Since the `ForkingExecutor` class seems to be pretty slow on Rubinius
due to DRb (c.f. http://git.io/xIVg), let's avoid running tests with
it on this platform.
Also, the `parallelize_me!` call make the suite to output a bunch of
errors due to rubinius/rubinius#2934 since there are thread-safety
problems with autoloading.
I found that nil values were being incorrectly sent to
`parse_nested_query` in Rack. Originally it was thought that Rails was
doing something incorrect but it was actually rack-test. This was fixed
in brynary/rack-test@4a4b2c1 and is no longer an issue. This commit
bumps rack-test in Rails so changes to Rack don't cause failures.
See rack/rack#813 for more information.
In f6e293ec54f02f83cdb37502bea117f66f87bcae we avoided a segfault in the
tests, however I think we should try to avoid the crash, as it may
happen in user code as well.
Here is what I distiled the bug down to:
```ruby
# Rails case - works on 2.0, 2.1; crashes on 2.2
require 'action_dispatch'
ActionDispatch::Response.new(200, "Content-Type" => "text/xml")
# General case - works on 2.0, 2.1; crashes on 2.2
def foo(optional = {}, default_argument: nil)
end
foo('quux' => 'bar')
```
Introduced in f6e293e ActionPack tests began sefaulting. I found that it
was the kwargs and the test causing the seg fault was missing the new
default_headers argument.
Most session stores offer an :expire_after option, but it's largely
undocumented. Cookie store also supports a number of options via
rack (these used to be documented in rails 2.3)
Haven't diagnosed yet. No similarly failing tests in Rails to work from.
cc @tenderlove, @eileencodes
Revert "there is always an integration session, so remove the check"
Revert "lazily create the integration session"
Revert "use before_setup to set up test instance variables"
This reverts commits 4cf3b8ac47f109fa83a6f66eb97d6cb0eace0d05, 303567e554de26822f3107be55c471d6477a745f, and fa63448420d3385dbd043aca22dba973b45b8bb2.
Partitioning of all the routes is currently being done during the
first request. Since there is no need to clear the cache for
`partitioned_routes` when adding a new route. We can move the
partitioning of the routes during setup time.
Fixes regression in #18423. Merge default headers for new responses,
but don't merge when creating a response from the last session request.
hat tip @senny ❤️
Fixed an issue where the `RAILS_RELATIVE_URL_ROOT` environment
variable is not prepended to the path when `url_for` is called.
If `SCRIPT_NAME` (used by Rack) is set, it takes precedence.
As part of #19029, in future `skip_before_action`, `skip_after_action` and
`skip_around_action` will raise an ArgumentError if the specified
callback does not exist. `skip_action_callback` calls all three of these
methods and will almost certainly result in an ArgumentError. If anyone
wants to remove all three callbacks then they can still call the three
individual methods. Therefore let's deprecate `skip_action_callback` now
and remove it when #19029 is merged.
Change filter on /rails/info/routes to use an actual path regexp from rails
and not approximate javascript version. Oniguruma supports much more
extensive list of features than javascript regexp engine.
Fixes#18402.
Collections can take advantage of `multi_read` if they render one template
and their partials begin with a cache call.
The cache call must correspond to either what the collections elements are
rendered as, or match the inferred name of the partial.
So with a notifications/_notification.html.erb template like:
```ruby
<% cache notification %>
<%# ... %>
<% end %>
```
A collection would be able to use `multi_read` if rendered like:
```ruby
<%= render @notifications %>
<%= render partial: 'notifications/notification', collection: @notifications, as: :notification %>
```
This has 2 effects:
1. RoutesProxy is CRAZY faster because it's no longer creating a new
Module each time method_missing is hit.
2. It bypasses an existing bug in ruby that makes `class << obj` unsafe
to be used in threading contexts.
Moves `X-Request-ID`, `action_dispatch.request_id` and
`HTTP_X_REQUEST_ID` strings to constants and freezes them.
We are freezing these strings to reduce the number of allocations in
Rails integration tests. The tests are spending a lot of time in GC and
this reduces the amount of time spent from 12% to 9% (in combination
with Rack PR that also freezes some strings).
Number of allocations before this change: 1030722
Number of allocations after this change: 967722
Add http_cache_forever to ActionController, so we can cache results
forever.
Things like static pages are a good candidate for this type of caching.
This cache only controls caching headers, so it is up to the browser to
cache those requests.
The methods `fresh_when` and `stale?` from ActionController::ConditionalGet
accept a single record as a short form for a hash. For instance
```ruby
def show
@article = Article.find(params[:id])
fresh_when(@article)
end
```
is just a short form for:
```ruby
def show
@article = Article.find(params[:id])
fresh_when(etag: @article, last_modified: @article.created_at)
end
```
This commit extends `fresh_when` and `stale?` to also accept a collection
of records, so that a short form similar to the one above can be used in
an `index` action. After this commit, the following code:
```ruby
def index
@article = Article.all
fresh_when(etag: @articles, last_modified: @articles.maximum(:created_at))
end
```
can be simply written as:
```ruby
def index
@article = Article.all
fresh_when(@articles)
end
```
PR #18772 changed the parameters of `stale?` to use `kwargs`.
[As for this comment](https://github.com/rails/rails/pull/18872/files#r24456288)
the default value for the `etag` parameter should be `record`, not `nil`.
This commit fixes the code and introduces a test that:
- passed before #18872
- fails on the current master (after #18772)
- passes again after setting the default value of `etag` to `record`.
In match_head_routes, deleted the routes in which request.request_method was empty (matches all HTTP verbs) when responding to a HEAD request. This prevents catch-all routes (such as Racks) from intercepting the HEAD request.
Fixes#18698
`url_helpers` used to be memoized. This was lost in a refactoring and
this PR adds it back. We noticed this while investigating why
integration tests are slower than controller tests.
This is an issue brought up by @daniel-rikowski in rails/web-console#91.
Citing his PR proposal here:
> Prior to this, backtrace lines were simply split by a single colon.
>
> Unfortunately that is also the drive letter delimiter in Windows paths
> which resulted in a lot of empty source fragments of "C:0". ("C" from
> the drive letter and 0 from "/path/to/rails/file.rb:16".to_i)
>
> Now the trace line is split by the first colon followed by some digits,
> which works for both Windows and Unix path styles.
Now, the PR was sent against web-console, because of the templates copy
issue we used to had. Instead of bothering the contributor to reopen the
issue against upstream Rails itself, I will make sure he gets the credit
by putting his name in [rails-contributors/hard_coded_authors.rb][].
[rails-contributors/hard_coded_authors.rb]: (https://github.com/fxn/rails-contributors/blob/master/app/models/names_manager/hard_coded_authors.rb).
Non-kwargs requests are deprecated now.
Guides are updated as well.
`post url, nil, nil, { a: 'b' }` doesn't make sense.
`post url, params: { y: x }, session: { a: 'b' }` would be an explicit way to do the same
Inside a controller functional test after the last flash is deleted it
still persists the flash because to_session_value is nil. We should
delete it from the session when the serialized version is nil, same as
the flash middleware.
Fixes an issue that would cause default_url_options to be lost when generating
URLs with fewer positional arguments than parameters in the route definition.
I found delegate to be a bottleneck during integration tests. Here is
the test case:
```ruby
require 'test_helper'
class DocumentsIntegrationTest < ActionDispatch::IntegrationTest
test "index" do
get '/documents'
assert_equal 200, response.status
end
end
Minitest.run_one_method(DocumentsIntegrationTest, 'test_index')
StackProf.run(mode: :wall, out: 'stackprof.dump') do
3000.times do
Minitest.run_one_method(DocumentsIntegrationTest, 'test_index')
end
end
```
Top of the stack:
```
[aaron@TC integration_performance_test (master)]$ stackprof stackprof.dump
==================================
Mode: wall(1000)
Samples: 23694 (7.26% miss rate)
GC: 1584 (6.69%)
==================================
TOTAL (pct) SAMPLES (pct) FRAME
7058 (29.8%) 6178 (26.1%) block in Module#delegate
680 (2.9%) 680 (2.9%) ActiveSupport::PerThreadRegistry#instance
405 (1.7%) 405 (1.7%) ThreadSafe::NonConcurrentCacheBackend#[]
383 (1.6%) 383 (1.6%) Set#include?
317 (1.3%) 317 (1.3%) ActiveRecord::Base.logger
281 (1.2%) 281 (1.2%) Rack::Utils::HeaderHash#[]=
269 (1.1%) 269 (1.1%) ActiveSupport::Notifications::Fanout::Subscribers::Evented#subscribed_to?
262 (1.1%) 262 (1.1%) block (4 levels) in Class#class_attribute
384 (1.6%) 246 (1.0%) block (2 levels) in Class#class_attribute
```
According to @eileencodes's tests, this speeds up integration tests so
that they are only 1.4x slower than functional tests:
Before:
INDEX: Integration Test: 153.2 i/s - 2.43x slower
After:
INDEX: Integration Test: 275.1 i/s - 1.41x slower
Shorthand route match is when controller and action are taken literally from path.
E.g.
get '/foo/bar' # => will use 'foo#bar' as endpoint
get '/foo/bar/baz' # => will use 'foo/bar#baz' as endpoint
Not any path with level two or more of nesting can be used as shortcut.
If path contains any characters outside of /[\w-]/ then it can't be
used as such.
This commit ensures that invalid shortcuts aren't used.
':controller/:action/postfix' - is an example of invalid shortcut
that was previosly matched and led to exception:
"ArgumentError - ':controller/:action' is not a supported controller name"