number of tag groups will not be huge
This can cause unbound CPU usage in some cases, and excessive logging in other cases. This commit moves redis readonly information into the local process, but maintains the DistributedCache for postgres readonly state.
* DEV: Add a new way to run specs in parallel with better output
This commit:
1. adds a new executable, `bin/interleaved_rspec` which works much like
`rspec`, but runs the tests in parallel.
2. adds a rake task, `rake interleaved:spec` which runs the whole test
suite.
3. makes autospec use this new wrapper by default. You can disable this
by running `PARALLEL_SPEC=0 rake autospec`.
It works much like the `parallel_tests` gem (and relies on it), but
makes each subprocess use a machine-readable formatter and parses this
output in order to provide a better overall summary.
(It's called interleaved, because parallel was taken and naming is
hard).
* Make popen3 invocation safer
* Use FileUtils instead of shelling out
* DRY up reporter
* Moved summary logic into Reporter
* s/interleaved/turbo/g
* Move Reporter into its own file
* Moved run into its own class
* Moved Runner into its own file
* Move JsonRowsFormatter under TurboTests
* Join on threads at the end
* Acted on feedback from eviltrout
This fixes a condition where an intermittent db connection could cause
invalid site settings to be stored
It also removes a catch all we had.
Somewhere around Rails 5 `db:create` started wanting full environment
this is a problem for Discourse since it needs to boot up data from the
db.
This removes the catch all and surgically adds a db / redis bypass to
db:create task.
In sidekiq, jobs are run in multiple threads within the same process. `cd` affects the entire process, so can cause unexpected issues in other running jobs.
Sometimes we would like to create a base image without any DB access, this
assists in creating custom base images with custom plugins that already
includes `public/assets`
Following this change set you can run:
```
SPROCKETS_CONCURRENT=1 DONT_PRECOMPILE_CSS=1 SKIP_DB_AND_REDIS=1 RAILS_ENV=production bin/rake assets:precompile
```
Then it is straight forward to create a base image without needing a DB or
Redis.
This previously was a hot path in topic view. Avoids an expensive active
record operation and instead perform SQL directly which is far more
targeted and efficient
* Support private uploads in S3
* Use localStore for local avatars
* Add job to update private upload ACL on S3
* Test multisite paths
* update ACL for private uploads in migrate_to_s3 task
* Expose a new plugin outlet. Pass group model to the group-member-dropdown so it can be accessed by plugins
* Added controller tests for group custom fields. update custom fields when updating a group
We now show if a queued or flagged post is a reply to another when in
the review queue. It's especially helpful for queued posts where
normally they are linked to the topic where they are created, and you
have no context about the reply.
Note that this will only apply to new queued posts going forward.
Previously queued posts will not show the "in reply to"
During profiling looking up topic users popped up as a hot path, this
change more than halved the amount of work it does
It reduces object allocations and method calls and avoids repeate translation
of common terms
This adds support for DISCOURSE_ENABLE_PERFORMANCE_HTTP_HEADERS
when set to `true` this will turn on performance related headers
```text
X-Redis-Calls: 10 # number of redis calls
X-Redis-Time: 1.02 # redis time in seconds
X-Sql-Commands: 102 # number of SQL commands
X-Sql-Time: 1.02 # duration in SQL in seconds
X-Queue-Time: 1.01 # time the request sat in queue (depends on NGINX)
```
To get queue time NGINX must provide: HTTP_X_REQUEST_START
We do not recommend you enable this without thinking, it exposes information
about what your page is doing, usually you would only enable this if you
intend to strip off the headers further down the stream in a proxy
Previously as soon as any override was defined we would regress to the slow
path for locale lookups. Additionally if `raise: true` was specified which
rails likes to add in views we would bypass the cache
The new design manages to use the fast path for many more cases
Before the locking here was added, replying to a post and liking a post
at roughly the same time could cause a deadlock.
Liking a post grabs an update lock on the post and then on the topic (to
update like counts).
We now lock the replied to post before getting the topic lock so that we
can update the replied to post later without causing a deadlock.
This was a fairly serious regression on sites with large (mega) topics,
however it was limited to staff.
The issue here is the query was using filtered_post_ids which I'd
assumed was already windowed to the current page, when in fact it was
all the ids in the topic. This fix corrects it by using the correct
windowed collection.
You can now add javascript files under `/javascripts/*` in a theme, and they will be loaded as if they were included in core, or a plugin. If you give something the same name as a core/plugin file, it will be overridden. Support file extensions are `.js.es6`, `.hbs` and `.raw.hbs`.
Previously theme setting descriptions were defined in the `settings.yml` file like this:
```
setting_name:
default: "My Default Value"
description:
en: "English description"
fr: "French description"
```
This commit allows developers to store the localised descriptions in the theme locale files instead:
```
en:
theme_metadata:
description: Theme Description
settings:
setting_name: "The localised description for setting_name"
```
SSO uses a special param to username suggester that whitelists a username
due to previous work we amended our lookup logic and started ignoring this
whitelist.
The fix ensures we always respect it, and also improves on the original
implementation that forgot to normalize the username.
This also corrects FileHelper.download so it supports "follow_redirect"
correctly (it used to always follow 1 redirect) and adds a `validate_url`
param that will bypass all uri validation if set to false (default is true)
This new `DISCOURSE_MAXMIND_BACKUP_PATH` can be used a secondary location
for maxmind db. That way a build machine, for example can cache it on the
host and reuse between builds.
Also per 5bfeef77 added proper error raising for download fails from
dedicated rake task
This also moves "refresh_maxmind_db_during_precompile_days" to a global
setting, it did not make sense in a site setting
* FEATURE: detect theme errors and catch them
* Bump COMPILER_VERSION
* Feedback
* Override eslint no console for one line
* Can't use our ajax method
* remove emoji from translation file
`rake posts:recover_uploads_from_index`
Searches through all missing uploads in the cluster, if it finds one it
tries to find it in the "upload index file" and creates a new upload for
it.
Previously we were only catching one type of data export, the new job will
catch every csv export we have
Job is pretty safe as it filters on system user id / pm with a particular
slug
Historically we would keep the user data export posts around but delete
the uploads.
This leaves a lot of broken uploads in the system.
This rake task allows us to clean up old mess.
Filename on disk may mismatch sha of file in some old 1X setups. This will
attempt to recover file even if sha1 mismatches. We had an old bug that
caused this.
This also adds `uploads:fix_relative_upload_links` which attempts to replace
urls of the format `/upload/default/...` with `upload://`
Rebaking posts can be expensive instead of blocking here simply mark posts
for rebake.
We can then work through them faster in other jobs, plus this should not
hold of a datacenter migration.
Previously this rake job would only run on a single site which is a bit
misleading
This also adds `VERBOSE=1 rake posts:missing_uploads` that will provide a
full report of missing uploads
This allows you to wait up to N seconds for the smoke test url to come up
in some cases you want to kick off the smoke test prior to having the smoke
test env ready to accept connections
Net::HTTP always returns ASCII-8BIT encoding. File.read auto-detects the encoding. This leads to an encoding inconsistency between a fresh download, and a cached download. This commit ensures all downloaded files are treated equally, by always returning the cached version from the filesystem, even during initial download.
One symptom of this problem is during theme exports: https://meta.discourse.org/t/116907
Related ruby ticket: https://bugs.ruby-lang.org/issues/2567
Having different behavior for staff and regular users can make it confusing for admins to understand how their configuration changes affect regular users
Previously username suggester would give up after 100 attempts at getting
a username and fallback to random string.
This amends the logic so we do all the work of figuring out a good username
in SQL and avoids a large amount of queries in cases where a lot of usernames
were used up.
This corrects an issue on sites with large numbers of anon users
Use the cooked version of the post and the quote to compare their content in
order to take into account the "typographer" option of the markdown pipeline.
We were blocking user registrations with same username and password,
but allowing usernames to be changed to be same as password later.
Also disallow names to be the same as password.
* English shouldn't fallback to any other locale
* Calculate fallback for default locale if it isn't English (useful for en_US)
* Reuse the fallback locale list when outputting translations to JavaScript
This reduces chances of errors where consumers of strings mutate inputs
and reduces memory usage of the app.
Test suite passes now, but there may be some stuff left, so we will run
a few sites on a branch prior to merging
The instagram onebox sometimes surrounds the image with an `<a>` tag, which was breaking the aspect ratio logic, and therefore causing posts to change height on load.
This is a feature that used to be present in discourse-assign but is
much easier to implement in core. It also allows a topic to be assigned
without it claiming for review and vice versa and allows it to work with
category group reviewers.
If creating a topic via the api as an admin and the category you specify
cannot be found an error will now be returned instead of just creating
the topic with no category. This will prevent accidental public topic
creation originally intended for a private category.
This commit is follow up to 535c594891 and
still allows for the creation of topics where the category param is
blank.
`#find` raises an error if the id given to it is invalid. As a result,
the conditional to check whether a `group` or `badge` is `present?` will
not be executed if any of the ids are invalid.
Follow up to
6ba914033c.
If creating a topic via the api as an admin and the category you specify
cannot be found an error will now be returned instead of just creating
the topic with no category. This will prevent accidental public topic
creation originally intended for a private category.
Since 5bfe051e, Discourse user agents are marked as non-crawlers (to avoid accidental blacklisting). This makes sure pageviews for these agents are tracked as crawler hits.
#b9d82818 makes enormous improvements to our bootstrap time, however going
to still keep compress for now despite the cost and watch it for a few weeks
* Do not brotli all locales in precompile
* Try without gzip
* uglify without compressing, always gzip
* skip uglify for unused locales
* FIX: Uglifier needs harmony for ES6 compatibility
* Use node uglifier if available
* Minor refactor
We found score hard to understand. It is still there behind the scenes
for sorting purposes, but it is no longer shown.
You can now filter by minimum priority (low, med, high) instead of
score.
No point moving all optimized image files to tombstone when the store is
changing. Also, `destroy_all` can easily blow memory since we are no
loading in batches.
This removes all uses of both `send` and `public_send` from consumers of
SiteSetting and instead introduces a `get` helper for dynamic lookup
This leads to much cleaner and safer code long term as we are always explicit
to test that a site setting is really there before sending an arbitrary
string to the class
It also removes a couple of risky stubs from the auth provider test
This change shows a notification number besides the flag icon in the
post menu if there is reviewable content associated with the post.
Additionally, if there is pending stuff to review, the icon has a red
background.
We have also removed the list of links below a post with the flag
status. A reviewer is meant to click the number beside the flag icon to
view the flags. As a consequence of losing those links, we've removed
the ability to undo or ignore flags below a post.
After careful analysis of large data-sets it became apparent that avg_time
had no impact whatsoever on "best of" topic scoring. Calculating avg_time
was a very costly operation especially on large databases.
We have some longer term plans of introducing other weighting that is read
time based into our scoring for "best of" and "top" topics, but in the
interim to stop a large amount of work that is not achieving any value we
are removing the jobs.
Column removal will follow once we decide on a new replacement metric.
Previously autospec would not pick up save if you saved a plugin in a
symlinked path, this broke quite a few workflows
We now maintain a reverse map so we can correctly re-run specs in plugins
`Upload#url` is more likely and can change from time to time. When it
does changes, we don't want to have to look through multiple tables to
ensure that the URLs are all up to date. Instead, we simply associate
uploads properly to `UserProfile` so that it does not have to replicate
the URLs in the table.
Minor fixes to add Rails 6 support to Discourse, we now will boot
with RAILS_MASTER=1, all specs pass
Only one tiny deprecation left
Largest change was the way ActiveModel:Errors changed interface a
bit but there is a simple backwards compat way of working it
This change automatically resizes icons for various purposes. Admins can now upload `logo` and `logo_small`, and everything else will be auto-generated. Specific icons can still be uploaded separately if required.
## Core
- Adds an SiteIconManager module which manages automatic resizing and fallback
- Icons are looked up in the OptimizedImage table at runtime, and then cached in Redis. If the resized version is missing for some reason, then most icons will fall back to the original files. Some icons (e.g. PWA Manifest) will return `nil` (because an incorrectly sized icon is worse than a missing icon).
- `SiteSetting.site_large_icon_url` will return the optimized version, including any fallback. `SiteSetting.large_icon` continues to return the upload object. This means that (almost) no changes are required in core/plugins to support this new system.
- Icons are resized whenever a relevant site setting is changed, and during post-deploy migrations
## Wizard
- Allows `requiresRefresh` wizard steps to reload data via AJAX instead of a full page reload
- Add placeholders to the **icons** step of the wizard, which automatically update from the "Square Logo"
- Various copy updates to support the changes
- Remove the "upload-time" resizing for `large_icon`. This is no longer required.
## Site Settings UX
- Move logo/icon settings under a new "Branding" tab
- Various copy changes to support the changes
- Adds placeholder support to the `image-uploader` component
- Automatically reloads site settings after saving. This allows setting placeholders to change based on changes to other settings
- Upload site settings will be assigned a placeholder if SiteIconManager `responds_to?` an icon of the same name
## Dashboard Warnings
- Remove PWA icon and PWA title warnings. Both are now handled automatically.
## Bonus
- Updated the sketch logos to use @awesomerobot's new high-res designs
We had quite a few cases in core where inputs are being mutated as a side
effect of calling a method.
This handles all the cases where specs caught this.
Mutating inputs makes code harder to reason about. Eg:
```
frog = "frog"
jump(frog)
puts frog
"fly" # ?????
```
This commit is part of a followup commit that adds # frozen_string_literal
to all our specs.
Previous behaviour was to silently remove tags that
belonged to a group with a parent tag that was missing.
The "required parent tag" feature is meant to guide people
to use the correct tags and show scoped results in the tag
input field, and to help create topic lists of related
tags. It isn't meant to be a strict requirement in the
composer that should trigger errors or restrictions.
* DEV: Replace site_setting_saved DiscourseEvent with site_setting_changed
site_setting_saved is confusing for a few reasons:
- It is attached to the after_save of the ActiveRecord model. This is confusing because it only works 'properly' with the db_provider
- It passes the activerecord model as a parameter, which is confusing because you get access to the 'database' version of the setting, rather than the ruby setting. For example, booleans appear as 'y' or 'n' strings.
- When the event is called, the local process cache has not yet been updated. So if you call SiteSetting.setting_name inside the event handler, you will receive the old site setting value
I have deprecated that event, and added a new site_setting_changed event. It passes three parameters:
- Setting name (symbol)
- Old value (in ruby format)
- New value (in ruby format)
It is triggered after the setting has been persisted, and the local process cache has been updated.
This commit also includes a test case which describes the confusing behavior. This can be removed once site_setting_saved is removed.
Since Rails 5.2, the behavior of `attribute_changed?` inside `after_save` callbacks has changed, so we need to use `saved_change_to_attribute` instead. The site setting local_process_provider in test mode was covering up the issue.
If you turn it on now, default all users to approved since they were
previously. Also support approving a user that doesn't have a reviewable
record (it will be created first.)
This also includes a refactor to move class method calls to
`DiscourseEvent` into an initializer. Otherwise the load order of
classes makes a difference in the test environment and some settings
might be triggered and others not, randomly.
This new site setting determines the maximum age of unread topics in
suggested. By default if you have any unread topics older than 90 days
they will be omitted from suggested.
This change was added for 2 reasons:
1. A performance safeguard, some users tend to collect a huge amount of
read state so it becomes super expensive to find unread
2. People who collect a large amount of unread are much more interested in
recent unread topics vs ancient unread topics, this makes suggested more
relevant
Also, this is a minor speed up for tests cause 3 expensive tests became 1.
Theme developers can include any number of scss files within the /scss/ directory of a theme. These can then be imported from the main common/desktop/mobile scss.
* FIX: correctly retrieve 'login required' setting value on wizard
FEATURE: extract 'invite only' setting in a separate checkbox control
* Update invite_only checkbox locale on wizard.
Co-Authored-By: techAPJ <arpit@techapj.com>
The compress brotli functionality is no longer optional, this has worked
well for years. The name of the ENV var is also confusing cause it does
not have a `DISCOURSE_` prefix which caused issues with the web upgrader
Brotli support is now unconditionally on
Previously jobs would fail silently in test mode. Now they will raise the exception and cause the relevant test to fail. This identified a few broken tests, which I will fix in a followup commit
- Plugin developers using OpenID2.0 should migrate to OAuth2 or OIDC. OpenID2.0 APIs will be removed in v2.4.0
- For sites requiring Yahoo login, it can be implemented using the OpenID Connect plugin: https://meta.discourse.org/t/103632
For more information, see https://meta.discourse.org/t/113249
Previously we tried using a sub-query which has some terrible performance
implications, by adding 1 extra query we can eliminate the PG issue
A join to user_stats also is prone to the same slowdown
This optimisation avoids large scans joining the topics table with the
topic_users table.
Previously when a user carried a lot of read state we would have to join
the entire read state with the topics table. This operation would slow down
home page and every topic page. The more read state you accumulated the
larger the impact.
The optimisation helps people who clean up unread, however if you carry
unread from years ago it will only have minimal impact.
A new checkbox has been added to the Tags tab of the category settings modal
which is used when some tags and/or tag groups are restricted to the category,
and all other unrestricted tags should also be allowed.
Default is the same as the previous behaviour: only allow the specified set of
tags and tag groups in the category.
Bianca Nenciu
Josh Moore
Penar Musaraj
Daniel Waterworth
Sam Saffron
Régis Hanol
Arpit Jalan
Guo Xiang Tan
David Taylor
Joffrey JAFFEUX
Robin Ward
Neil Lalonde
Vinoth Kannan
Dan Ungureanu
Gerhard Schlager
Roman Rizzi
Maja Komel
Rafael dos Santos Silva
Andrew Schleifer
Bianca Nenciu
Ralph Rooding
Blake Erickson
Michael Brown
Michael Brown
Osama Sayegh
Tim Lange
Tim Lange
Vinoth Kannan
Jeff Wong
Tarek Khalil
Erick Guan
Dax74