This PR was merged into the 5.3-dev branch.
Discussion
----------
[Serializer] Add a Custom End Of Line in CSV File
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| License | MIT
| Doc PR | -
Commits
-------
20f03677e3 [Serializer] Add a Custom End Of Line in CSV File
This PR was merged into the 5.3-dev branch.
Discussion
----------
Add `#[As-prefix]` to service attributes
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
This PR renames all existing attributes with the `As` prefix, as I proposed several times already.
This should help autocompletion, and it's required to not collide with existing class names (eg the `Command` class, but also the old `Controller` class, etc.)
I think this `As` prefix is a convention for the better.
Commits
-------
4f1318963a Add `#[As-prefix]` to service attributes
This PR was merged into the 5.3-dev branch.
Discussion
----------
[HttpKernel] Add `#[AsController]` attribute for declaring standalone controllers on PHP 8
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
This PR adds an `#[AsController]` attribute to allow autoconfiguring controllers even when they don't extend `AbstractController`. This should allow removing the line about `controller.service_arguments` in `services.yaml` to cover this need (but that is noise to most.)
Commits
-------
437f5feb45 [HttpKernel] Add `#[AsController]` attribute for declaring listeners on PHP 8
* 5.2:
[travis] use packagist API v2
[translation] Fix update existing key with existing +int-icu domain
[Contracts] Fix branch name in README.md links
* 4.4:
[travis] use packagist API v2
[translation] Fix update existing key with existing +int-icu domain
[Contracts] Fix branch name in README.md links
This PR was merged into the 4.4 branch.
Discussion
----------
[Translation] Fix update existing key with existing +int-icu domain
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | none
| License | MIT
| Doc PR | none
Using php-translation webui interface, I was trying to update an existing +int-icu domain but a regular domain was created and the existing key wasn't updated. Looks like in the method I modified, we should have been looking for a potential domain+intl-icu domain and try update it first
Commits
-------
2a196ca0dc [translation] Fix update existing key with existing +int-icu domain
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Notifier] Move abstract test cases to namespace
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no (only fixes builds)
| New feature? | yes (not sure about a changeling entry)
| Deprecations? | no
| Tickets | Fix #... <!-- prefix each issue number with "Fix #", no need to create an issue if none exist, explain below instead -->
| License | MIT
| Doc PR | no
This enables BC promise for the abstract test classes, like for other ones.
As we ignore current `/Tests` directory in `.gitattributes`, builds will fail if `symfony/notifier` is installed as dist.
Commits
-------
33ba3c755d [Notifier] Move abstract test cases to namespace
This PR was merged into the 5.2 branch.
Discussion
----------
Fixed parsing deprecated definitions without message key
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
### Steps to reproduce
1. Create service/alias definition and mark it as deprecated without specifying message key as shown in https://symfony.com/doc/5.2/service_container/alias_private.html#deprecating-service-aliases. For example:
```yaml
services:
service_without_deprecation_message:
class: Foo
deprecated:
package: vendor/package
version: 1.1
alias_without_deprecation_message:
alias: foobar
deprecated:
package: vendor/package
version: 1.1
```
2. Clear cache by running `php bin/console cache:clear`
### Actual behaviour
`Symfony\Component\Config\Exception\LoaderLoadException` exception is thrown with the following message: `Warning: Undefined array key "message" (...)`
See https://gist.github.com/adamwojs/5bac85d4326d8fbe6f52a357fa6e57e6 for full stactrace.
### Excepted behaviour
Given definitions are loaded correctly. Default service/alias deprecation message is used as shown in linked documentation page.
Commits
-------
d585b64953 Fixed parsing deprecated definitions without message key
This PR was merged into the 4.4 branch.
Discussion
----------
[Security] Handle properly 'auto' option for remember me cookie security
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40471
| License | MIT
| Doc PR | n/a
Manually setting remember_me cookie secure as auto is still is being set as secure one even if used over http.
This PR fixes this behaviour by converting auto to null prior setting it up for service.
Commits
-------
2bcf69c071 [Security] Handle properly 'auto' option for remember me cookie security
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[Validator] Avoid triggering the autoloader for user-input values
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
Following-up to https://twitter.com/seldaek/status/1372450636361502721 - mostly to see if the build passes or if this breaks some undocumented/unclear-to-me assumptions.
Essentially using the `Valid` constraint should only validate objects if they exist as objects. If a user sends a string and that gets assigned to a property, `Valid` should not attempt autoloading that user-given string.
As far as I can tell, this is used in two places:
- acb32dd396/src/Symfony/Component/Validator/Validator/RecursiveContextualValidator.php (L364-L365) where non-objects are anyway ignored, so this change is harmless there.
- acb32dd396/src/Symfony/Component/Validator/Validator/RecursiveContextualValidator.php (L652-L660) where it's explicitly passing anything in there to get the proper exception, so my change makes sure that exception is thrown *before* autoloading attempts. I am just not 100% sure if there are cases where validateGenericNode will receive a class name as a string to validate in $value. I can't imagine why it would but that doesn't mean it's true.
Commits
-------
e45eb23ba2 [Validator] Avoid triggering the autoloader for user-input values
This PR was merged into the 5.2 branch.
Discussion
----------
[FrameworkBundle] ensure TestBrowserToken::$firewallName is serialized
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | n/a
| License | MIT
| Doc PR | n/a
Fixes a bug introduced in #40368. When `TestBrowserToken` in unserialized, `$firewallName` is `null` so trying to access it via `getFirewallName` throws a `TypeError`.
Commits
-------
8ba12ece57 [FrameworkBundle] ensure TestBrowserToken::$firewallName is serialized
This enables BC promise for the abstract test classes, like for other ones.
As we ignore current `/Tests` directory in `.gitattributes`, builds will fail if `symfony/notifier` is installed as dist.
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Runtime] make GenericRuntime ... generic
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
This PR will allow #40436 to move to https://github.com/symfony/psr-http-message-bridge
For the record, it builds on a prototype I wrote almost one year ago at https://github.com/tchwork/bootstrapper.
This PR makes the `GenericRuntime` implementation able to auto-discover runtime implementations for specific types.
It uses the autoloader for the discovery: when a closure-app requires or returns a type `Vendor\Foo`, it will use a convention and check if the class `Symfony\Runtime\Vendor\FooRuntime` exists. If yes, it will use it to resolve the corresponding type. Such runtime classes have to extend `GenericRuntime` so that they can use the protected API it provides. This requirement is aligned with the fact that the very convention proposed here is an implementation detail that works when using a `GenericRuntime` as the main runtime (This behavior can be overridden by providing explicit entries in the new `runtimes` option when booting the `GenericRuntime`.)
`SymfonyRuntime` can be used as both the main runtime or a type-specific runtime:
- when used as the main runtime, it configures the typical global-state for Symfony and has a fast codepath for Symfony types, while still being generic.
- it can also be used in another runtime as a way to resolve Symfony types (would typically be useful to #40436 for running Console apps in a PSR-based web app.)
Commits
-------
33e371e24d [Runtime] make GenericRuntime ... generic
* 5.2:
[HttpKernel] do is_file check before include
[PhpUnitBridge] fix reporting deprecations from DebugClassLoader
[FrameworkBundle] skip deprecation in integration tests
enable HTTP method overrides as early as possible with the HTTP cache
* 4.4:
[HttpKernel] do is_file check before include
[PhpUnitBridge] fix reporting deprecations from DebugClassLoader
[FrameworkBundle] skip deprecation in integration tests
This PR was merged into the 4.4 branch.
Discussion
----------
[HttpKernel] Add is_file() check before include
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes (?)
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
I recently noticed that on every deployment I got warnings reported by sentry from the `cache:clear` command.
> Warning: include(/var/www/html/var/cache/pro_/App_KernelProdContainer.php): failed to open stream: No such file or directory
In `Symfony\Component\HttpKernel\Kernel` [line 469](https://github.com/symfony/symfony/blob/5.x/src/Symfony/Component/HttpKernel/Kernel.php#L469)
(on 4.4 it's on [line 536](https://github.com/symfony/symfony/blob/4.4/src/Symfony/Component/HttpKernel/Kernel.php#L536))
This is because the code tries to include the `$cachePath` without checking if it exists/is a file first. It seems like something similar was fixed some time ago already (https://github.com/symfony/symfony/pull/27249) some lines above.
This PR fixes the mentioned warnings for me.
I'm running on Symfony 5.2.5 at the moment, but the docs said that bugfixes should target branch 4.4 and new features should target 5.x, and I guess this could rather be seen as a bugfix than a new feature, so I branched off of 4.4 and also set this as target for the PR, I hope that's correct.
Commits
-------
8efcc17573 [HttpKernel] do is_file check before include
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[Form] Add "form_attr" FormType option
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? |no
| New feature? | yes
| Deprecations? | no
| Tickets | N/A
| License | MIT
| Doc PR | [#15108](https://github.com/symfony/symfony-docs/pull/15108)
## What is this about
This PR add support for [`form` attribute](https://html.spec.whatwg.org/multipage/form-control-infrastructure.html#attr-fae-form) to Symfony Form ([browser compatibility](https://caniuse.com/form-attribute)).
The `form` attribute allows form elements to override their associated form (which is their nearest ancestor form element by default). This is extremely useful to solve **nested form problem** and allows **form children** to be **rendered outside form tag** while still working as expected.
## New "form_attr" FormType option
#### form_attr
**type**: `bool` or `string` **default**: `false`
If set to `true` on a **root form**, adds [`form` attribute](https://html.spec.whatwg.org/multipage/form-control-infrastructure.html#attr-fae-form) on every children with their root **form id**.
This allows you to render form children outside the form tag and avoid **nested form problem** in some situations while keeping the form working properly.
If set to `true` on a **child**, adds [`form` attribute](https://html.spec.whatwg.org/multipage/form-control-infrastructure.html#attr-fae-form) on it with its **root form id**.
This allows you to render **that child** outside the form tag and avoid **nested form problem** in some situations while keeping the form working properly.
If root form has no `id` (this may happen by create an *unnamed* form), you can set it to a `string` identifier to be used at `FormView` level to link children and root form anyway.
## Usage on Root Form Example
#### Form Type
Enable the feature by setting `form_attr` to `true` on the root form.
```php
use Symfony\Component\Form\Extension\Core\Type;
class ListFilterType extends AbstractType
{
public function buildForm(FormBuilderInterface $builder, array $options): void
{
$builder
->add('search', Type\SearchType::class)
->add('orderBy', Type\ChoiceType::class, [
'choices' => [
// ...
],
])
->add('perPage', Type\ChoiceType::class, [
'choices' => [
// ...
],
])
}
public function configureOptions(OptionsResolver $resolver): void
{
$resolver->setDefault('form_attr', true); // <--- Set this to true
}
}
```
#### Twig
The following Twig template **works properly** even if form children are **outside** their form tag ([browser compatibility](https://caniuse.com/form-attribute)).
```twig
<div class="header-filters">
{{ form_errors(form) }}
{{ form_row(form.search) }} {# has attribute form="list_filter" #}
{{ form_row(form.orderBy) }} {# has attribute form="list_filter" #}
</div>
<!-- -->
<!-- Some other HTML content, like a table or even another Symfony form -->
<!-- -->
<div class="footer-filters">
{{ form_row(form.perPage) }} {# has attribute form="list_filter" #}
</div>
{{ form_start(form) }} {# id="list_filter" #}
{{ form_end(form) }}
```
✅ Every form elements work properly even outside form tag.
## Usage on Form Child Example
Enable the feature by setting `form_attr` to `true` on selected child.
```php
use Symfony\Component\Form\Extension\Core\Type;
class ListFilterType extends AbstractType
{
public function buildForm(FormBuilderInterface $builder, array $options): void
{
$builder
->add('search', Type\SearchType::class)
->add('orderBy', Type\ChoiceType::class, [
'choices' => [
// ...
],
])
->add('perPage', Type\ChoiceType::class, [
'form_attr' => true, // <--- Set this to true
'choices' => [
// ...
],
])
}
}
```
#### Twig
The following Twig template **works properly** even if `form.perPage` is **outside** form tag ([browser compatibility](https://caniuse.com/form-attribute)).
```twig
<div class="header-filters">
{{ form_start(form) }} {# id="list_filter" #}
{{ form_errors(form) }}
{{ form_row(form.search) }}
{{ form_row(form.orderBy) }}
{{ form_end(form, {'render_rest': false}) }}
</div>
<!-- -->
<!-- Some other HTML content, like a table or even another Symfony form -->
<!-- -->
<div class="footer-filters">
{{ form_row(form.perPage) }} {# has attribute form="list_filter" #}
</div>
```
✅ `form.perPage` element work properly even outside form tag.
Commits
-------
5f913cec74 [Form] Add "form_attr" FormType option
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[Validator] Add normalizer option to Unique constraint
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | #37451
| License | MIT
| Doc PR | todo
Hello! Reopening my PR #37507 with target branch `5.x`
This PR is about https://github.com/symfony/symfony/issues/37451. The idea is to make that constraint more flexible and able to process business rules, especially when working with objects. We can think about it as the similar feature in UniqueEntity constraint, when we declare on which attributes we are applying the constraint. But in our case it is more general - we pass a callable with whatever logic we want to apply to collection elements before we apply 'uniqueness check' :)
Looks like no BC breaks. Thanks! :)
Commits
-------
44e1e8bc9b [Validator] Add normalizer option to Unique constraint
This PR was merged into the 5.3-dev branch.
Discussion
----------
[DebugBundle] Remove warning of ServerDumpPlaceholderCommand in console
| Q | A
| ------------- | ---
| Branch? | 5.2.5
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40495
| License | MIT
| Doc PR |
In 5.2.5 the console commands are lazy. (https://github.com/symfony/symfony/pull/39851)
With this change the ServerDumpCommand::$defaultName is used which isn't set in the placeholder command.
If no vardump-server in debug.dump_destination is defined, this will lead to a warning and not adding the command to the console list.
Commits
-------
37b2a19989 [DebugBundle] Add $defaultName to PlaceholderCommand
* 5.2:
[Console] Fix line wrapping for decorated text in block output
[Inflector] Fixed pluralize "coupon"
[PhpUnitBridge] fix compat with symfony/debug
[VarDumper] Adds support for ReflectionUnionType to VarDumper
Correctly clear lines for multi-line progress bar messages.
[Security] Add XML support for authenticator manager
* 4.4:
[Console] Fix line wrapping for decorated text in block output
[Inflector] Fixed pluralize "coupon"
[PhpUnitBridge] fix compat with symfony/debug
[VarDumper] Adds support for ReflectionUnionType to VarDumper
Correctly clear lines for multi-line progress bar messages.
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[Console] Fix line wrapping for decorated text in block output
| Q | A
| ------------- | ---
| Branch? |
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| License | MIT
Fixed bug that caused decorated text to be wrapped too early in SymfonyStyle->block().
Commits
-------
760be88310 [Console] Fix line wrapping for decorated text in block output
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Runtime] Remove "docs" from readme
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR | https://github.com/symfony/symfony-docs/pull/15081
Make the readme of the Runtime component link to the docs. This PR is blocked by the doc PR.
Commits
-------
6f4552fcec [Runtime] Remove "docs" from readme
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[VarDumper] Adds support for ReflectionUnionType to VarDumper
Fixes a bug with VarDumper when dumping a ReflectionUnionType.
> PHP Error: Call to undefined method ReflectionUnionType::isBuiltin()
Notes:
* One of the existing tests relies on its position in the test file. I had to modify its expected line number.
* There is an existing trailing space around line 367 in an expected value.
I'm not sure if this was left for BC reasons but it seems like a bug if the dumper is leaving trailing spaces.
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
This commit fixes a crash when dumping ReflectionUnionType. The code is minimal but uses a few extra lines to preserve key order for bc and consistency. Additionally, there is an else condition that is currently unreachable but is defensive should they add additional subtypes of ReflectionType. Tests are included. Please let me know if you have any questions or suggestions. Thanks for Symfony it's a wonderful project.
Commits
-------
1a11491f6e [VarDumper] Adds support for ReflectionUnionType to VarDumper
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Mailer] Add supported auth modes to exception
| Q | A
| ------------- | ---
| Branch? | 4.4+
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
If there is no match for the available authenticators the thrown exception now includes the supported auth methods as reported by the server.
Commits
-------
d9c99d86ab [Mailer] Add supported auth modes to exception
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Security] Fix BC layer
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
Tests were broken because no password hasher was configured for the new `InMemoryUser` class in configs on lower branches.
This PR fixes it by making the new class extend the deprecated one, so that the password hasher factory is able to find a password hasher for it (instanceof-based match).
Commits
-------
2d7f7b5072 [Security] Fix BC layer
This PR was merged into the 5.3-dev branch.
Discussion
----------
[DependencyInjection] accept null index in #[TaggedItem]
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
As hinted by @stof in https://github.com/symfony/symfony/pull/40248#discussion_r595065941
Commits
-------
6d16fac703 [DI] accept null index in #[TaggedItem]
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Security] Rename User to InMemoryUser
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | yes
| Tickets | Closes#26348
| License | MIT
| Doc PR | -
This PR aims to clarify that the `User` class should only be used by the `InMemoryUserProvider`, as documented:
c06a76c384/src/Symfony/Component/Security/Core/User/User.php (L15-L17)
It also renames `UserChecker` to `InMemoryUserChecker` because it only works with the in-memory user class:
c06a76c384/src/Symfony/Component/Security/Core/User/UserChecker.php (L31-L32)
Commits
-------
55b51d3f90 [Security] Rename User to InMemoryUser
This PR was merged into the 5.3-dev branch.
Discussion
----------
Deprecate configuring tag names and service ids in compiler passes
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | -
| License | MIT
| Doc PR | -
This PR is aimed at reducing the code complexity by hardcoding the name of tags and service ids that compiler passes have to deal with.
I think making these names configurable only adds boilerplate and maintenance overhead for no benefit:
- for the practice: the need to use a pass with a renamed tag/id should be extremely rare (do yo know any?)
- for the theory: a decorating pass could still rename before/after the processing, so this use case is still supported.
Side note: I skipped updating changelog+upgrade files. This would be just noise to me (nobody uses this possibility anyway ;) )
Commits
-------
6fe82d8be7 Deprecate configuring tag names and service ids in compiler passes
* 5.2:
[ErrorHandler] Fix error caused by `include` + open_basedir
[FrameworkBundle] Make the TestBrowserToken interchangeable with other tokens
[Console] ProgressBar clears too many lines on update
[FrameworkBundle] Exclude unreadable files when executing About command
[Bridge\Twig] Add 'form-control-range' for range input type
Be explicit about transparent background color of links in toolbar
[Translation] fix test case name
[Cache] Fix wrong namespace in test
[DependencyInjection] Fix return type
* 4.4:
[Console] ProgressBar clears too many lines on update
[FrameworkBundle] Exclude unreadable files when executing About command
[Bridge\Twig] Add 'form-control-range' for range input type
Be explicit about transparent background color of links in toolbar
[Translation] fix test case name
[Cache] Fix wrong namespace in test
[DependencyInjection] Fix return type
This PR was squashed before being merged into the 5.2 branch.
Discussion
----------
[FrameworkBundle] Make the TestBrowserToken interchangeable with other tokens
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
For some reason most tokens implement getFirewallName (and the deprecated getProviderKey), but not all. This adds it to TestBrowserToken because it caused issues in my tests when upgrading to Symfony 5's `Client::loginUser()`.
I am also happy to add this to AnonymousToken but I am more worried about BC impacts there, so I left it alone for now. All other token classes implement it, so I am not sure why this isn't in TokenInterface to begin with.
Commits
-------
a4958ae7ad [FrameworkBundle] Make the TestBrowserToken interchangeable with other tokens
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[Console] ProgressBar clears too many lines on update
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? |no <!-- please update src/**/CHANGELOG.md files -->
| Deprecations? |no <!-- please update UPGRADE-*.md and src/**/CHANGELOG.md files -->
| License | MIT
The ProgressBar incorrectly calculates line lengths when updating, including non-displayable characters such as ANSI colors. This causes it to clear too many lines if the terminal width is greater than the displayed line length but less than the line length including non-displayed characters. An example of this bug in action is https://github.com/acquia/cli/issues/467
Commits
-------
2aa3df0c74 [Console] ProgressBar clears too many lines on update