While invalidating a channel, we send out a cluster message
from the app layer as well as from the localcachelayer.
This causes multiple redundant messages to be sent out.
And additionally multiple cluster handlers were being registered
from the app layer as well as the localcachelayer.
We fix this by removing the cluster broadcast from the app layer
and directly calling the store method. We also remove
the duplicate event handler registration.
* MM-21336: Improve channel rename restrictions
Non-direct channels should not have __ in them.
* Fix order of i18n extraction
* Incorporate review comments
Use a more relaxed check by actually checking for valid userIds.
* Improve error message
When push notification setting was selected to "All Activity", we were
just sending the unread count for the channel from where the message was sent
and not for all channels.
After discussion with @enahum and @migbot, we decided to keep the badge count
to only refer to mentions and not unread posts.
Therefore, we only take the unread count for the user irrespective of the
notify_props settings.
Updated the tests to reflect that.
Co-authored-by: mattermod <mattermod@users.noreply.github.com>
* MM-21209: Use the LRU cache for UserStore.Get call too
We already have userProfileByIdsCache to store the user profiles by Id.
It just wasn't being used for the (*UserStore).Get method. We add a wrapper
method in LocalCacheUserStore to intercept that call and check for the
presence of the user Id in the cache.
There is no need to add any code for invalidation as all of that is already
present.
* Fix nil check for rootstore
* Fix TestUserStore test
Added an invalidate call. The invalidation was being done
from the app level. Hence we have to do it manually here for the test.
Display LHS bot Icon in web app. As part of mentioned task, Added LastIconUpdate variable in model.bot to store last update time of icon. Also added code to update/delete value of the mentioned variable when setting/deleting bot icon.
* Make WebSocketEvent type immutable
* Update code to use updated immutable WebSocketEvent type
* Export WebSocketEvent fields and mark them as deprecated
* migrated termsOfServiceCache from sqlstore to localcachelayer, and mocked store tests
* revert gitignore local change
* fixed caching in termsOfService Get, and added basic tests for termsOfServiceCache
* added a test for cache save, and fixed call to Store tests
* fixed GeLatest termsOfService from cache test
* added license headers to terms_of_service cache files
* using doStandardReadFromCache and doStandardAddToCache when reading and writing to cache
* removed unused variable, termsOfServiceCacheName
* added special key for the latest termsOfService value in termsOfServiceCache
* updated license information on localcachelayer termsOfServiceCache files
* fixed not updating latest termsOfServiceCache on Get by ID, and invalidating cache cluster on termsOfServiceCache save
Co-authored-by: mattermod <mattermod@users.noreply.github.com>
The `channelByNameCache` in the `GetByNames` query checks each teamId+channelName
combination in the cache and if any of the keys is not present, it queries the DB.
In this case, this is effectively a cache miss from the point of view of user.
But since the cache is at a teamId+channelName level, there may be several hits and some misses.
This is misleading as it does not generate proper metrics and might lead to a
false assumption of how effective the cache is.
We create separate metrics which account for all or none cache hits/misses.
This removes the old metrics which are used by other queries. So they are now separated,
and it will improve both the old and new metric. Because the old metric is not contaminated,
and the new metric is also separate.
* Migrate channelCache cache from store/sqlstore/channel_store.go to the new store/localcachelayer
* fix import for new path
* remove unnecessary code, refactor cache constant
* resolve conflicts
* fix missing variable and import after merging
* refactor code, fix current issue
Co-authored-by: mattermod <mattermod@users.noreply.github.com>
* Channel Member update event
* Address PR comments and extend tests
* golang ci addressed
* Update api4/channel_test.go
Co-Authored-By: Ben Schumacher <ben.schumacher@mattermost.com>
* Implement allowing multiple SAML Libraries
* fix unit tests
* updates based on review feedback
* Only set on startup, not config changes
* update unit tests, reload enterprise
* need to reset for previous tests
* update license statement
* fix licensing line
* MM-14675: Upserts GroupTeam when GroupChannel is linked.
MM-14675: Delete all associated GroupChannels when deleting a GroupTeam.
MM-14675: Using replica DB where possible.
MM-14675: Updates create method used in tests.
* MM-14675: Removes unnecessary DB retrieval of GroupSyncable record.
* MM-21118: Use querybuilder instead of fmt.Sprintf in channelStore
We use the querybuilder API instead of just building the query string
manually using fmt.Sprintf
* Replaced comma with colon
* Add MLOG for sql store
* Update printf function
* Update store/sqlstore/supplier.go
Co-Authored-By: Martin Kraft <martin@upspin.org>
* Update go mod
* update go mod and tidy it up
* update vendor folder
Go 1.13 introduced this new flag which removes all absolute file system paths
from the binary and just keeps the module path / GOPATH. To debug binary
crashes or stack traces, we only need the path to the code and not the full
file path. Hence this is a quick way to reduce the binary size without any information
loss.
Shaves off around 750KB from the server binary.
* MM-21103: change plugin signature path
Save as `<plugin_id>.tar.gz.sig` instead of `<plugin_id>.sig`. The latter was a relic of the previous design to support multiple plugin signatures, but now creates an inconsistency with how the original source files were supplied as `<some_name>.tar.gz` and `<some_name>.tar.gz.sig`.
Fixes: https://mattermost.atlassian.net/browse/MM-21103
* relax signature matches to avoid assuming signatures always exist
* MM-21116: Improve channelSearch SQL query generation
The likeTerm from buildFulltextClause was being built inefficient using strings.
We change it to use squirrel to build the query.
And while at it, we also change the len check with checking with
an empty string which is more idiomatic. Both compile to the same
code, so there is no difference performance wise.
* Use a better variable name
