This PR introduces a new `enable_experimental_backup_uploads` site setting (default false and hidden), which when enabled alongside `enable_direct_s3_uploads` will allow for direct S3 multipart uploads of backup .tar.gz files.
To make multipart external uploads work with both the S3BackupStore and the S3Store, I've had to move several methods out of S3Store and into S3Helper, including:
* presigned_url
* create_multipart
* abort_multipart
* complete_multipart
* presign_multipart_part
* list_multipart_parts
Then, S3Store and S3BackupStore either delegate directly to S3Helper or have their own special methods to call S3Helper for these methods. FileStore.temporary_upload_path has also removed its dependence on upload_path, and can now be used interchangeably between the stores. A similar change was made in the frontend as well, moving the multipart related JS code out of ComposerUppyUpload and into a mixin of its own, so it can also be used by UppyUploadMixin.
Some changes to ExternalUploadManager had to be made here as well. The backup direct uploads do not need an Upload record made for them in the database, so they can be moved to their final S3 resting place when completing the multipart upload.
This changeset is not perfect; it introduces some special cases in UploadController to handle backups that was previously in BackupController, because UploadController is where the multipart routes are located. A subsequent pull request will pull these routes into a module or some other sharing pattern, along with hooks, so the backup controller and the upload controller (and any future controllers that may need them) can include these routes in a nicer way.
PERF: Update like count in visible posts without an extra GET per like
Currently when a user is reading a topic and some post in it receive a
like from another user, the Ember app will be notified via MessageBus
and issue a GET to `/posts/{id}` to get the new like count. This worked
fine for us until today, but it can easily create a self-inflicted DDoS
when a topic with a large number of visitors gets a large number of
likes, since we will issue `visitors * likes` GET requests requests.
This patch optimizes this flow, by sending the new like count down in
the MessageBus notification, removing any need for the extra request.
It shouldn't cause any drift on the count because we send down the full
count instead of the difference too.
Possible follow-ups could include handling like removal.
We are only using list_multipart_parts right now in the
uploads controller for multipart uploads to check if the
upload exists; thus we don't need up to 1000 parts.
Also adding a note for future explorers that list_multipart_parts
only gets 1000 parts max, and adding params for max parts
and starting parts.
Support for invites alongside DiscourseConnect was added in 355d51af. This commit fixes the guardian method so that the bulk invite button functionality also works.
* FIX: Preserve field types when updating revision
When a post was edited quickly twice by the same user, the old post
revision was updated with the newest changes. To check if the change
was reverted (i.e. rename topic A to B and then back to A) a comparison
of the initial value and last value is performed. If the check passes
then the intermediary value is dismissed and only the initial value and
the last ones are preserved. Otherwise, the modification is dismissed
because the field returned to its initial value.
This used to work well for most fields, but failed for "tags" because
the field is an array and the values were transformed to strings to
perform the comparison.
* FIX: Reset last_editor_id if revision is reverted
If a post was revised and then the same revision was reverted,
last_editor_id was still set to the ID of the user who last edited the
post. This was a problem because the same person could then edit the
same post again and because it was the same user and same post, the
system attempted to update the last one (that did not exist anymore).
This commit introduces a new s3:ensure_cors_rules rake task
that is run as a prerequisite to s3:upload_assets. This rake
task calls out to the S3CorsRulesets class to ensure that
the 3 relevant sets of CORS rules are applied, depending on
site settings:
* assets
* direct S3 backups
* direct S3 uploads
This works for both Global S3 settings and Database S3 settings
(the latter set directly via SiteSetting).
As it is, only one rule can be applied, which is generally
the assets rule as it is called first. This commit changes
the ensure_cors! method to be able to apply new rules as
well as the existing ones.
This commit also slightly changes the existing rules to cover
direct S3 uploads via uppy, especially multipart, which requires
some more headers.
FinalDestination's follow_canonical mode used for embedded topics should work when canonical URLs are relative, as specified in [RFC 6596](https://datatracker.ietf.org/doc/html/rfc6596)
We are no longer able to display the image returned by Instagram directly within a Discourse site (either in the composer, or within a cooked post within a topic), so:
- Display an image placeholder in the composer preview
- A cooked post should use an iframe to display the Instagram 'embed' content
The `白名单` term becomes `名单 白名单` after it is processed by
cppjieba in :query mode. However, `白名单` is not tokenized as such by cppjieba when it
appears in a string of text. Therefore, this may lead to failed matches as
the search data generated while indexing may not contain all of the
terms generated by :query mode. We've decided to maintain parity for now
such that both indexing and querying uses the same :mix mode. This may
lead to less accurate search but our plan is to properly support CJK
search in the future.
In preparation for adding automatic CORS rules creation
for direct S3 uploads, I am adding tests here and moving the
CORS rule definitions into a dedicated class so they are all
in the one place.
There is a problem with ensure_cors! as well -- if there is
already a CORS rule defined (presumably the asset one) then
we do nothing and do not apply the new rule. This means that
the S3BackupStore.ensure_cors method does nothing right now
if the assets rule is already defined, and it will mean the
same for any direct S3 upload rules I add for uppy. We need
to be able to add more rules, not just one.
This is not a problem on our hosting because we define the
rules at an infra level.
If the initial `get`/`head` response within `resolve` returns a status code of `103`, attempt to fetch the same URL with the alternative `small_get` method.
* FIX: allowed_theme_ids should not be persisted in GlobalSettings
It was observed that the memoized value of `GlobalSetting.allowed_theme_ids` would be persisted across requests, which could lead to unpredictable/undesired behaviours in a multisite environment.
This change moves that logic out of GlobalSettings so that the returned theme IDs are correct for the current site.
Uses get_set_cache, which ultimately uses DistributedCache, which will take care of multisite issues for us.
* DEV: Sanitize HTML admin inputs
This PR adds on-save HTML sanitization for:
Client site settings
translation overrides
badges descriptions
user fields descriptions
I used Rails's SafeListSanitizer, which [accepts the following HTML tags and attributes](018cf54073/lib/rails/html/sanitizer.rb (L108))
* Make sure that the sanitization logic doesn't corrupt settings with special characters
This PR doesn't change any behavior, but just removes code that wasn't in use. This is a pretty dangerous place to change, since it gets called during user's registration. At the same time the refactoring is very straightforward, it's clear that this code wasn't doing any work (it still needs to be double-checked during review though). Also, the test coverage of UserNameSuggester is good.
* PERF: Remove JOIN on categories for PM search
JOIN on categories is not needed when searchin in private messages as
PMs are not categorized.
* DEV: Use == for string comparison
* PERF: Optimize query for allowed topic groups
There was a query that checked for all topics a user or their groups
were allowed to see. This used UNION between topic_allowed_users and
topic_allowed_groups which was very inefficient. That was replaced with
a OR condition that checks in either tables more efficiently.
By default, Rails only includes the Vary:Accept header in responses when the Accept: header is included in the request. This means that proxies/browsers may cache a response to a request with a missing Accept header, and then later serve that cached version for a request which **does** supply the Accept header. This can lead to some very unexpected behavior in browsers.
This commit adds the Vary:Accept header for all requests, even if the Accept header is not present in the request. If a format parameter (e.g. `.json` suffix) is included in the path, then the Accept header is still omitted. (The format parameter takes precedence over any Accept: header, so the response is no longer varies based on the Accept header)
Searching within a topic currently does not make use of PG search and
we're simply doing an `ilike` against the post raw. Furthermore,
`Post#post_number` is already unique within a topic so the other
ordering will never ever be used. This change simply makes the query
cleaner to read.
This new app event will fire whenever a bookmark is created,
edited, or deleted for a post or topic, and replaces these old
app events which had inconsistent APIs:
* page:bookmark-post-toggled
* topic:bookmark-toggled
When the event is triggered, the arguments are in this order:
1. bookmark - The bookmark record created or changed. Will be null
if the bookmark was deleted.
2. target - Object with target (post or topic) and targetId (post ID
or topic ID)
User API keys (not the same thing as admin API keys) are currently
leaked to redis when rate limits are applied to them since redis is the
backend for rate limits in Discourse and the API keys are included in
the redis keys that are used to track usage of user API keys in the last
24 hours.
This commit stops the leak by using a SHA-256 representation of the user
API key instead of the key itself to form the redis key.
We don't need to manually delete the existing redis keys that contain
unhashed user API keys because they're not long-lived and will be
automatically deleted within 48 hours after this commit is deployed to
your Discourse instance.
This removes all custom controllers and redis/messagebus logic from discourse-presence, and replaces it with core's new PresenceChannel system.
All functionality should be retained. This implementation should scale much better to large numbers of users, reduce the number of HTTP requests made by clients, and reduce the volume of messages on the MessageBus.
For more information on PresenceChannel, see 31db8352
Instead of leaking ordering of the posts all around the class, we
centralize it in a method making the code easier to understand. In a
future PR, we will also introduce a plugin API to allow custom ordering
and the change in this commit helps to faciliate that.
The method was only used for mega topics but it was redundant as the
first post can be determined from using the condition where
`Post#post_number` equal to one.
`/u/username/invited.json?filter=expired` and `/u/username/invited.json?filter=pending` APIs are already returning data to admins. However, the `can_see_invite_details?` boolean was false, which prevented the Ember frontend from showing the tabs correctly. This commit updates the guardian method to match reality.
Sometimes administrators want to permanently delete posts and topics
from the database. To make sure that this is done for a good reasons,
administrators can do this only after one minute has passed since the
post was deleted or immediately if another administrator does it.
When a tag with alot of topics is used, we end up allocating a Ruby
array of all the topic ids. Instead, we can just use a subquery here and
handle all of the exclusion logic in PG.
Follow-up to ae13839f98
We don't want to be using emails as source for username and name suggestions in cases when it's possible that a user have no chance to intervene and correct a suggested username. It risks exposing email addresses.
* FIX: Remove List-Post email header
This header is used for mailing lists and can confuse some email clients
such as Thunderbird to display wrong replying options.
* FIX: Replace reply_key in email custom headers
Admins can add custom email headers from site settings. Email sender
will try to replace the reply key if %{reply_key} exists or remove the
header if a reply key does not exist.
Invite is used in two contexts, when inviting a new user to the forum
and when inviting an existent user to a topic. The first case is more
complex and it involves permission checks to ensure that new users can
be created. In the second case, it is enough to ensure that the topic
is visible for both users and that all preconditions are met.
One edge case is the invite to topic via email functionality which
checks for both conditions because first the user must be invited to
create an account first and then to the topic.
A side effect of these changes is that all site settings related to
invites refer to inviting new users only now.
We aren't translating these settings, so it makes more sense to move them into the code. I added an instance method so plugins can add mappings for custom reasons.
- Allow the `/presence/get` endpoint to return multiple channels in a single request (limited to 50)
- When multiple presence channels are initialized in a single Ember runloop, batch them into a single GET request
- Introduce the `presence-pretender` to allow easy testing of PresenceChannel-related features
- Introduce a `use_cache` boolean (default true) on the the server-side PresenceChannel initializer. Useful during testing.
This commit fixes the `eval` and `evalsha` commands/methods and any other methods that don't have a wrapper in `DiscourseRedis` and expect keyword arguments. I noticed this problem in Logster when I was trying to fetch some log messages in JSON format using the rails console and saw the messages were missing the `env` field. Logster uses the `eval` command to fetch messages `env`s:
dc351fd00f/lib/logster/redis_store.rb (L250-L253)
and that code was not fetching anything because `DiscourseRedis` didn't pass the `keys` keyword arg to the redis gem.
Usually, when an email is received a user lookup is performed using the
email address found in the `From` header. When an email has an
`X-Original-From` header, if it is equal to `Reply-To` then it uses that
one instead. The comparison was sensitive to whitespaces and other
insignificant characters such as quotes because it reconstructed the
`From` header.
For the fixture added in this commit, it compared the reconstructed
`From` header `John Doe <johndoe@example.com>` with the `Reply-To`
header `"John Doe" <johndoe@example.com>`.
* DEV: Remove HTML setting type and sanitization logic.
We concluded that we don't want settings to contain HTML, so I'm removing the setting type and sanitization logic. Additionally, we no longer allow the global-notice text to contain HTML.
I searched for usages of this setting type in the `all-the-plugins` repo and found none, so I haven't added a migration for existing settings.
* Mark Global notices containing links as HTML Safe.
FinalDestination now supports the `follow_canonical` option, which will perform an initial GET request, parse the canonical link if present, and perform a HEAD request to it.
We use this mode during embeds to avoid treating URLs with different query parameters as different topics.
Previously, while retrieving each upload urls in a post S3 CDN urls with different path in prefix (external urls technically) are considered as uploaded url. It created issue while checking missing uploads.
This commit resolves refactors can_invite_to? to use
can_invite_to_forum? for checking the site-wide permissions and then
perform topic specific checkups.
Similarly, can_invite_to? is always used with a topic object and this is
now enforced.
There was another problem before when `must_approve_users` site setting
was not checked when inviting users to forum, but was checked when
inviting to a topic.
Another minor security issue was that group owners could invite to
group topics even if they did not have the minimum trust level to do
it.
It was possible to see notifications of other users using routes:
- notifications/responses
- notifications/likes-received
- notifications/mentions
- notifications/edits
We weren't showing anything private (like notifications about private messages), only things that're publicly available in other places. But anyway, it feels strange that it's possible to look at notifications of someone else. Additionally, there is a risk that we can unintentionally leak something on these pages in the future.
This commit restricts these routes.
The all inboxes was introduced in
016efeadf6 but we decided to roll it back
for performance reasons. The main performance challenge here is that PG
has to basically loop through all the PMs that a user is allowed to view
before being able to order by `Topic#bumped_at`. The all inboxes was not
planned as part of the new/unread filter so we've decided not to tackle
the performance issue for the upcoming release.
Follow-up to 016efeadf6
* PERF: Improve database query perf when loading topics for a category.
Instead of left joining the `topics` table against `categories` by filtering with `categories.id`,
we can improve the query plan by filtering against `topics.category_id`
first before joining which helps to reduce the number of rows in the
topics table that has to be joined against the other tables and also
make better use of our existing index.
The following is a before and after of the query plan for a category
with many subcategories.
Before:
```
QUERY PLAN
-------------------------------------------------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------
Limit (cost=1.28..747.09 rows=30 width=12) (actual time=85.502..2453.727 rows=30 loops=1)
-> Nested Loop Left Join (cost=1.28..566518.36 rows=22788 width=12) (actual time=85.501..2453.722 rows=30 loops=1)
Join Filter: (category_users.category_id = topics.category_id)
Filter: ((topics.category_id = 11) OR (COALESCE(category_users.notification_level, 1) <> 0) OR (tu.notification_level > 1))
-> Nested Loop Left Join (cost=1.00..566001.58 rows=22866 width=20) (actual time=85.494..2453.702 rows=30 loops=1)
Filter: ((COALESCE(tu.notification_level, 1) > 0) AND ((topics.category_id <> 11) OR (topics.pinned_at IS NULL) OR ((t
opics.pinned_at <= tu.cleared_pinned_at) AND (tu.cleared_pinned_at IS NOT NULL))))
Rows Removed by Filter: 1
-> Nested Loop (cost=0.57..528561.75 rows=68606 width=24) (actual time=85.472..2453.562 rows=31 loops=1)
Join Filter: ((topics.category_id = categories.id) AND ((categories.topic_id <> topics.id) OR (categories.id = 1
1)))
Rows Removed by Join Filter: 13938306
-> Index Scan using index_topics_on_bumped_at on topics (cost=0.42..100480.05 rows=715549 width=24) (actual ti
me=0.010..633.015 rows=464623 loops=1)
Filter: ((deleted_at IS NULL) AND ((archetype)::text <> 'private_message'::text))
Rows Removed by Filter: 105321
-> Materialize (cost=0.14..36.04 rows=30 width=8) (actual time=0.000..0.002 rows=30 loops=464623)
-> Index Scan using categories_pkey on categories (cost=0.14..35.89 rows=30 width=8) (actual time=0.006.
.0.040 rows=30 loops=1)
Index Cond: (id = ANY ('{11,53,57,55,54,56,112,94,107,115,116,117,97,95,102,103,101,105,99,114,106,1
13,104,98,100,96,108,109,110,111}'::integer[]))
-> Index Scan using index_topic_users_on_topic_id_and_user_id on topic_users tu (cost=0.43..0.53 rows=1 width=16) (a
ctual time=0.004..0.004 rows=0 loops=31)
Index Cond: ((topic_id = topics.id) AND (user_id = 1103877))
-> Materialize (cost=0.28..2.30 rows=1 width=8) (actual time=0.000..0.000 rows=0 loops=30)
-> Index Scan using index_category_users_on_user_id_and_last_seen_at on category_users (cost=0.28..2.29 rows=1 width
=8) (actual time=0.004..0.004 rows=0 loops=1)
Index Cond: (user_id = 1103877)
Planning Time: 1.359 ms
Execution Time: 2453.765 ms
(23 rows)
```
After:
```
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Limit (cost=1.28..438.55 rows=30 width=12) (actual time=38.297..657.215 rows=30 loops=1)
-> Nested Loop Left Join (cost=1.28..195944.68 rows=13443 width=12) (actual time=38.296..657.211 rows=30 loops=1)
Filter: ((categories.topic_id <> topics.id) OR (topics.category_id = 11))
Rows Removed by Filter: 29
-> Nested Loop Left Join (cost=1.13..193462.59 rows=13443 width=16) (actual time=38.289..657.092 rows=59 loops=1)
Join Filter: (category_users.category_id = topics.category_id)
Filter: ((topics.category_id = 11) OR (COALESCE(category_users.notification_level, 1) <> 0) OR (tu.notification_level > 1))
-> Nested Loop Left Join (cost=0.85..193156.79 rows=13489 width=20) (actual time=38.282..657.059 rows=59 loops=1)
Filter: ((COALESCE(tu.notification_level, 1) > 0) AND ((topics.category_id <> 11) OR (topics.pinned_at IS NULL) OR ((topics.pinned_at <= tu.cleared_pinned_at) AND (tu.cleared_pinned_at IS NOT NULL))))
Rows Removed by Filter: 1
-> Index Scan using index_topics_on_bumped_at on topics (cost=0.42..134521.06 rows=40470 width=24) (actual time=38.267..656.850 rows=60 loops=1)
Filter: ((deleted_at IS NULL) AND ((archetype)::text <> 'private_message'::text) AND (category_id = ANY ('{11,53,57,55,54,56,112,94,107,115,116,117,97,95,102,103,101,105,99,114,106,113,104,98,100,96,108,109,110,111}'::integer[])))
Rows Removed by Filter: 569895
-> Index Scan using index_topic_users_on_topic_id_and_user_id on topic_users tu (cost=0.43..1.43 rows=1 width=16) (actual time=0.003..0.003 rows=0 loops=60)
Index Cond: ((topic_id = topics.id) AND (user_id = 1103877))
-> Materialize (cost=0.28..2.30 rows=1 width=8) (actual time=0.000..0.000 rows=0 loops=59)
-> Index Scan using index_category_users_on_user_id_and_last_seen_at on category_users (cost=0.28..2.29 rows=1 width=8) (actual time=0.004..0.004 rows=0 loops=1)
Index Cond: (user_id = 1103877)
-> Index Scan using categories_pkey on categories (cost=0.14..0.17 rows=1 width=8) (actual time=0.001..0.001 rows=1 loops=59)
Index Cond: (id = topics.category_id)
Planning Time: 1.633 ms
Execution Time: 657.255 ms
(22 rows)
```
* PERF: Optimize index on topics bumped_at.
Replace `index_topics_on_bumped_at` index with a partial index on `Topic#bumped_at` filtered by archetype since there is already another index that covers private topics.
The file size error messages for max_image_size_kb and
max_attachment_size_kb are shown to the user in the KB
format, regardless of how large the limit is. Since we
are going to support uploading much larger files soon,
this KB-based limit soon becomes unfriendly to the end
user.
For example, if the max attachment size is set to 512000
KB, this is what the user sees:
> Sorry, the file you are trying to upload is too big (maximum
size is 512000KB)
This makes the user do math. In almost all file explorers that
a regular user would be familiar width, the file size is shown
in a format based on the maximum increment (e.g. KB, MB, GB).
This commit changes the behaviour to output a humanized file size
instead of the raw KB. For the above example, it would now say:
> Sorry, the file you are trying to upload is too big (maximum
size is 512 MB)
This humanization also handles decimals, e.g. 1536KB = 1.5 MB
Instead of going to the OP of the topic for topic-level bookmarks
(which are bookmarks where for_topic is true) when clicking on the
bookmark in the quick access menu or on the user bookmark list,
this commit takes the user to the last unread post in
the topic instead. This should be generally more useful than landing
on the unchanging OP.
To make this work nicely, I needed to add the last_read_post_number to
the BookmarkQuery based on the TopicUser association. It should not add
too much extra weight to the query, because it is limited to the user
that we are fetching bookmarks for.
Also fixed an issue where the bookmark serializer highest_post_number was
not taking into account whether the user was staff, which is when we
should use highest_staff_post_number instead.
Allows creating a bookmark with the `for_topic` flag introduced in d1d2298a4c set to true. This happens when clicking on the Bookmark button in the topic footer when no other posts are bookmarked. In a later PR, when clicking on these topic-level bookmarks the user will be taken to the last unread post in the topic, not the OP. Only the OP can have a topic level bookmark, and users can also make a post-level bookmark on the OP of the topic.
I had to do some pretty heavy refactors because most of the bookmark code in the JS topics controller was centred around instances of Post JS models, but the topic level bookmark is not centred around a post. Some refactors were just for readability as well.
Also removes some missed reminderType code from the purge in 41e19adb0d
The display name can have quotes around it, which does not work
with our current comparison of a from field (in this case Reply-To)
and another header (X-Original-From), because we are not comparing
the two values in the same way. This causes an issue where the
commit here: b88d8c8 will not
work properly; the forwarded email gets the From address instead
of the Reply-To address as intended.
We don't actually use the reminder_type for bookmarks anywhere;
we are just storing it. It has no bearing on the UI. It used
to be relevant with the at_desktop bookmark reminders (see
fa572d3a7a)
This commit marks the column as readonly, ignores it, and removes
the index, and it will be dropped in a later PR. Some plugins
are relying on reminder_type partially so some stubs have been
left in place to avoid errors.
This partially reverts commit ddb458343d.
Seeing performance degrade on larger sites so back to drawing board on
this one. Instead of the DISTINCT LEFT JOIN, we switch back to
IN(subquery).
First reported in https://meta.discourse.org/t/-/202482/19
There are two optimizations being applied here:
1. Fetch a user's group ids in a seperate query instead of including it
as a sub-query. When I tried a subquery, the query plan becomes very
inefficient.
1. Join against the `topic_allowed_users` and `topic_allowed_groups`
table instead of doing an IN against a subquery where we UNION the
`topic_id`s from the two tables. From my profiling, this enables PG to
do a backwards index scan on the `index_topics_on_timestamps_private`
index.
This commit fixes a bug where listing all messages was incorrectly
excluding topics if a topic has been archived by a group even if the
user did not belong to the group.
This commit also fixes another bug where dismissing private messages
selectively was subjected to the default limit of 30.
We don't need no stinkin' denormalization! This commit ignores
the topic_id column on bookmarks, to be deleted at a later date.
We don't really need this column and it's better to rely on the
post.topic_id as the canonical topic_id for bookmarks, then we
don't need to remember to update both columns if the bookmarked
post moves to another topic.
When copying an existing upload stub temporary object
on S3 to its final destination we were not copying across
its additional headers such as content-disposition and
cache-control, which led to issues like attachments not
downloading with their original filename when clicking
the download links in posts.
This is because the metadata_directive = REPLACE option
was not being passed to object.copy_from(), so only the
source object's headers were being used. Added an option
for apply_metadata_to_destination to apply this option
conditionally, because we may not always want to replace
this metadata, but we definitely do when copying a temporary
upload.
When a user archives a personal message, they are redirected back to the
inbox and will refresh the list of the topics for the given filter.
Publishing an event to the user results in an incorrect incoming message
because the list of topics has already been refreshed.
This does mean that if a user has two tabs opened, the non-active tab
will not receive the incoming message but at this point we do not think
the technical trade-offs are worth it to support this feature. We
basically have to somehow exclude a client from an incoming message
which is not easy to do.
Follow-up to fc1fd1b416
In order to include the new/unread count in the browse more message
under suggested topics, a couple of technical changes have to be made.
1. `PrivateMessageTopicTrackingState` is now auto-injected which is
similar to how it is done for `TopicTrackingState`. This is done so
we don't have to attempt to pass the `PrivateMessageTopicTrackingState`
object multiple levels down into the suggested-topics component. While
the object is auto-injected, we only fetch the initial state and start
tracking when the relevant private messages routes has been hit and only
when a private message's suggested topics is loaded. This is
done as we do not want to add the extra overhead of fetching the inital
state to all page loads but instead wait till the private messages
routes are hit.
2. Previously, we would stop tracking once the `user-private-messages`
route has been deactivated. However, that is not ideal since
navigating out of the route and back means we send an API call to the
server each time. Since `PrivateMessageTopicTrackingState` is kept in
sync cheaply via messageBus, we can just continue to track the state
even if the user has navigated away from the relevant stages.
When forwarding emails into the group inbox, we now use the
original sender email as the from_address since
2ac9fd9dff. However, we have not
been saving the original CC addresses of the forwarded email,
which are needed to include those recipients in on the conversation
when replying via the group inbox.
This commit captures the CC addresses on the incoming email, and
makes sure the emails are created as staged users and added to the
list of topic allowed users so they are included on CC's sent by
the GroupSmtpEmail and other jobs.
When 2ac9fd9dff was done, this
affected the small post that is created when forwarding an email
into the group inbox. Instead of using the name and the email of
the user who forwarded the email, it used the original from email
and name to create the small post. So instead of something like
"Discourse Team forwarded the above email" we ended up with
"John Smith forwarded the above email" which is incorrect.
This fixes the issue by creating a staged user for the forwarding
email address (if such a user does not yet exist) and uses that
for the "forwarded" small post instead.
Other locale characters in file names (e.g. é, ä) as well
as special characters can cause issues on S3, notably the S3
copy object operation does not support these special characters.
Instead of storing the original file name in the key, which is
unnecessary, we now generate a random file name with the original
extension for the temporary file and use that for all external
upload stub operations.
The seed-fu gem resets the sequence on all the tables it touches. In some situations, this can cause primary keys to be re-used. This commit introduces a freedom patch which ensures seed-fu only touches the sequence when it is less than the id of one of the seeded records.
PresenceChannel aims to be a generic system for allow the server, and end-users, to track the number and identity of users performing a specific task on the site. For example, it might be used to track who is currently 'replying' to a specific topic, editing a specific wiki post, etc.
A few key pieces of information about the system:
- PresenceChannels are identified by a name of the format `/prefix/blah`, where `prefix` has been configured by some core/plugin implementation, and `blah` can be any string the implementation wants to use.
- Presence is a boolean thing - each user is either present, or not present. If a user has multiple clients 'present' in a channel, they will be deduplicated so that the user is only counted once
- Developers can configure the existence and configuration of channels 'just in time' using a callback. The result of this is cached for 2 minutes.
- Configuration of a channel can specify permissions in a similar way to MessageBus (public boolean, a list of allowed_user_ids, and a list of allowed_group_ids). A channel can also be placed in 'count_only' mode, where the identity of present users is not revealed to end-users.
- The backend implementation uses redis lua scripts, and is designed to scale well. In the future, hard limits may be introduced on the maximum number of users that can be present in a channel.
- Clients can enter/leave at will. If a client has not marked itself 'present' in the last 60 seconds, they will automatically 'leave' the channel. The JS implementation takes care of this regular check-in.
- On the client-side, PresenceChannel instances can be fetched from the `presence` ember service. Each PresenceChannel can be used entered/left/subscribed/unsubscribed, and the service will automatically deduplicate information before interacting with the server.
- When a client joins a PresenceChannel, the JS implementation will automatically make a GET request for the current channel state. To avoid this, the channel state can be serialized into one of your existing endpoints, and then passed to the `subscribe` method on the channel.
- The PresenceChannel JS object is an ember object. The `users` and `count` property can be used directly in ember templates, and in computed properties.
- It is important to make sure that you `unsubscribe()` and `leave()` any PresenceChannel objects after use
An example implementation may look something like this. On the server:
```ruby
register_presence_channel_prefix("site") do |channel|
next nil unless channel == "/site/online"
PresenceChannel::Config.new(public: true)
end
```
And on the client, a component could be implemented like this:
```javascript
import Component from "@ember/component";
import { inject as service } from "@ember/service";
export default Component.extend({
presence: service(),
init() {
this._super(...arguments);
this.set("presenceChannel", this.presence.getChannel("/site/online"));
},
didInsertElement() {
this.presenceChannel.enter();
this.presenceChannel.subscribe();
},
willDestroyElement() {
this.presenceChannel.leave();
this.presenceChannel.unsubscribe();
},
});
```
With this template:
```handlebars
Online: {{presenceChannel.count}}
<ul>
{{#each presenceChannel.users as |user|}}
<li>{{avatar user imageSize="tiny"}} {{user.username}}</li>
{{/each}}
</ul>
```
The generate_presigned_put endpoint for direct external uploads
(such as the one for the uppy-image-uploader) records allowed
S3 metadata values on the uploaded object. We use this to store
the sha1-checksum generated by the UppyChecksum plugin, for later
comparison in ExternalUploadManager.
However, we were not doing this for the create_multipart endpoint,
so the checksum was never captured and compared correctly.
Also includes a fix to make sure UppyChecksum is the last preprocessor to run.
It is important that the UppyChecksum preprocessor is the last one to
be added; the preprocessors are run in order and since other preprocessors
may modify the file (e.g. the UppyMediaOptimization one), we need to
checksum once we are sure the file data has "settled".
This bug was introduced by f66007ec83.
In PostJobsEnqueuer we previously did not fire the after_post_create
event and after_topic_create event for private message topics. This was
changed in the above commit in order to publish message bus messages
for topic tracking state updates. Unfortunately this caused the
NotifyMailingListSubscribers job to be enqueued for all posts including
private messages, and admins and the users involved in the PMs got
emailed the contents of the PMs if they had mailing list mode enabled.
Luckily the impact of this was mitigated by a Guardian#can_see? check
for each mailing list mode user in the NotifyMailingListSubscribers job.
We never want to notify mailing list mode subscribers for private messages
so an early return has been added there, plus the logic in PostJobsEnqueuer
has been fixed, and tests have been added to that class where there were
none before.
This is unnecessary, as when the temporary key is created
in S3Store we already include the s3_bucket_folder_path, and
the key will always start with temp/ to assist with lifecycle
rules for multipart uploads.
This was affecting Discourse.store.object_from_path,
Discourse.store.signed_url_for_path, and possibly others.
See also: e0102a5
This can be used to change the list of topic posters. For example,
discourse-solved can use this to move the user who posted the solution
after the original poster.
There are certain design decisions that were made in this commit.
Private messages implements its own version of topic tracking state because there are significant differences between regular and private_message topics. Regular topics have to track categories and tags while private messages do not. It is much easier to design the new topic tracking state if we maintain two different classes, instead of trying to mash this two worlds together.
One MessageBus channel per user and one MessageBus channel per group. This allows each user and each group to have their own channel backlog instead of having one global channel which requires the client to filter away unrelated messages.
Previously we had temp/ in the middle of the S3 key path like so
* /uploads/default/temp/randomstring/test.png (normal site)
* /sitename/uploads/default/temp/randomstring/test.png (s3 folder path site)
* /standard10/uploads/sitename/temp/randomstring/test.png (multisite site)
However this necessitates making a lifecycle rule to clean up incomplete
S3 multipart uploads for every site, something which we cannot do. It makes
much more sense to have a structure with /temp at the start of the key,
which is what this commit does:
* /temp/uploads/default/randomstring/test.png (normal site)
* /temp/sitename/uploads/default/randomstring/test.png (s3 folder path site)
* /temp/standard10/uploads/sitename/randomstring/test.png (multisite site)
This pull request introduces the endpoints required, and the JavaScript functionality in the `ComposerUppyUpload` mixin, for direct S3 multipart uploads. There are four new endpoints in the uploads controller:
* `create-multipart.json` - Creates the multipart upload in S3 along with an `ExternalUploadStub` record, storing information about the file in the same way as `generate-presigned-put.json` does for regular direct S3 uploads
* `batch-presign-multipart-parts.json` - Takes a list of part numbers and the unique identifier for an `ExternalUploadStub` record, and generates the presigned URLs for those parts if the multipart upload still exists and if the user has permission to access that upload
* `complete-multipart.json` - Completes the multipart upload in S3. Needs the full list of part numbers and their associated ETags which are returned when the part is uploaded to the presigned URL above. Only works if the user has permission to access the associated `ExternalUploadStub` record and the multipart upload still exists.
After we confirm the upload is complete in S3, we go through the regular `UploadCreator` flow, the same as `complete-external-upload.json`, and promote the temporary upload S3 into a full `Upload` record, moving it to its final destination.
* `abort-multipart.json` - Aborts the multipart upload on S3 and destroys the `ExternalUploadStub` record if the user has permission to access that upload.
Also added are a few new columns to `ExternalUploadStub`:
* multipart - Whether or not this is a multipart upload
* external_upload_identifier - The "upload ID" for an S3 multipart upload
* filesize - The size of the file when the `create-multipart.json` or `generate-presigned-put.json` is called. This is used for validation.
When the user completes a direct S3 upload, either regular or multipart, we take the `filesize` that was captured when the `ExternalUploadStub` was first created and compare it with the final `Content-Length` size of the file where it is stored in S3. Then, if the two do not match, we throw an error, delete the file on S3, and ban the user from uploading files for N (default 5) minutes. This would only happen if the user uploads a different file than what they first specified, or in the case of multipart uploads uploaded larger chunks than needed. This is done to prevent abuse of S3 storage by bad actors.
Also included in this PR is an update to vendor/uppy.js. This has been built locally from the latest uppy source at d613b849a6. This must be done so that I can get my multipart upload changes into Discourse. When the Uppy team cuts a proper release, we can bump the package.json versions instead.
When emails were forwarded to a group inbox by the email address
of the group, for example when an email ends up in spam and must
be manually forwarded to the group+site@discoursemail.com address,
the OP of the topic ended up being the group's email address instead
of the sender who originally sent the email to the group inbox.
This commit detects that an email has been forwarded using existing
tools, and if the from address matches one of the group incoming
email addresses, then we look at the forwarded email's from address
and use that instead for the incoming email from address as well as
the staged/regular user used for the Topic.user.
This will make it much cleaner to forward emails into a group inbox,
and will prevent issues with PostAlerter where the OP is double-notified
for these emails.
Currently, pinned topics are ordered by the `bumped_at` column. This behavior is not desired because it gives admins no control over the order of pinned topics. This PR makes pinned topics ordered by the `pinned_at` column. A topic that is pinned last appears first in topic lists. If an admin wants an already pinned topic to appear first in the list of pinned topics, they'll have to unpin that topic and pin it again.
Meta topic: https://meta.discourse.org/t/how-do-i-set-the-order-of-pinned-topics/16935/23?u=osama.
Emails can include the marker in a different language, depending on
site and user settings. The email receiver always looked for the marker
in default language.
Uploads can be reused between site settings. This change allows the same
upload to be exported only once and then the same file is reused. The
same applies to import.
Allow admins to configure exceptions to our Rails rate limiter.
Configuration happens in the environment variables, and work with both
IPs and CIDR blocks.
Example:
```
env:
DISCOURSE_MAX_REQS_PER_IP_EXCEPTIONS: >-
14.15.16.32/27
216.148.1.2
```
