* Introducing unit (not integration) tests for the app layer
* Initial support for unit tests at the API
* Adding unit tests support to the store layer
* Add unit tests support in commands
* Adding last tests needed for run unit tests properly
* Fixing govet
* Removing some duplication
* Fixing tests
* Fixing tests
* Not compiling test helpers with the main module for api
* Revert "Not compiling test helpers with the main module for api"
This reverts commit 36a199bbe0.
* Fixing tests
* Fixing unit tests
* More consistency between api4/apiteslib.go and app/helper_test.go
* Renaming things to make more obvious the new Setup functions purpose
* Reverting change in go.sum
* Start with empty mock for app layer
* Start with empty mock for api layer
* Start with empty mock for web layer
* Renaming SetupWithStoreMockConfig to SetupConfigWithStoreMock
* Fixing tests on web package
* Removing unnecesary function
* Run all tests fasts in postgres
* Moving postgres config to a file
* Addressing PR review comments
* Testing against mysql in the CI to keep things less changed
* Revert "Testing against mysql in the CI to keep things less changed"
This reverts commit fc940c111d.
* Fixing a test broken in postgres but working in mysql
* Fixing some tests
* Fixing some config tests with postgres
* Fixing the rest of config tests
* Fixing govet error
* Fixing search tests for postgres
* Consistent license message for all the go files
* Fixing the last set of unconsistencies with the license headers
* Addressing PR review comments
* Fixing busy.go and busy_test.go license header
* Refactor more code to use testutils.WasCalled
* incorporate review comments
* Revert watcher_test.go changes
The file is in config package and the other tests use config_test.
So it is not visible.
* MM-18636: limit configuration writes to 4Mb
By default, MySQL silently truncates writes that exceed the column type in question. Change the column type from `TEXT` to `MEDIUMTEXT` to allow writes to the `Configurations` and `ConfigurationFiles` table to exceed 65535 bytes. This is a backwards compatible migration, but does require a rewrite of the table.
However, MySQL is further constrained by the default `max_allowed_packet` value of 4Mb, so limit writes accordingly.
Fixes: https://mattermost.atlassian.net/browse/MM-18636
* simplify unit tests
* fix import
* [MM-16719] fixing reflection: now recursing on value, not interface
adding defaults to cfgWithoutEnvOverrides
adding tests
* don't modify the minimalConfig
* create/update config.json using go generate
* added default config generator
added config-reset to Jenkins and make package, updated defaults to consider 'isNew' flag
* corrections after code review
* fixed Config.isValid to handle empty encryption keys
* fixed Config.isValid to handle empty encryption keys
* fixed Config.isValid to handle empty encryption keys
* isUpdate now only checks for nil
* Addressed review comments, added unit testing for default config generator
* err shadowing
* license
* provide output file for config generator via ENV variable, since go generate doesn't support arguments and we need two output paths (config-reset and package)
* cleanup
* proper defaults for PushNotificationServer and SendPushNotifications
* corrected generating defaults for TrustedProxyIPHeader to be consistent with default.json in master
* Check for empty SiteURL as well as nil
* corrected SiteURL settings and checking
* crazy typos fixed
* corrected tests to newly expected values
* relaxed the checks
* fixed formatting
* MM-11697: Environment overrides do not overwrite config.json on save
#10388
The config store now keeps a copy of the config as loaded from the store
without environment overrides. Whenever persisting, we now check if the
current setting is different from the loaded setting. If it is, then use
the loaded setting instead.
As described in the comments to `removeEnvOverrides` in `common.go`,
this behavior will have to change if we ever let the user change a
setting that has been environmentally overriden.
This was interesting because the `load` function in `common.go` also
persists, so we have to tee the provided `io.ReadCloser` and construct a
config that doesn't have the environment overrides. And then we have to
find the path to the (maybe) changed variable in the config struct
using reflection.
Possible WIP: I had to expose a `GetWithoutEnvOverrides` function in the
Store interface just for the tests -- this is because the `file_test`
and `database_test`s are in the config_test package instead of the
`config` package.
* added function documentation
* fixed a small problem with tests
* MM-11697: big cleanup based on Jesse's PR comments
* MM-11697: edits per PR feedback
* MM-11697: licence header
* MM-11697: now testing that on disk config is not changed by env
overrides
* MM-11697: remove unneeded exports
* TestGetLicenseFileFromDisk: avoid using fileutils.FindConfigFile
* config: abstract config-related file access, extend memory store
* simplify config validate to avoid file knowledge
* fix relative file tests
* cluster: fix ConfigChanged event
The old and new configurations were swapped when notifying the enterprise code of configuration changes, creating needless instability in propagating config updates across a cluster.
* config/database: ignore duplicates
* test cleanup
* remove unnecessary Save() in test
* MM-14145: The config store Set will now Save automatically
When UpdateConfig (and configStore.Set) is called in admin.go and
config.go, commonStore.Set now takes a store-specific persist function.
It uses that persist function to save the configuration automatically.
Removed: Now callers do not have to call configStore.Save or
app.PersistConfig, and those functions have been removed.
Possible downside: this means a "failed to persist config" error can now
be thrown during a app.UpdateConfig or commonStore.Set call. But
considering application code never really sets a config without saving
it (except in the test cases, which were testing that -- see below), it
seems fine to group these responsibilities.
Also removed: tests for 'set without save'. Since that can not happen
anymore, the tests are not needed.
* Removed Save completely, cleaned up formatting, joined save test with
set tests.
* fixed shadowed variable error
* [MM-13828] Running tests from a new temp folder with all test resources
Possible fix for #10132
All packages which have a TestMain and use testlib.MainHelper will have a new current working directory which will have all the test
resources copied.
Note: default.json is copied as config.json as well to make sure tests don't have any impact due to changes in config by devs
* [MM-13828] Added TestMain to remaining packages to use testlib.MainHelper
This makes sure tests from all packages run with same test resources, setup in a new temp folder for each package
* Updated Jenkins file to not not config/default.json
This makes sure CI has same config files as a dev's machine
* [MM-13828] Changes requested from code review
Added accessor methods to testlib.MainHelper for accessing members
Fixed some broken tests due to change in cwd while tests run
Some other code refactoring and improvements
* [MM-13828] Added new factory method with options for creating test main helper and some code refactoring
testlib.NewMainHelperWithOptions supports options to turn on/off test dependencies and environment setup
Some other code refactoring
* Exporting members of testlib.MainHelper to make enterprise tests work
* Fixed gofmt error
* [MM-13828] removed unwanted dependency on plugins directory while setting up test resources
* [MM-13828] Fixed some tests failing due to them being running from temp folder
* [MM-13828] Some code changes suggested in PR review
* Fixed gofmt error
* vendor github.com/jmoiron/sqlx
* MM-11262: introduce a database store
* revert unnecessary fmt.Errorf
* simplify unit test helper methods
* remote TODO re: retry
* relocate initializeConfigurationsTable for clarity
* factor out a commonStore
* acquire database config lock on close for safety
* add missing header
* fix lock comment
