... when the user is destroyed.

1. Normally, for a given awardable and award emoji name, a user is only allowed
   to create a single award emoji.

2. This validation needs to be removed for ghost users, since:

   - User A and User B have created award emoji - with the same name and against
     the same awardable
   - User A is deleted. Their award emoji is moved to the ghost user
   - User B is deleted. Their award emoji needs to be moved to the ghost user.
     However, this breaks the uniqueness validation, since the ghost user is
     only allowed to have one award emoji of a given name for a given awardable

... when the user is destroyed.

To clarify, this regards abuse reports that the to-be-deleted user has _reported_.

Introduction
------------

1. The foreign key was not explicitly specified on the association.
2. The `AbuseReport` model contains two references to user - `reporter_id` and
   `user_id`
3. `user.abuse_report` is supposed to return the single abuse report where
   `user_id` refers to the given user.

Bug Description
---------------

1. `user.abuse_report` would return an abuse report where `reporter_id` referred
   to the current user, if such an abuse report was present.

2. This implies a slightly more serious bug as well:

   - Assume User A filed an abuse report against User B
   - We have an abuse report where `reporter_id` is User A and `user_id` is User B
   - If User A is updated (`user_a.block`, for example), the abuse report would
     also be updated, such that both `reporter_id` _and_ `user_id` point to User A.

Fix
---

Explicitly declare the foreign key `user_id` in the `has_one` declaration

... when the user is destroyed.

1. When the user is deleted.

2. Refactor out code relating to "migrating records to the ghost user" into a
   `MigrateToGhostUser` concern, which is tested using a shared example.

Rename Rake task from geo:add_primary_node to geo:set_primary_node

See merge request !1579

[ci skip]

[ci skip]

[ci skip]

[ci skip]

[ci skip]

[ci skip]

Read track from deployment and visualise canary deployments

See merge request !1551

Burndown Charts

Closes #91

See merge request !1540

Add check for deploy boards

See merge request !1555

Fix elasticsearch docs: ID=1 sudo ... is ineffective

See merge request !1573

Add a Rake task to make the current node the primary Geo node

See merge request !1533

Build issues for merge request when there is an issue template

Closes #1968

See merge request !1550

Return 404 instead of a 500 error on API status endpoint if Geo tracking DB is not enabled

See merge request !1571

Strip trailing forward slashes from elasticsearch URLs

Closes #2039

See merge request !1566

Ee fix groups long url

See merge request !1568

Fix intermittent spec failure in spec/lib/gitlab/mirror_spec.rb

Closes #2081

See merge request !1570

If you attempt to run this spec between 00:00 and 01:00, one of
the examples fails because:

DateTime.now.beginning_of_hour == DateTime.now.at_beginning_of_day

Closes #2081
Signed-off-by: Rémy Coutable <remy@rymai.me>

Resolve "Elasticsearch: Highlighting in the code results does not work"

Closes #1567

See merge request !1564

[ci skip]

[ci skip]

If the tracking database is not configured, the secondary attempts to read from
the Geo DB but fails spectacularly because the tables don't exist.

To support #1664, we need a way to add a Geo Node to the database with an input SSH key.