This commit introduces an advisory lock at the lfs_objects_model level
to ensure uniqueness of new records (only one lfs_objects_project
per repository type).

After this, we need to add a migration that deduplicates
lfs_objects_projects by lfs_object_id, project_id and repository_type.
After the migration is finalized, we need to add a unique index on all
three columns to ensure data integrity since the unique validation
already exists at the model level.

Changelog: other

Changelog: performance

Only in rspec context

Since the `disable_git_http_fetch_writes` FF does not define
a feature flag actor, it is currently not possible to increment
the fetch project statistics without enabling or disabling it
for all projects.

This commit introduces the `allow_git_http_fetch_writes` FF
that allow us to control this and unskip the statistics
for specific projects.

Changelog: changed

Changelog: changed

Co-authored-by: Terri Chu <tchu@gitlab.com>

This mutation allows to trigger the creation of a security policy
project. We also introduce the securityPolicyProjectCreated subscription.
It will be triggered as soon as the project is created.

Changelog: added
EE: true

Changelog: changed

Co-authored-by: David Fernandez <dfernandez@gitlab.com>

Removes feature flag as we have successfully rolled out the new
feature.

Changelog: removed

When publishing a npm package that its archive has been bundled with a custom root folder name, an error is raised stating that package.json not found. In this MR, we fix that by validating the archive as long as it has package.json file in its root regardless of the root folder name.

Changelog: fixed

Add a feature flag to control prompts migration
Use the new FF to route to AiGatewayCodeCompletionMessage
Adjust the base prompt in AiGatewayCodeCompletionMessage to be nil

Changelog: changed
EE: true

Remove flag references in code

Changelog: fixed
MR: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/160190

Calling `IO::Buffer.new` produces this warning. It looks like the async
gem calls `IO::Buffer.new` on `require`.

So we silence warnings while requiring the async gem. In Ruby 3.3, the
gem no longer calls `IO::Buffer.new` so we can skip silencing.

Set logger level to INFO for the google cloud storage library
used to fetch package metadata from a gcp bucket.

Changelog: changed

Co-authored-by: Sarah Yasonik <syasonik@gitlab.com>

https://github.com/elastic/elasticsearch-ruby/pull/1360 added a
Elasticsearch version validation step, but unfortunately this causes
the client to fail with OpenSearch. Patch the client to maintain
compatibility.

Update the tests to fix test failure issues present in the latest
gem. The tests were failing because elasticsearch-transport attempted
to inspect the raw string for GZIP headers, but the stubbed response
was a Ruby array so this failed.

Changelog: removed
EE: true

Changelog: added

This change removes the
store_first_reviewer_assignment_timestamp_in_metrics feature flag.

Changelog: removed

Removes diff_preview_in_email feature flag, fully enabling the feature.

Changelog: other

The FF has been enabled by default since 16.11
Removing the flag as we have seen no issues
reported by users.

Changelog: other

Sets claude 3.5 sonnet by default
Removes 'claude_3_code_generation_haiku' feature flag
Removes 'claude_3_5_code_generation_sonnet' feature flag

Changelog: changed

Create callout component and add unit tests.

This is part of the work for consolidating all feature flags involved in
work item epics

Changelog: other
EE: true

Rename metrics group ide -> remote_development

This commit adds syncing of writes to
new audit event tables behind a feature flag

EE: true

Changelog: added
EE: true

Changelog: other

Changelog: changed

Changelog: changed
EE: true

Changelog: removed
EE: true

This removes the :jira_connect_proxy_create_branch feature flag as the feature has been globally enabled on production for a week with no issues.

Changelog: fixed

This refactors the flag name so that it will be a flag for the core
user mapping feature, to allow 1 flag per importer we roll the feature
out to.

- `importer_user_mapping`, applies to the core user mapping feature
- Have one flag per importer we roll user mapping out to, example
  `bulk_import_user_mapping`. We would always additionally check the
  flag state of `importer_user_mapping` with these flag checks.

This lets us globally switch user mapping off for all importers, if we
need to (could be handy if we find user mapping strains production
services while we’re evaluating it) but still control the rollout to
individual importers in a granular way.