Most importantly, the original request thread must yield its share lock
while waiting for the live thread to commit -- otherwise a request's
base and live threads can deadlock against each other.
Referencing Rails.env without checking if it's defined couples
us to Railties.
Fix by avoiding the line breaks if we don't have an env check
to rely on.
Tests can (and do) access the database from the main thread. In this
case they were starting a transaction, then making a request. The
request would create a new thread, which would allocate a new database
connection. Since the main thread started a transaction that contains
data that the new thread wants to see, the new thread would not see it
due to data visibility from transactions. Spawning the new thread in
production is fine because middleware should not be doing database
manipulation similar to the test harness. Before 603fe20c it was
possible to set the database connection id based on a thread local, but
603fe20c changes the connection lookup code to never look at the
"connection id" but only at the thread object itself. Without that
indirection, we can't force threads to use the same connection pool as
another thread.
Fixes#23483
Some places were saying filter, while others said filter_options, spare the ambiguity
and use filter throughout.
This inlines a needless local variable and clarifies a route filter consists of defaults
and values to match against.
Assume the filter is a string, if it wasn't a hash and isn't nil. Remove needless else
and rely on Ruby's default nil return.
Add spaces within hash braces.
Add two options: `-c` and `-g`.
`-g` option returns the urls name, verb and path fields that match the pattern.
`-c` option returns the urls for specific controller.
Fixes#18902, and Fixes#20420
[Anton Davydov & Vipul A M]
There are some cases when the `body` in `response_body=` can be set to
nil. One of those cases is in `actionpack-action_caching` which I found
while upgrading it for Rails 5.
It's not possible to run `body.each` on a `nil` body so we have to
return after we run `response.reset_body!`.
we never use this custom array outside the mime type `parse` method. We
can reduce the interaction to just a regular array, so we should use
that instead (IOW, there was nothing special about AcceptList so we
should remove it).
This works on OSX but for some reason travis is throwing a
```
1) Error:
ExpiresInRenderTest#test_dynamic_render_with_absolute_path:
NoMethodError: undefined method `unlink' for nil:NilClass
```
Looking at other tests in Railties the file has a name and we close
it before unlinking, so I'm going to try that.
Test that we are not allowing you to grab a file with an absolute path
outside of your application directory. This is dangerous because it
could be used to retrieve files from the server like `/etc/passwd`.
Remove nonsense definition of == from `AcceptItem`. The definition only
compared names and not `q` values or even object identity. The only use
was in the `assort!` method that really just wanted the index of the
item given the item's name. Instead we just change the caller to use
`index` with the block form.
The cache for `render file:` seems to also be used in the case of
`render(string)`. If one is supposed to be a hit and the other is
supposed to be a miss, and they both reference the same file, then the
cache could return incorrect values. This commit clears the cache
between runs so that we get non-cached behavior.
* 5-0-beta-sec:
bumping version
fix version update task to deal with .beta1.1
Eliminate instance level writers for class accessors
allow :file to be outside rails root, but anything else must be inside the rails view directory
Don't short-circuit reject_if proc
stop caching mime types globally
use secure string comparisons for basic auth username / password
- Fixes bug #23142.
- Bug was occurring only with ActionController::API, because `_process_options` wasn't being run for API requests, even though it was being run for normal app requests.
The Mapper build_path method marks routes where path parameters are part
of a path segment as custom routes by altering the regular expression, e.g:
get '/foo-:bar', to: 'foo#bar'
There were some edge cases where certain constructs weren't being picked
up and this commit fixes those.
Fixes#23069.
This reverts commit 5d1b7c3b441654e8008dcd303f5367883ec660a6.
The change here didn't actually fix the issue it was trying to fix, and
this isn't the correct way to fix either issue. The problem is switching
from the builder to grouping with find_all/regex is now very dependent
on how you structure your path pattern.
We can provide a more flexible upgrade experience by warning users they are using unsafe methods instead of forcing the safe API by deprecating before removal. This PR provides this functionality.
In commit d993cb3 `build_path` was changed from using `grep` to
`find_all` to save array allocations.
This change was a little too aggressive in that when the dash comes
before the symbol like `/omg-:song` the symbol is skipped.
Removing the check for `n.right.left.literal?` fixes this issue, but
does add back some allocations. The number of allocations are still well
less than before.
I've added a regression test to test this behavior for the future.
Fixes#23069.
Array allocations as of d993cb3:
```
{:T_SYMBOL=>11}
{:T_REGEXP=>17}
{:T_STRUCT=>6500}
{:T_MATCH=>12004}
{:T_OBJECT=>91009}
{:T_DATA=>100088}
{:T_HASH=>114013}
{:T_STRING=>159637}
{:T_ARRAY=>321056}
{:T_IMEMO=>351133}
```
Array allocations after this change:
```
{:T_SYMBOL=>11}
{:T_REGEXP=>1017}
{:T_STRUCT=>6500}
{:T_MATCH=>12004}
{:T_DATA=>84092}
{:T_OBJECT=>87009}
{:T_HASH=>110015}
{:T_STRING=>166152}
{:T_ARRAY=>322056}
{:T_NODE=>343558}
```
Instead of checking whether the class has recycle! or not, we can just
always add the method to all controller classes when the test harness is
loaded. Technically this means that the controller test harness will
not work with controllers that do not inherit from AC::Metal, but then,
I'm not sure that is supported anyway.
Mixing in the module one will ensure that we don't break method caches,
and eliminates a runtime check so it should speed up tests (slightly).
- `CONTROLLER` argument can now be supplied in different ways (Rails::WelcomeController, Rails::Welcome, rails/welcome)
- If `CONTROLLER` argument was supplied but it does not exist, will warn the user that this controller does not exist
- If `CONTROLLER` argument was supplied and no routes could be found matching this filter, will warn the user that no routes were found matching the supplied filter
- If no routes were defined in the config/routes.rb file, will warn the user with the original message
- The root method is defined and documented in Base module and
decorated in Resources module.
- The documentation in Base module actually talks about method
signature of decorated method from Resources module.
- Argument handling was moved to decorated method in
977455cc2e
to handle options such as :as with directly passed path parameter.
- To avoid the confusion, removed original root method from Base module
and only kept overridden version in Resources module.
- References - https://github.com/rails/rails/pull/12208 &
https://github.com/rails/rails/pull/12208#issuecomment-24350897.
- Earlier only Hash was allowed as params argument to url_helpers.
- Now ActionController::Parameters instances will also be allowed.
- If the params are not secured then it will raise an ArgumentError to
indicate that constructing URLs with non-secure params is not recommended.
- Fixes#22832.
- AC::Parameters#convert_parameters_to_hashes should return filtered or
unfiltered values based on whether it is called from `to_h` or `to_unsafe_h`
instead of always defaulting to `to_h`.
- Fixes#22841
The `cache` helper methods should pass any extra options
to the cache store. For example :expires_in would be a
valid option if memcache was the cache store. The change
in commit da16745 broke the ability to pass any options
other than :skip_digest and :virtual_path. This PR
restores that functionality and adds a test for it.
During the `5.0.0.beta1` release, the CHANGELOGs got an entry like the
following:
```
* No changes.
```
It is kinda confusing as there are indeed changes after it. Not a
biggie, just a small pass over the CHANGELOGs.
[ci skip]
The ActionPack test suite had a handful of these warnings when run. This
was due to `assert_response` being tested outside the context of a
controller instance where those instance variables would already have
been initialized.
If you're not familiar with how the `Referer` header works, you likely
won't understand why you need to provide a fallback or under what
circumstances it would be used.
Hopefully this clarifies things a bit.
When calling `to_h` on an `ActionController::Parameters` instance it would
`deep_dup` its internal parameters.
This inadvertently called `dup` on a passed Active Record model which would
create new models. Fix by only dupping Ruby's Arrays and Hashes.
Applications that use `redirect_to :back` can be forced to 500 by
clients that do not send the HTTP `Referer` (sic) header.
`redirect_back` requires the user to consider this possibility up front
and avoids this trivially-caused application error.
`redirect_to :back` is a somewhat common pattern in Rails apps, but it
is not completely safe. There are a number of circumstances where HTTP
referrer information is not available on the request. This happens often
with bot traffic and occasionally to user traffic depending on browser
security settings.
When there is no referrer available on the request, `redirect_to :back`
will raise `ActionController::RedirectBackError`, usually resulting in
an application error.
`redirect_back` takes a required `fallback_location` keyword argument
that specifies the redirect when the referrer information is not
available. This prevents 500 errors caused by
`ActionController::RedirectBackError`.
Introduced in e56c63542780fe2fb804636a875f95cae08ab3f4, `CacheHelper#fragment_cache_key` is a duplicate of `ActionController::Caching::Fragments#fragment_cache_key`.
We now require the view to provide this method on its own (as with `view_cache_dependencies`); `ActionController::Caching::Fragments` exports its version as a `helper_method`.
In Rails 5.1 `ActionController::TestCase` will be moved out of Rails
into it's own gem.
Please use `ActionDispatch::IntegrationTest` going foward.
Because this will be moved to a gem I used `# :stopdoc:` instead of
deleting the documentation. This will remove it from the Rails
documentation but still leave the method documented for when we move it
to a gem.
Guides have been updated to use the routing structure used in Integration
and all test examples have been updated to inherit from
`ActionDispatch::IntegrationTest` instead of `ActionController::TestCase.
Fixes#22496
Follow-up to PR #19977, which helpfully added the redirection path to the error message of assert_response if response is a redirection, but which removed the response code, obscuring the type of redirect.
This PR:
- brings back the response code in the error message,
- updates the tests so the new messages can be tested,
- and adds test cases for the change.
I did this change but it is affecting how the request params end up
after being processed by the router.
To be in the safe side, I just take the format from the extension in the
URL when is not present in those params and it's being used only for the
`Request#formats` method
All of our tests were testing the `ActionController::Live` behavior in a
standalone environment, without going through the router or behaving
like a real application.
This resulted in `ActionController::Live` throwing the exception
`undefined method 'request' for #<ActionDispatch::Request:0x00000003ad1148>`
because `make_response!` was expecting a response instead of a request.
The expectation of a response came from `set_response!` in non-router
tests setting the response and passing it to `make_response!`. In the
case of an application we would hit `serve` in `RouteSet` first which
would send us to `make_response!` with a request sent instead of a
response.
The changes here remove `set_response!` so `make_response!` always
receives a request.
Thanks to KalabiYau for help with the investigation and solution.
Fixes#22524
[Eileen M. Uchitelle & KalabiYau]
Per this comment
https://github.com/rails/rails/pull/18334#issuecomment-69234050 we want
`protect_from_forgery` to default to `prepend: false`.
`protect_from_forgery` will now be insterted into the callback chain at the
point it is called in your application. This is useful for cases where you
want to `protect_from_forgery` after you perform required authentication
callbacks or other callbacks that are required to run after forgery protection.
If you want `protect_from_forgery` callbacks to always run first, regardless of
position they are called in your application, then you can add `prepend: true`
to your `protect_from_forgery` call.
Example:
```ruby
protect_from_forgery prepend: true
```
We don't want to commit the cookie jar more than once because then we
will be attempting to modify a frozen hash.
Fixes Railties test failure caused by 492b134.
This adds a new method to request and response so we don't need to
violate the law of demeter.
We are changing `Request` and `Response` so that they always have a
`cookie_jar`
This is a continuation on work to combine integration and controller
test code bases in Rails.
We want to get rid of the `Live::Response` so we are consolidating methods
from `Live::Response` and `Response` by merging them together.
This adds an `#empty` method to the request so we don't need to
hard-code the empty array each time we call an empty
`ActionDispatch::Request`.
The work here is a continuation on combining controller and integration
test code bases into one.
According to pr #22443 in the guides there's always a dollar sign before every command, so why is in the main README a `$` and in every submodule a `%`?
Just eye candy..
- `assert_predicate` appends its own error message at the end of message
generated by `assert_response` and because of that the error message
displays the whole `response` object.
- For eg.
Expected response to be a <success>, but was a redirect to <http://test.host/posts>.
Expected #<ActionDispatch::TestResponse:0x007fb1cc1cf6f8....(lambda)>}>> to be successful?.
- Complete message can be found here -
https://gist.github.com/prathamesh-sonpatki/055afb74b66108e71ded#file-gistfile1-txt-L19.
- After this change the message from `assert_predicate` won't be
displayed and only message generated by `assert_response` will be shown
as follows:
Expected response to be a <success>, but was a redirect to <http://test.host/posts>
There were two places where the tld_length default
was hard-coded to 1, both overriding the real default
value of ActionDispatch::Http::URL.tld_length in this
set of tests.
This commit removes both of those, relying on the
actual value of ActionDispatch::Http::URL.tld_length,
unless it's specifically overridden.
- If the assert_response is checking for any non-redirect response like
:success and actual response is :redirect then, the error message displayed
was -
Expected response to be a <success>, but was <302>
- This commit adds the redirect path to the error message of
assert_response if the response is :redirect.
So above message is changed to -
Expected response to be a <success>, but was a redirect to <http://test.host/posts/lol>
Closes#21230 by following the indication of @rafaelfranca:
> I think the output change would be simpler.
> What is really important to show is the class of the middleware, so we should change the output to show that.
The configuration for `config.static_cache_control`, and its replacement
`config.public_file_server.headers` are implemented in Railties.
People would configure this in environment files, which is Railties domain too.
When a request is made with AJAX and an error occurs, Rails will render
a text-template for the exception instead of the HTML error page
(#11960).
The `.text.erb` variant of the `_source` template is currently missing,
causing HTML to be rendered in the response. This commit adds the text
template.
To keep the page scannable we only only show the first three source
extracts.
Related to #14745.
Before:
```
~/testing-exceptions ᐅ curl 'http://localhost:3000/' -H
'X-Requested-With: XMLHttpRequest'
RuntimeError in PostsController#index
<div class="source " id="frame-source-0">
<div class="info">
Extracted source (around line <strong>#3</strong>):
</div>
<div class="data">
<table cellpadding="0" cellspacing="0" class="lines">
<tr>
```
After:
```
~/testing-exceptions ᐅ curl 'http://localhost:3000/' -H
'X-Requested-With: XMLHttpRequest'
RuntimeError in PostsController#index
Extracted source (around line #3):
*3 raise
```
It used to behave like this:
url_for(controller: 'x', action: 'y', q: {})
# -> "/x/y?"
We previously avoided empty query strings in most cases by removing
nil values, then checking whether params was empty. But as you can
see above, even non-empty params can yield an empty query string. So
I changed the code to just directly check whether the query string
ended up empty.
(To make everything more consistent, the "removing nil values"
functionality should probably move to ActionPack's Hash#to_query, the
place where empty hashes and arrays get removed. However, this would
change a lot more behavior.)
rather than an action name and *args. The *args were not being used in regular
applications outside tests. This causes a backwards compatibility
issue, but reduces array allocations for most users.
`dispatch` sets the request and response on the controller for us
automatically, so the test harness doesn't need to know the internals of
how request / response is set.
Conflicts:
actionpack/lib/action_controller/test_case.rb
For ActionController::Base we write the cookies in a middleware if it
was not yet committed no matter if the response was committed or not. [1]
For ActionController::Live we write the cookies before the response is
committed. [2]
We already mimic ActionController::Live in
ActionController::TestCase but we don't mimic the ActionController::Base
behavior because we were checking if the response was committed before
writing the cookies.
Now we are matching the behavior of the middleware and writing the
cookies if it was not written before.
[1]: 80c6b901d4/actionpack/lib/action_dispatch/middleware/cookies.rb (L599-L604)
[2]: 80c6b901d4/actionpack/lib/action_controller/metal/live.rb (L218-L223)
This commit follows up of ea9bc06c9a47b839d5e2db94ba6bf7e29c8f0ae9.
To check `@response.content_type.to_s` is ended with `"xml"`,
to use `\z` is sufficient.
Prior to this change, given a route:
# config/routes.rb
get ':a' => "foo#bar"
If one pointed to http://example.com/%BE (param `a` has invalid encoding),
a `BadRequest` would be raised with the following non-informative message:
ActionController::BadRequest
From now on the message displayed is:
Invalid parameter encoding: hi => "\xBE"
Fixes#21923.
When an application has multiple root entries with different
constraints, the current solution is to use `get '/'`. Example:
**Currently I have to do:**
```ruby
get '/', to: 'portfolio#show', constraints: ->(req) { Hostname.portfolio_site?(req.host) }
get '/', to: 'blog#show', constraints: ->(req) { Hostname.blog_site?(req.host) }
root 'landing#show'
```
**But I would like to do:**
```ruby
root 'portfolio#show', constraints: ->(req) { Hostname.portfolio_site?(req.host) }
root 'blog#show', constraints: ->(req) { Hostname.blog_site?(req.host) }
root 'landing#show'
```
Other URL matchers such as `get`, `post`, etc, already allows this, so I
think it's fair that `root` also allow it since it's just a shortcut for
a `get` internally.
