This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[FrameworkBundle] AnnotationsCacheWarmer should support doctrine/annotations:^1.13
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40353
| License | MIT
| Doc PR |
The `AnnotationsCacheWarmer` is creating a cached a `new CachedReader` in the `doWarmUp()`. But this will fail if doctrine/cache is not installed. In #40338 I added the `kernel.cache_warmer` tag on `annotations.cache_warmer` service even though the doctrine/cache is not installed.
This PR will make sure `AnnotationsCacheWarmer` is using the `PsrCacheReader` when available.
This bug was not found in the tests because doctrine/cache is always installed.
----------
Big golden star to @jrushlow because you test dev-master.
Commits
-------
fb1cc72b18 [FrameworkBundle] AnnotationsCacheWarmer should support doctrine/annotations:^1.13
This PR was merged into the 4.4 branch.
Discussion
----------
[Security] Refresh original user in SwitchUserListener
Fixes#39991
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#39991
| License | MIT
| Doc PR | NA
Fix SwitchUserListener to update original token with refreshed user. This prevents a non-refreshed user from causing problems elsewhere, such as in Voters.
Commits
-------
42453454c9 Refresh original user in SwitchUserListener.
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[WebProfilerBundle] Disable CSP if dumper was used
| Q | A
| ------------- | ---
| Branch? | 5.x for features
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | Fix#29084
| License | MIT
| Doc PR | -
Disables a configured Content Security Policy if the dumper was used to avoid breaking the toolbar. This is a less invasive alternative to #29155 which fixes#29084 (there is a project with a test case linked there).
Commits
-------
5dc2637263 [WebProfilerBundle] Disable CSP if dumper was used
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[twig-bridge] Allow NotificationEmail to be marked as public
| Q | A
| ------------- | ---
| Branch? | 5.x for features
| Bug fix? |no
| New feature? | yes
| Deprecations? |no
| Tickets | Fix#37879
| License | MIT
| Doc PR | -
Closes https://github.com/symfony/symfony/issues/37879
`NotificationEmail` can be created public from the `asPublicEmail` method in any context but with the Notifier component. In this case it is created by it as not public and therefore displays the importance in the subject of the email.
For end users, aka not admin, the importance in the subject's email is not necessary.
This PR will allow if needed to set a `NotificationEmail` as public even after being created, wherever it was created.
I am not sure how to handle the version in the changelog.
For the tests I don't know what's the policy so I just added a new case in each test case.
Commits
-------
8f753d27f2 [twig-bridge] Allow NotificationEmail to be marked as public
* 5.2:
bug #40427 [Console] Stop accepting ints as InputOption defaults
Fix fingerprint when context is not serializable
Fix `ConstraintViolation#getMessageTemplate()` to always return `string`
This PR was merged into the 4.4 branch.
Discussion
----------
[Console] Fix type of InputOption::$default
Options can also be `int`s. So add this type to the PHPDoc parameter
type annotation of `InputInterface::setOption` and the return type
annotation of `InputInterface::getOption`.
| Q | A
| ------------- | ---
| Branch? | 4.4 or 5.2 for bug fixes
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40427
| License | MIT
Commits
-------
3a2d1b4a0a bug #40427 [Console] Stop accepting ints as InputOption defaults
This PR was merged into the 4.4 branch.
Discussion
----------
[TwigBridge] Fix "Serialization of 'Closure'" error when rendering an TemplatedEmail
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40445
| License | MIT
| Doc PR |
When context contains a closure, it can't be serialized. In that case, we now assume that fingerprint is always different, and in that case, email will always be re-rendered
Commits
-------
c3e30ebda2 Fix fingerprint when context is not serializable
The types accepted and provided by `InputInterface::getOption`
and `setOption` do not include `int` (and passing something like
`--option=123` will set the option to the string `"123"`, not
to the integer `123`).
The `InputOption` default types should match this.
This PR was merged into the 5.2 branch.
Discussion
----------
Fix `ConstraintViolation#getMessageTemplate()` to always return `string`
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
`ConstraintViolation#getMessageTemplate()`'s inherited signature states that `string` is
to be returned by it at all times, yet the implementation returns `null` when no message
template had been provided at instantiation.
This patch obviates it, returning an empty string when the
message template is `null`.
Ref: https://github.com/symfony/symfony/pull/40415#issuecomment-792839512
Commits
-------
72a464e449 Fix `ConstraintViolation#getMessageTemplate()` to always return `string`
* 5.2:
Bump Symfony version to 5.2.6
Update VERSION for 5.2.5
Update CHANGELOG for 5.2.5
Update translations for Norwegian Nynorsk (nn) #38756
Fix eventListener initialization when eventSubscriber constructor dispatch an event
[FrameworkBundle] fix XSD
clear unchecked choice radio boxes even if clear missing is set to false
Fix `ConstraintViolation#getPropertyPath()` to always return `string`
[ErrorHandler] Added missing type annotations to FlattenException
[TwigBridge] Allow version 3 of the Twig extra packages
Fix FrameworkBundle PropertyAccess definition when not in debug
* 4.4:
Update translations for Norwegian Nynorsk (nn) #38756
Fix eventListener initialization when eventSubscriber constructor dispatch an event
clear unchecked choice radio boxes even if clear missing is set to false
[ErrorHandler] Added missing type annotations to FlattenException
[TwigBridge] Allow version 3 of the Twig extra packages
Fix FrameworkBundle PropertyAccess definition when not in debug
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Runtime] a new component to decouple applications from global state
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | https://github.com/symfony/symfony-docs/pull/15081
Follow up of #36652, see discussion there.
What if we could decouple the bootstrapping logic of our apps from any global state?
This PR makes it possible via a new proposed `symfony/runtime` component.
The immediate benefit this provides is easier maintenance of Symfony apps: code that is currently shipped by recipes will be able to move to `vendor/`. Read the previous sentence twice, this is big :)
Check the following PR to see how far this goes: https://github.com/symfony/recipes/pull/787
The longer-term benefit is being able to run the exact same app under several runtimes: PHP-FPM, CLI, but also PHP-PM and similar. Thanks to the proposed interface, this benefit could span to any PHP apps; not only to apps using the Symfony HttpKernel/HttpFoundation components. This part could be moved to `symfony/contracts` in the future.
Performance-wise, I measured no significant difference with the current way of running apps.
RuntimeInterface
----------------
The core of this component is the `RuntimeInterface` which describes a high-order
runtime logic.
It is designed to be totally generic and able to run any application outside of
the global state in 6 steps:
1. the main entry point returns a callable that wraps the application;
2. this callable is passed to `RuntimeInterface::getResolver()`, which returns a
`ResolverInterface`; this resolver returns an array with the (potentially
decorated) callable at index 0, and all its resolved arguments at index 1;
3. the callable is invoked with its arguments; it returns an object that
represents the application;
4. that object is passed to `RuntimeInterface::getRunner()`, which returns a
`RunnerInterface`: an instance that knows how to "run" the object;
5. that instance is `run()` and returns the exit status code as `int`;
6. the PHP engine is exited with this status code.
This process is extremely flexible as it allows implementations of
`RuntimeInterface` to hook into any critical steps.
Autoloading
-----------
This package registers itself as a Composer plugin to generate a
`vendor/autoload_runtime.php` file. This file shall be required instead of the
usual `vendor/autoload.php` in front-controllers that leverage this component
and return a callable.
Before requiring the `vendor/autoload_runtime.php` file, set the
`$_SERVER['APP_RUNTIME']` variable to a class that implements `RuntimeInterface`
and that should be used to run the returned callable.
Alternatively, the class of the runtime can be defined in the `extra.runtime.class`
entry of the `composer.json` file.
A `SymfonyRuntime` is used by default. It knows the conventions to run
Symfony and native PHP applications.
Examples
--------
This `public/index.php` is a "Hello World" that handles a "name" query parameter:
```php
<?php
require_once dirname(__DIR__).'/vendor/autoload_runtime.php';
return function (array $request, array $context): void {
// $request holds keys "query", "body", "files" and "session",
// which map to $_GET, $_POST, $_FILES and &$_SESSION respectively
// $context maps to $_SERVER
$name = $request['query']['name'] ?? 'World';
$time = $context['REQUEST_TIME'];
echo sprintf('Hello %s, the current Unix timestamp is %s.', $name, $time);
};
```
This `bin/console.php` is a single-command "Hello World" application
(run `composer require symfony/console` before launching it):
```php
<?php
use Symfony\Component\Console\Command\Command;
use Symfony\Component\Console\Input\InputInterface;
use Symfony\Component\Console\Output\OutputInterface;
require_once dirname(__DIR__).'/vendor/autoload_runtime.php';
return function (Command $command) {
$command->addArgument('name', null, 'Who should I greet?', 'World');
return $command->setCode(function (InputInterface $input, OutputInterface $output) {
$name = $input->getArgument('name');
$output->writeln(sprintf('Hello <comment>%s</>', $name));
});
};
```
The `SymfonyRuntime` can resolve and handle many types related to the
`symfony/http-foundation` and `symfony/console` components.
Check its source code for more information.
Commits
-------
61b32ab2a3 [Runtime] a new component to decouple applications from global state
This PR was merged into the 5.3-dev branch.
Discussion
----------
Don't use sprintf in trigger_deprecation() calls
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
Old habits die hard :)
Commits
-------
9ba8f0567d Don't use sprintf in trigger_deprecation() calls
`ConstraintViolation#getMessageTemplate()`'s inherited signature states that `string` is
to be returned by it at all times, yet the implementation returns `null` when no message
template had been provided at instantiation.
This patch obviates it, returning an empty string when the
message template is `null`.
Ref: https://github.com/symfony/symfony/pull/40415#issuecomment-792839512
This PR was merged into the 5.3-dev branch.
Discussion
----------
[HttpKernel] Deprecate returning a `ContainerBuilder` from `KernelInterface::registerContainerConfiguration()`
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | -
| License | MIT
| Doc PR | -
This behavior was introduced in 38aef98694 (diff-7753dd9020accbf6eb7c3dbb86a992340ce9dff4a0258458d7bbdd0a3d396939R261)
But this is never used nor tested AFAIK, and the interface doesn't document it.
Commits
-------
4449b553a8 [HttpKernel] Deprecate returning a `ContainerBuilder` from `KernelInterface::registerContainerConfiguration()`
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[DependencyInjection] Add support an integer return for default_index_method
| Q | A
| ------------- | ---
| Branch? | 5.x for features
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | #40319
| License | MIT
| Doc PR | TODO
Commits
-------
f0922c70d6 [DependencyInjection] Add support an integer return for default_index_method
This PR was merged into the 5.2 branch.
Discussion
----------
Fix `ConstraintViolation#getPropertyPath()` to always return `string`
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
`ConstraintViolation#getPropertyPath()`'s inherited signature states that `string` is
to be returned by it at all times, yet the implementation returns `null` when no property
path had been provided at instantiation.
This patch obviates it, returning an empty string when the
property path is `null`.
Commits
-------
7d1029b907 Fix `ConstraintViolation#getPropertyPath()` to always return `string`
This PR was merged into the 4.4 branch.
Discussion
----------
[DoctrineBridge] Fix eventListener initialization when eventSubscriber constructor dispatch an event
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40365
| License | MIT
| Doc PR | -
The issue occurred, when an EventSubscriber (lazyLoaded) dispatch an event when constructed. In that case, the state of the `ContainerAwareEventManager` become inconsistent for the triggered event:
- the `listener` property contains both listener instance and `serviceId` meaning it's not fully initialized
- the `initialized` property contains `true` meaning the listeners are initialized
Sorry for this PR without test, But it's really to hard to reproduce the issue :(
@parijke @michanismus @fliespl @reypm could you please check if this patch fixes the issue for you?
Commits
-------
b3ee29244f Fix eventListener initialization when eventSubscriber constructor dispatch an event
This PR was merged into the 4.4 branch.
Discussion
----------
[FrameworkBundle] Fix PropertyAccess definition when not in debug
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| License | MIT
The signature of the `PropertyAccessor::createCache()` method specifies that the first argument must be a string but in the `FameworkBundle` configures its DIC with a first argument to `null`. Replacing `null` by `''` allows to respect the contact of `PropertyAccessor::createCache()` and removes the following error when the `symfony/framework-bundle` is blocked in `4.4` but the symfony/property-access goes up to `5.2`.
```
Argument 3 passed to Symfony\Component\PropertyAccess\PropertyAccessor::createCache() must be of the type string, null given, called in /usr/src/app/var/cache/prod/ContainerDX7KWI4/getCache_PropertyAccessService.php on line 12
```
Commits
-------
116c54a554 Fix FrameworkBundle PropertyAccess definition when not in debug
This PR was merged into the 4.4 branch.
Discussion
----------
[Form] clear unchecked choice radio boxes even if clear missing is set to false
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#16802
| License | MIT
| Doc PR |
Commits
-------
e7b4851ea0 clear unchecked choice radio boxes even if clear missing is set to false
This PR was merged into the 5.3-dev branch.
Discussion
----------
[PropertyAccess] use bitwise flags to configure when the property accessor should throw
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | Fix#31126
| License | MIT
| Doc PR |
Commits
-------
a50cfcb49d use bitwise flags to configure when the property accessor should throw
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Cache] boost perf by wrapping keys validity checks with `assert()`
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
PSR-6 has one perf hog: checking the validity of keys.
But in practice, an invalid key should never happen in production: encoding/cleanup is a must-have, and it's a step that should be identified *during dev*.
That's why I think we're safe wrapping these checks with `assert()`.
On an `ArrayAdapter`, this doubles the throughput of the pool when getting items.
I didn't use `assert()` in constructors when not on the hot path.
This PR also makes some callable properties static, as they should be from the beginning.
Commits
-------
8f03a1f555 [Cache] boost perf by wrapping keys validity checks with `assert()`
This PR was merged into the 4.4 branch.
Discussion
----------
[ErrorHandler] Added missing type annotations to FlattenException
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40385
| License | MIT
| Doc PR | N/A
This class has no constructor where the properties' types could be inferred from. If we want potential contributors (and our serializer, see #40385) to read that code, I think it's only fair that we document the property types via doc blocks.
Commits
-------
d68832e1b9 [ErrorHandler] Added missing type annotations to FlattenException
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Console] Avoid unneeded preg_replace_callback
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| License | MIT
Motivations for this change:
* Avoid an unneded preg call, explode+implode is faster
* The previous regex created to suboptimal expressions,
due to the pipe that caused empty to be matched.
That means an input like `foo:bar`
was translated into `foo[^:]*[^:]*:bar[^:]*[^:]*`
instead of simply `foo[^:]*:bar[^:]*`
Commits
-------
daaa760fd2 Avoid unneeded preg_replace_callback in console application
Motivations for this change:
* Avoid an unneded preg call, explode+implode is faster
* The previous regex created to suboptimal expressions,
due to the pipe that caused empty to be matched.
That means an input like `foo:bar`
was translated into `foo[^:]*[^:]*:bar[^:]*[^:]*`
instead of simply `foo[^:]*:bar[^:]*`
This PR was merged into the 4.4 branch.
Discussion
----------
[TwigBridge] Allow version 3 of the Twig extra packages
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | N/A
| License | MIT
| Doc PR | N/A
Commits
-------
61b0be9564 [TwigBridge] Allow version 3 of the Twig extra packages
`ConstraintViolation#getPropertyPath()`'s inherited signature states that `string` is
to be returned by it at all times, yet the implementation returns `null` when no property
path had been provided at instantiation.
This patch obviates it, returning an empty string when the
property path is `null`.
