This was always the intended API as per discussion in https://github.com/rails/rails/issues/36916, however versions before Rails 6 always passed a symbol. This means that some apps relied on symbols to be passed, this change is intended to note the change in behaviour so that apps can respond in kind.

* add leading `#` before `=>` since hash rocket is valid Ruby code
* add backticks
* remove trailing spaces
* and more

Before 34cc301f, type casting by boolean attribute when querying is a
no-op, so finding by truthy boolean string (i.e.
`where(value: "true") # => value = 'true'`) didn't work as expected
(matches it to FALSE in MySQL #32624). By type casting is ensured, a
value on boolean attribute is always serialized to TRUE or FALSE.

In PostgreSQL, `where(value: :false) # => value = 'false'` was a valid
SQL, so 34cc301f is a regresson for PostgreSQL since all symbol values
are serialized as TRUE.

I'd say using `:false` is mostly a developer's mistake (user's input
basically comes as a string), but `:false` on boolean attribute is
serialized as TRUE is not a desirable behavior for anybody.

This allows falsy boolean symbols as false, i.e.
`klass.create(value: :false).value? # => false` and
`where(value: :false) # => value = FALSE`.

Fixes #35676.

- I feel `i18n_customize_full_messages` explains the meaning of the
  config better.
- Followup of https://github.com/rails/rails/pull/32956

* Update RAILS_VERSION
* Bundle
* rake update_versions
* rake changelog:header

When assigning a hash to a time attribute that's missing a year
component (e.g. a `time_select` with `:ignore_date` set to `true`)
then the year defaults to 1970 instead of the expected 2000. This
results in the attribute changing as a result of the save.

Before:

    event = Event.new(start_time: { 4 => 20, 5 => 30 })
    event.start_time # => 1970-01-01 20:30:00 UTC
    event.save
    event.reload
    event.start_time # => 2000-01-01 20:30:00 UTC

After:

    event = Event.new(start_time: { 4 => 20, 5 => 30 })
    event.start_time # => 2000-01-01 20:30:00 UTC
    event.save
    event.reload
    event.start_time # => 2000-01-01 20:30:00 UTC

Related to https://github.com/rails/rails/pull/34817#issuecomment-451508668

Generally followed the pattern for https://github.com/rails/rails/pull/32034

* Removes needless CI configs for 2.4
* Targets 2.5 in rubocop
* Updates existing CHANGELOG entries for fewer merge conflicts
* Removes Hash#slice extension as that's inlined on Ruby 2.5.
* Removes the need for send on define_method in MethodCallAssertions.

Fixing code block rendering, indentation, backticks, etc.

The purpose of fe9547b6 is to work type casting to value from database.

But that was caused not to use the value before type cast even except
Active Record.

There we never guarantees that the value before type cast was going to
the used in this validation, but we should not change the behavior
unless there is some particular reason.

To restore original behavior, still use the value before type cast if
`came_from_user?` is undefined (i.e. except Active Record).

Fixes #33651.
Fixes #33686.

Related to #31503

Since we have `has_secure_token`, it is too confusing to use `_token`
suffix with `has_secure_password`.
Context https://github.com/rails/rails/pull/33307#discussion_r200807185

Add mention about default value of `config.active_model.i18n_full_message`.

We only add the header when releasing to avoid some conflicts.

[ci skip]

Skipping over 2.4.0 to sidestep the `"symbol_from_string".to_sym.dup` bug.

References #32028

:tada::tada::tada:

We don't implement much custom marshalling logic for these objects, but
the proc default case needs to be handled separately. Unfortunately
there's no way to just say "do what you would have done but with this
value for one ivar", so we have to manually implement `marshal_load` as
well.

The test case is a little bit funky, but I'd really like an equality
test in there, and there's no easy way to add one now that this is out
of AR (since the `attributes` method isn't here)

Fixes #31216

I found a bug that validation callbacks don't fire on multiple context.
So I've fixed it.

Example:

```ruby
class Dog
  include ActiveModel::Validations
  include ActiveModel::Validations::Callbacks

  attr_accessor :history

  def initialize
    @history = []
  end

  before_validation :set_before_validation_on_a, on: :a
  before_validation :set_before_validation_on_b, on: :b
  after_validation :set_after_validation_on_a, on: :a
  after_validation :set_after_validation_on_b, on: :b

  def set_before_validation_on_a; history << "before_validation on a"; end
  def set_before_validation_on_b; history << "before_validation on b"; end
  def set_after_validation_on_a;  history << "after_validation on a" ; end
  def set_after_validation_on_b;  history << "after_validation on b" ; end
end
```

Before:

```
d = Dog.new
d.valid?([:a, :b])
d.history # []
```

After:

```
d = Dog.new
d.valid?([:a, :b])
d.history # ["before_validation on a", "before_validation on b", "after_validation on a", "after_validation on b"]
```

time.to_date conversion happens considering leap years
so a conversion of "Day.new({'day(1i)'=>'1', 'day(2i)'=>'1', 'day(3i)'=>'1'})" results in saving the date as Mon, 03 Jan 0001
which might seem weird on the user level, hence falling back to parsing on string level resolves this data mismatch
Fixes #28521

This brings the Length validator in line with the Numericality
validator, which currently supports Proc & Symbol arguments

ActiveModel::Errors#merge! allows ActiveModel::Errors to append errors from
a separate ActiveModel::Errors instance onto their own.

Example:

    person = Person.new
    person.errors.add(:name, :blank)

    errors = ActiveModel::Errors.new(Person.new)
    errors.add(:name, :invalid)

    person.errors.merge!(errors)
    puts person.errors.messages
    # => { name: ["can't be blank", "is invalid"] }

[ci skip]