GitLab

Exclude current offenses in `.rubocop_manual_todo.yml` and link the
follow-up issue.

Ensure using `User#can_read_all_resources?` or
`User#can_admin_all_resources?` instead of `User#admin?`

Enable new cops in RuboCop 0.9.3.

We bring this cop into the main config, and set a lower threshold.

Some files are excluded (migration specs, which we don't want to touch),
others are fixed, and some have explicit disablements.

The threshold of 28 is chosen as it accounts for the vast majority of
our usage. We should aim to bring this down to 20, and eventually closer
to 10.

Note that at present let_it_be is not counted.

This avoids autoloading non-production tooling code in production.
Signed-off-by: Rémy Coutable <remy@rymai.me>

This allows to move permanent `Exclude`s from todo files into the main
configuration file.

These were added because of Rails eagerload.
These files are now moved out of Rails eagerload path,
so the conditionals are not necessary.

This is redundant with Rubocop's Style/FrozenStringLiteralComment cop,
which we didn't have enabled when this was introduced.

To preserve the same behaviour we also:

- Switch to `EnforcedStyle: always_true`, to forbid `false` values.
- Replace the wildcard list with an explicit list of all violations,
  so we still catch all newly added files.

- add in deprecate cop helper

This completes the elimination of todos for Graphql/ResolverType,
which is now a fully-enabled Rubocop rule.

Prior this change we were introducing new offense without notice.

Add documentation in-class with bad and good examples as well.

Previously our GraphQL styleguide stated that description strings should
not end in periods. Descriptions provided by `ruby-graphql` objects all
do end in periods, so our overall API is inconsistent in how it presents
its schema descriptions.

This change switches our description styleguide to recommend that we add
periods to the end of description strings.

It also adds to the existing `GraphQL/Descriptions` cop to enforce this
which can auto-correct offenses.

We generate the data file using ruby-warnings

Adds rubocop:

- Only match if method name matches, not just line number
- Match only file path from root so that files from other machines work

Adds autocorrection

- Add autocorrection by inserting ** before.
- Special correction case for literal hash args

Mark cop as unsafe as relies on data source

Fix correction from splat type to double splat

This new rubocop rule enforces that all resolvers declare their type.

This is important for several reasons:

- developers can read a resolver and know what it does, by understanding
  what it is meant to return.
- we can enable better testing of resolvers, by creating fields from
  resolvers using `Resolver.field_options`.
- Resolvers become the SSOT for field metadata, enabling us to eliminate
  repetition and boilerplate.

- keep up to date and add new cops.

- Run autocorrection
- Add failing specs to todo

This makes changing the base class in future easier.

Let's use the more restrictive Types::GlobalIDType when possible

To avoid possible security issues we need to enforce the usage of the
`expected_type` parameter whenever possible when converting a GID string
into an object.

Avoid programmatic checks because they are slower as well.

It is discouraged to create a table along with more than one foreign
key at a time so registering a cop to check this violation would
eliminate the need of waiting for a feedback from db maintainer to
catch the violation.

`UploadedFile.from_params` should only be used by `multipart.rb`

MR fixes the cop for the files

Adds changelog for change

Add a new cop that checks that index operations in migrations like
remove_index, remove_concurrent_index, and index_exists? use an explicit
index name rather than relying on rails behavior to infer the index name
from the index definition.

Add a new cop that detects instances where migrations create or remove
a complex indexes that do not specify an explicit index name.

This change also detects GraphQL/JSONType offenses in mutations and
resolvers. Previously it was limited to detecting offenses only in
types.

See

https://gitlab.com/gitlab-org/gitlab/-/merge_requests/38125#note_389671877

Move existing offences to todo