Adjust skip trial copy in trial sign up flow for SaaS users who are
logged in

Update windows tags

See merge request gitlab-org/gitlab!23882

Replace pipeline custom action array header with slot and buttons

Closes #195352

See merge request gitlab-org/gitlab!22839

Closes #195352

https://gitlab.com/gitlab-org/gitlab/-/merge_requests/22839

Move Jupyter notebooks from repository to packages

See merge request gitlab-org/gitlab!23556

Include 'license_scanning' in Usage data

See merge request gitlab-org/gitlab!23690

Add deploy_token_type column to the deploy_tokens table

See merge request gitlab-org/gitlab!23530

Added migration to add column.
Added data migration and related spec file.
Added enum for deploy_token_type in model.

Revert: Add milestone issues display limit

See merge request gitlab-org/gitlab!24004

Use NodeUpdateService in update API

Closes #199112

See merge request gitlab-org/gitlab!23894

Adds a format for relative and open date ranges

See merge request gitlab-org/gitlab!23584

This reverts merge request !23102

Upgrade to Rails 6

Closes #30030 and #31034

See merge request gitlab-org/gitlab!19891

Fix Geo Sidebar missing `active` class

Closes #199238

See merge request gitlab-org/gitlab!23885

Spec

Tests and lint

Move the clone button to the tree controls area

See merge request gitlab-org/gitlab!17752

https://gitlab.com/gitlab-org/gitlab/-/merge_requests/17752

- Improve validations of time ranges
- Use deconstructing for named parameters in functions
- Improve time range examples in jsdoc
- Switch to placing the functions in `const`

There are basically 2 big updates in this commit:

1) Split up each range type handler into a separate function
    (plus a few helper functions).
2) Export a type infer-er

The type infer-er is useful internally, but also helps me
feel more confident about code changes. In theory no 
type should ever collide with another type, so this 
function is kind of like the "canary in the coal mine" 
in that we should always be able to test a 1:1 range 
type from this function. If we can test the types 
thoroughly, we can be more confident each respective 
handler is working on the right set of data.

On that note, splitting up the logic for each range type 
makes me feel much more confident that each one is 
handled in a way that doesn't conflict with any other. 
More importantly, any potential changes to an existing 
range type or additions of new range types will: 
A) Not cause any changes to the existing ones and
B) Scale gracefully

On that final note of graceful scaling, I find the (very
common) method of basically extending a long 
if/else if/else block to be frightening from the 
perspective of ever needing to add things. I've found 
the method I use here (a dictionary lookup, in a bunch 
of cases) to be much easier to reason about: every 
addition adds a single dictionary entry, and the JS 
selects it (O(1)) when it's needed.

The date ranges can be used to represent periods of time
relative to arbitrary points in time, such as a fixed date or the
current moment.

Create backend route for dedicated Release page

See merge request gitlab-org/gitlab!23792

Remove existing to clarify examples

See merge request gitlab-org/gitlab!23987

Refactor merge request routing

See merge request gitlab-org/gitlab!23782

Added Monaco-based editor lite version

Closes #198618

See merge request gitlab-org/gitlab!23767

This introduces the bare-minimum Monaco editor for single-file edits
like those for Snippets, blobs and CI/CD configurations

Merge branch '197973-staged-and-unstaged-modification-tooltip-for-file-with-staged-changes-in-web-ide' into 'master'

Resolve "Staged and unstaged modification" tooltip

See merge request gitlab-org/gitlab!23847

When credentials are changed, the field isn't marked as dirty
which results in the model not being saved or updated.

The query was passing a string for file_type when it is an int.
This is actually an unused column but was working in Rails 5
because it was just coerced to 0 which is the default value.

In Rails 6, it is now nil so no records match our query

In Rails 6, after_commit callbacks are no longer called
unless the model was saved

The behavior of this is going to change in Rails 6.1
where.not(a: 1, b: 2) is going to be:

WHERE (a != 1 OR b !=2 ) instead of WHERE (a != 1 AND b != 2)

Previously, having the `where` turned this into a relation
so `ActiveRecord::Relation#delete_all` is called.

In Rails 6, `ActiveRecord::Associations::CollectionProxy#delete_all`
is called causing an error in the query.

* Disable locking at the class-level for `update_all`

* Fixes DEPRECATION WARNING: `ActiveRecord::Result#to_hash` has been
  renamed to `to_a`. `to_hash` is deprecated and will be removed in
  Rails 6.1.

* Fix lookup of CE views

  `view_paths=` was removed so we just create a new LookupContext
  here instead of duplicating then modifying the paths

* Fix some view related spec failures

* Fix database configuration for connection pools

  ActiveRecord::Base.configurations now returns a
  ActiveRecord::DatabaseConfigurations object

We use `present_members` here so that associations are preloaded.
We do this even if there's only one record here because it raises
an exception when user.status isn't preloaded.

In Rails 6, content_type now returns the complete header. We use
media_type to just get the MIME type portion

See https://github.com/rails/rails/pull/36034

Removes patches that are already in Rails 6:
Gitlab::ContentDisposition patch
ActiveRecord::QueryCache patch
ActiveJob::Arguments patch
Gitlab::ActionViewOutput patch

Updates Activerecord::Locking::Optimistic patch to follow changes in
https://github.com/rails/rails/blob/v6.0.1/activerecord/lib/active_record/locking/optimistic.rb

Fix missing timezone for datetime_with_timezone by adding this custom
type to ActiveRecord's time_zone_aware_attributes