https://gitlab.com/gitlab-org/gitlab-foss/merge_requests/31741, shipped
in GitLab 12.2, reduced Gitaly RPCs by generating commit data only if
hooks or services were actually active. However, GitLab plugins
(https://docs.gitlab.com/ee/administration/plugins.html), are executable
programs that get launched inside `SystemHooksService#execute_hooks`
regardless of whether any hooks are actually active. As a result of this
change, users have to configure a Webhook for plugins to fire.

We restore the previous behavior by checking whether plugins are
available and executing `SystemHooksService#execute_hooks` if any
plugins are available.

Closes https://gitlab.com/gitlab-org/gitlab/issues/194131

Update foreign key constraint for personal access tokens

Closes #191262

See merge request gitlab-org/gitlab!22305

Docs: Remove duplicate deploy action from aws.md

See merge request gitlab-org/gitlab!22365

Use Ruby 2.6.5 CI images

See merge request gitlab-org/gitlab!22372

fix include template code examples for SAST

See merge request gitlab-org/gitlab!22339

Update github import test

Closes #193485

See merge request gitlab-org/gitlab!22353

It needs to click the authenticate button, and then the repos are listed

Docs: Remove extraneous text

See merge request gitlab-org/gitlab!22403

Replace breakpoints with GlBreakpointInstance in issuable_context.js

See merge request gitlab-org/gitlab!22008

Check that uploaded file content matches the file extension

See merge request gitlab-org/gitlab!20697

Handle when extension is missing
Handle ambiguous file type checks
- Content of text files often don't define their type.
- A text file may be missing a file extension.
- We may encounter a file with no extension that we can't magically
type.
Optionally check integrity of specific mime types
Ensure upload integrity by file extension
Apply integrity check to AvatarUploader and FaviconUploader

Migrate deployment_instance_spec.js spec to jest

See merge request gitlab-org/gitlab!22169

Remove ignored columns from Plan model

See merge request gitlab-org/gitlab!22302

[ci skip]

[ci skip]

Optimize Group#for_epics scope

Closes #37368 and #37748

See merge request gitlab-org/gitlab!22375

Fix tests failing in async mode

Closes #193491

See merge request gitlab-org/gitlab!22361

This commit adds sync: false to our tests, preparing
for VTU upgrade

Previously the foreign key constraint for `user_id` on the
`personal_access_tokens` table did not have the ON DELETE CASCADE
option enabled, which made deleting users that had generated a
PAT impossible.

This migration adds a new foreign key constraint with this option and
drops the old one.

Closes https://gitlab.com/gitlab-org/gitlab/issues/191262

- ee/spec/javascripts/environments/deploy_board_component_spec.js
- ee/spec/javascripts/vue_shared/components/deployment_instance_spec.js

Prevent stack overflowing with a lot of user references

Closes #119333

See merge request gitlab-org/gitlab!22247

Replace search by stub name with real components

Closes #191270

See merge request gitlab-org/gitlab!22300

Add support for operator in filter bar

See merge request gitlab-org/gitlab!19011

Drop redundant index on ci_pipelines.project_id

See merge request gitlab-org/gitlab!22325

When an issue, MR, or epic contains a lot of user references (even to
the same user), we use the method
Banzai::ReferenceParser::BaseParser.collection_objects_for_ids to load
those users and maintain the loaded users for the current request in a
RequestStore cache.

This method calls Hash#values_at with the splatted IDs - so one argument
per ID passed in. If enough IDs are passed in, it's possible for the
arguments here to overflow the stack.

Let's say we have:

    ids = [1] * 1_000_000; ids.length
    #=> 1000000
    cache = {}
    #=> {}
    cache.values_at(*ids).compact
    # SystemStackError: stack level too deep

We could do this `cache.values_at(*ids.uniq).compact`, but that would
fail in this case:

    distinct_ids = 1.upto(1_000_000).to_a; distinct_ids.length
    #=> 1000000
    cache.values_at(*distinct_ids.uniq).compact
    # SystemStackError: stack level too deep

So the best option seems to be to just write a slower, but
non-stack-consuming, version:

    Benchmark.realtime { ids.uniq.map { |id| cache[id] }.compact }
    #=> 0.15192799968644977
    Benchmark.realtime { distinct_ids.uniq.map { |id| cache[id] }.compact }
    #=> 0.3621319998055696

To test this locally, you can replace the last issue's description with
a lot of mentions of a user (here I've created a user with the username
't'), but it takes a long time:

    problem = ('@t ' * 1_000_000)[0...1_000_000]; problem.length
    #=> 1000000
    Issue.last.update!(description: problem_string)

Visiting that issue in the UI will then overflow the stack. With this
change, it won't (although it is very slow).

Pass through TOP_UPSTREAM_SOURCE_REF for qa pipeline

See merge request gitlab-org/gitlab!22263

Signed-off-by: Balasankar "Balu" C <balasankarc@autistici.org>

Signed-off-by: Balasankar "Balu" C <balasankarc@autistici.org>

Remove optimized_groups_user_can_read_epics_method feature flag and keep
the optimized code in the codebase.

Closes https://gitlab.com/gitlab-org/gitlab/issues/37748

Fix shallowMount + stubs behavior change in `@vue/test-utils`

Closes #145615

See merge request gitlab-org/gitlab!22286

Provide proper mock data for test

Closes #191453

See merge request gitlab-org/gitlab!22316

Replace trigger with emit in test

Closes #191430

See merge request gitlab-org/gitlab!22315

Make sure we are searching components by reference
Not by generated stub name

Test for no offenses only makes sense with offense

See merge request gitlab-org/gitlab!22301

Add ajax wait for requests

Closes #55286

See merge request gitlab-org/gitlab!22291

Stop exposing MR refs in favor of persistent pipeline refs

Closes #35918 and #35140

See merge request gitlab-org/gitlab!22198

Add weight change tracking

See merge request gitlab-org/gitlab!21515

This commit fixes the inconsistent behavior of MR refs