EE: true
Changelog: changed

Auto-corrected via:
  rubocop -a --only Style/RedundantFreeze scripts spec tooling qa

Some places are calling strong_memoize_with() without parameters. They
should be using strong_memoize_attr() instead. Update the RuboCop to
reflect this.

The `Graphql/GIDExpectedType` cop enforces that we pass an
`expected_type:` param when calling `Gitlab.object_from_id`, but our
Global IDs validate that the object resolved from a Global ID matches
the model which does the same check as the :expected_type param.

- https://gitlab.com/gitlab-org/gitlab/-/blob/45b3c596ef8b181bc893bd3b71613edf66064936/app/graphql/types/global_id_type.rb#L98
- https://gitlab.com/gitlab-org/gitlab/-/blob/ac2116a0abbf98019145cbffa0fbe569c7e2f978/app/graphql/gitlab_schema.rb#L128

https://gitlab.com/gitlab-org/gitlab/-/issues/417093

Fixes some `Naming/InclusiveLanguage` offenses.

This commit:
- Moves some of the settings Service Ping metrics to instrumentation
  classes.
- Adds PrometheusMetric to handle metrics querying Prometheus (being
  wrapped by the #with_prometheus call) without resorting to evaluate
  the metric value block in another context.
- Fixes some specs that were broken by migrating the metrics to the
  instrumentation class framework.

Auto-corrected via:

  rubocop -a --only Performance/RegexpMatch ...

Auto-corrections manually reviewed and verified.

Signed-off-by: Rémy Coutable <remy@rymai.me>

It prevents the usage of this deprecated method

This migrates two Jira usage metrics in Service Ping to instrumentation
classes framework.

They have been subsequently removed from `Gitlab::UsageData`, as they
are generated directly from instrumentation classes.

* Rename Rubocop cop
* Consider `let_it_be` aliases
  * For example: `let_it_be_with_reload`
* Ensure that example group alternatives are caught
  * For example: `it_behaves_like` and `include_examples`

Previously, we only retained HAML exclusions which ended with
`html.haml.rb`. We also need to retain exclusions with different format
like `text`, `atom`, or no format (by mistake).

HAML files (ending with `.html.haml.rb`) added (manually) to
`.rubocop_todo/**/*.yml` now are retained and not removed.

This is useful because `haml-lint` (which use RuboCop rules) cannot
exclude HAML files per RuboCop. So, we tend to disable whole cop rules in
`haml-lint.yml`.

With this change we can now start enabling "temporarily" disabled cop
rules in add HAML files as TODOs.

It catches exceptions being rescued without being properly
re-raised inside the Job Class.

Changelog: changed
EE: true

We allow `add_column` and `create_table` methods inside the `down`
method in post-deploy migrations. But if these methods were wrapped
inside a statement like `if` then the cop was raising an error.
For example, in the following, the 2nd example was being incorrectly
flagged as an offense:

```
def down
  add_column(:table, :column)
end
```

```
def down
  add_column(:table, :column) unless column_exists(:table, :column)
end
```

Use Pathname API as much as possible and resolve all RuboCop offenses.

If specs change in {,ee/}/spec/migrations/**/*_spec.rb change RuboCop's
cache should be invalidated for this cop to avoid flaky results in CI.

Also, use RESTRICT_ON_SEND to define target method which is used by RuboCop
internally to speed-up scans.

Adds database migration and model definitions necessary to add
visibility toggle for Model experiments

Changelog: added

Previously, this cop only flagged `self.ignored_columns += [...]`.

Because of this change the cop rule is put back in "grace period" to
maintain `master` stability.

Flag local assignments during factory "load time". This leads to
static data definitions.

Move these definitions into attribute block or
`transient` block to ensure that the data is evaluated during
"runtime" and remains dynamic.

Example:
  # bad
  factory :foo do
    random = rand(23)
    baz { "baz-#{random}" }

    trait :a_trait do
      random = rand(23)
      baz { "baz-#{random}" }
    end

    transient do
      random = rand(23)
      baz { "baz-#{random}" }
    end
  end

  # good
  factory :foo do
    baz { "baz-#{random}" }

    trait :a_trait do
      baz { "baz-#{random}" }
    end

    transient do
      random { rand(23) }
    end
  end

Current offenses are excludes via
  rake rubocop:todo:generate[Graphql/ResourceNotAvailableError]

Put this cop in grace period.

Show the full source code in the offense message instead of its AST.

This same functionality is now covered by Gettext/StaticString.

This cop ensures that only static strings are passed to gettext
translation methods `_()`, `s_()`, `n_()`, `N_()`.

Any kind of interpolation, formatting, or concatenation is disallowed.

The following definitions are allowed:
* Local variables
* Method calls which are not mentioned above
* Constants

Generate TODOs for pending offenses in Ruby code.

This primarily targets ActionMailer::MailDeliveryJob which our mailers
send to Sidekiq. The ActiveJob wrapper does not include
ApplicationWorker which makes it tricky to configure `data_consistency`.

This MR introduces another path of logic to handle active jobs and
enable db load balancing for those jobs.

Changelog: other

Changelog: added
EE: true

An `authorize` statement with no arguments is the same as not having an
authorize statement at all. So, we should flag an offense if `authorize`
does not have any arugments.

Also:

- Moves geo database docs to ee/db/geo/docs/
- Adds support for embedding database docs in ee/db/embedding/docs/

Relates to https://gitlab.com/gitlab-org/gitlab/-/issues/404396

This MR removes usage of old method from the codebase.
Also removed corresponding rubocop check

- ensure good practice to try and leverage the non
  primary database