If timestamp column have the precision, it need to format according to
the precision of timestamp column.

Respect the database default charset for `schema_migrations` table.

Add `auto_increment?` instead of `extra == 'auto_increment'`

Remove unused line

Missing `@` [ci skip]

The charset of `version` column in `schema_migrations` table is depend
on the database default charset and collation rather than the encoding
of the connection.

This predicate is only used in `query_attribute`, and is relatively easy
to remove without adding a bunch of is a checks.

This predicate was only to figure out if it's safe to do case
insensitive comparison, which is only a problem on PG. Turns out, PG can
just tell us whether we are able to do it or not. If the query turns out
to be a problem, let's just replace that method with checking the SQL
type for `text` or `character`. I'd rather not burden the type objects
with adapter specific knowledge.

The *real* solution, is to deprecate this behavior entirely. The only
reason we need it is because the `:case_sensitive` option for
`validates_uniqueness_of` is documented as "this option is ignored for
non-strings". It makes no sense for us to do that. If the type can't be
compared in a case insensitive way, the user shouldn't tell us to do
case insensitive comparison.

The type code is actually quite accessible, and I'm planning to
encourage people to look at the files in the `type` folder to learn more
about how it works. This will help reduce the noise from code that is
less about type casting, and more about random AR nonsense.

It only existed to make sure the subclasses of `Delegator` were YAML
serializable. As of Ruby 2.2, these are YAML dumpable by default, as it
includes
https://github.com/tenderlove/psych/commit/2a4d9568f7d5d19c00231cf48eb855cc45ec3394

This allows us to remove `Type::Value#klass`, as it was only used for
multi-parameter assignment to reach into the types internals. The
relevant type objects now accept a hash in addition to their previous
accepted arguments to `type_cast_from_user`. This required minor
modifications to the tests, since previously they were relying on the
fact that mulit-parameter assignement was reaching into the internals of
time zone aware attributes. In reaility, changing those properties at
runtime wouldn't change the accessor methods for all other forms of
assignment.

NameError#missing_name? can just use NameError#name if the arg is a Symbol

remove noise from AM tests

Use the Rails favicon.ico instead of the ruby favicon.ico for rails guides [ci skip]

NameError#name returns a missing name as a symbol, so if the given name
is a symbol, it doesn't have to use #missing_name to get the last constant
name in the error message.

make zones_map private

Conflicts:
	activesupport/lib/active_support/values/time_zone.rb

use kwargs instead of xhr method. refs #18771.

[ci skip]

Closes #16538

Fixed undefined method `i18n_key' for nil:NilClass for labels in non AR ...

Refactoring at #18647 broke using non active record objects in form_for. This patch
restores the original behaviour where we only compute i18 key when object.respond_to?(:to_model)

README.rdoc was generated to support the doc:app task. Now that
this task is gone we can switch to Markdown, which is nowadays
a better default.

This patch removes the tasks doc:app, doc:rails, and doc:guides.

In our experience applications do not generate APIs using doc:app.
Methods may be certainly documented for maintainers, annotated
with YARD tags, etc. but that is intended to be read with the
source code, not in a separate website. Then, teams also have
typically selected topics written down in Markdown files, or in
a GitHub wiki... that kind of thing.

If a team absolutely needs to generate application documentation
for internal purposes, they can still easily write their own task.

Regarding doc:rails and doc:guides, we live in 2015. We are used
to go to online docs all the time. If you really want access to the
API offline RubyGems generates it for every Rails component unless
you tell it not to, and you can checkout the Rails source code to
read the guides as Markdown, or download them for a Kindle reader.

All in all, maintaining this code does not seem to be worthwhile
anymore.

As a consequence of this, guides (+3 MB uncompressed) won't be
distributed with the rails gem anymore. Of course, guides and API
are going to be still part of releases, since documentation is
maintained alongside code and tests.

Also, time permitting, this will allow us to experiment with novel
ways to generate documentation in the Rails docs server, since
right now we were constrained by being able to generate them in
the user's environment.

nil or false should not be valid argument to the merge method.

Closes #12264

Relation#merge checks if the argument is an array and the only possible
returns of build_default_scope is nil or a Relation.

Doing this we can raise an ArgumentError when Relation#merge receive a
nil value.

Add an `:only` option to `perform_enqueued_jobs` to filter jobs based on type.

type.

This allows specific jobs to be tested, while preventing others from
being performed unnecessarily.

Example:

    def test_hello_job
      assert_performed_jobs 1, only: HelloJob do
        HelloJob.perform_later('jeremy')
        LoggingJob.perform_later
      end
    end

An array may also be specified, to support testing multiple jobs.

Example:

    def test_hello_and_logging_jobs
      assert_nothing_raised do
        assert_performed_jobs 2, only: [HelloJob, LoggingJob] do
          HelloJob.perform_later('jeremy')
          LoggingJob.perform_later('stewie')
          RescueJob.perform_later('david')
        end
      end
    end

Fixes #18802.

Trim space and document :only option.

The `&block` form is more than twice as fast as the manual form of
delegation (and is the code I'd rather write anyway). Unfortunately,
it's still twice as slow on MRI. However, this is enough of a hotspot to
justify giving JRuby special treatment.

I can't currently provide benchmarks in the context of Active Record,
since the JDBC adapters still aren't updated for 4.2, but the actual
work performed (assuming it's been read at least once already) will have
nearly the same performance characteristics as
https://gist.github.com/sgrif/b86832786551aaee74de.

We already validate the keys, so it is better to use the built-in
feature to do this