* 5.2:
[TwigBridge] Install symfony/intl to run tests on Travis
[Translation] Make `name` attribute optional in xliff2
[Security] #[CurrentUser] argument should resolve to null when it is anonymous
This PR was merged into the 4.4 branch.
Discussion
----------
[TwigBridge] Install symfony/intl to run tests on Travis
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | bi
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
The hard dependency on `symfony/intl` was removed from the Form component in 5.3-dev (#40298). I suggest to add the explicit dev dependency on TwigBridge on 4.4 already.
Commits
-------
b2970456bf [TwigBridge] Install symfony/intl to run tests on Travis
This PR was merged into the 5.3-dev branch.
Discussion
----------
[Routing] Construct Route annotations using named arguments
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | N/A
| License | MIT
| Doc PR | Not needed
This PR proposes to bump the `doctrine/annotations` library to 1.12 to gain access to its emulation layer for named arguments. Furthermore, constructing a `Route` annotation the old way by passing an array of parameters is deprecated.
### Reasons for this change
The constructors of our annotation classes have become unnecessarily complicated because we have to support two ways of calling them:
* An array of parameters, passed as first argument, because that's the default behavior `doctrine/annotations`.
* A set of named arguments because that's how PHP 8 attributes work.
Since we can now tell the Doctrine annotation reader to use named arguments as well, we can simplify the constructors of our annotations significantly.
### Drawback
After this change, there is no easy way anymore to construct instances of the `Route` annotation class directly on PHP 7. The PR has been built under the assumption that instances of this class are usually created using either Doctrine annotations or a PHP 8 attribute. Thus, most applications should be unaffected by this change.
Commits
-------
29b0f96046 [Routing] Construct Route annotations using named arguments
This PR was merged into the 5.3-dev branch.
Discussion
----------
Deprecate passing null as $message or $code to exceptions
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | N/A
| License | MIT
| Doc PR | N/A
Follow-up to #40271.
Following the example of the PHP core, this PR introduces deprecation warnings that are triggered if a developer attempts to pass null as `$code` or `$message` to an exception constructor.
Commits
-------
8e3058d95a Deprecate passing null as $message or $code to exceptions
This PR was squashed before being merged into the 5.3-dev branch.
Discussion
----------
[Form] Remove hard dependency on symfony/intl
| Q | A
| ------------- | ---
| Branch? | 5.x (or 6.0)
| Bug fix? |
| New feature? | no
| Deprecations? | yes
| Tickets | Fix#39596
| License | MIT
| Doc PR | https://github.com/symfony/symfony-docs/pull/15026
This was voted down in 2018 (#29229) and will revert #29720 by @chalasr. I reopen it because the Form component is way less dependent on Intl component now.
Im hesitant if we should do this in 5.x or 6.0. If a user don't have `symfony/intl` installed, they will get an error in runtime. That is something that speaks for doing it in 6.0.
Could I get some opinions?
### TODO
- [x] Update `UPGRADE-x.x.md`
Commits
-------
f90d3ec203 [Form] Remove hard dependency on symfony/intl
This PR was submitted for the 5.2 branch but it was merged into the 4.4 branch instead.
Discussion
----------
[Translation] Make `name` attribute optional in xliff2
Do not set a fake `name` attribute on `unit` element from xliff2 to allow using `source` attribute and avoid missing translation error
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes/no
| New feature? | no
| Deprecations? | no
| Tickets | Fix#37055
| License | MIT
| Doc PR | symfony/symfony-docs#... <!-- required for new features -->
When `xlf` translations are loaded, if a name exists on `unit` element, the segment's source is ignored:
```foreach ($xml->xpath('//xliff:unit') as $unit) {
foreach ($unit->segment as $segment) {
$attributes = $unit->attributes();
$source = $attributes['name'] ?? $segment->source;
```
At the same time, when dumping translations, the segment's source is copied into the unit's name attribute, unless it's longer than 80 characters. In that case, `substr(md5($source), -7)` is set into the name attribute.
This results in a missing translation error, because the source is ignored and the name is a random string.
Suggested solution: only set the name attribute if the string is less than 80 characters.
Commits
-------
97058559cc [Translation] Make `name` attribute optional in xliff2
This PR was merged into the 5.2 branch.
Discussion
----------
[Security] #[CurrentUser] arguments should resolve to null for "anon."
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
The UserValueResolver should only resolve `UserInterface` (or subtype) typed arguments:
bc9e946a56/src/Symfony/Component/Security/Http/Controller/UserValueResolver.php (L54-L55)
When using the `#CurrentUser` attribute with an AnonymousToken in the storage, the resolved argument value is `anon.`. This PR fixes it.
/cc @jvasseur
Commits
-------
8d3078dd35 [Security] #[CurrentUser] argument should resolve to null when it is anonymous
This PR was merged into the 5.2 branch.
Discussion
----------
[Config] Switched to non-null defaults in exception constructors
| Q | A
| ------------- | ---
| Branch? | 5.2
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | N/A
| License | MIT
| Doc PR | N/A
Follow-up to #40271 on the 5.2 branch.
Commits
-------
2e865ac057 Switched to non-null defaults in exception constructors
This PR was merged into the 5.3-dev branch.
Discussion
----------
Remove service session.storage.mock_file when user configure factory
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40282
| License | MIT
| Doc PR | -
When deprecating the session.storage services, and user configures a factory, the FrameworkExtension removes statefull services, but I forgot to also remove the `session.storage.mock_file` service.
Commits
-------
483f840a08 Remove service session.storage.mock_file when user configure factory
This PR was submitted for the 5.x branch but it was merged into the 5.2 branch instead.
Discussion
----------
[FrameworkBundle] Allow x-forwarded-prefix trusted header in config
| Q | A
| ------------- | ---
| Branch? | 5.2 (as requested by @nicolas-grekas)
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| License | MIT
Support for `X_FORWARDED_PREFIX` has been added in PR https://github.com/symfony/symfony/pull/37734.
However, it is impossible to use it because the configuration doesn't allow the `x-forwarded-prefix` value in `framework.yaml`.
Commits
-------
95fdd90491 Allow x-forwarded-prefix trusted header.
* 5.2:
Move github templates at the org level
[Cache] Fix Redis TLS scheme `rediss` for Redis connection
In calls to mb_ functions, silently transform arg into string
Switched to non-null defaults in exception constructors
[Routing] fix conflict with param named class in attribute
[Cache] fix setting items' metadata on commit()
* 4.4:
Move github templates at the org level
[Cache] Fix Redis TLS scheme `rediss` for Redis connection
In calls to mb_ functions, silently transform arg into string
This PR was merged into the 4.4 branch.
Discussion
----------
[Console] Handle calls to mb_ functions with non string arguments
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40200
| License | MIT
| Doc PR | no
In PHP8.1, a number of functions who were accepting null arguments will only accept
string ones.
(see https://wiki.php.net/rfc/deprecate_null_to_scalar_internal_arg)
In the polyfill, mb_* functions are already declared with a strict type checking of "string".
Therefore, it is necessary to get rid of the use of non string arguments when calling mb_* functions,
so that it won't break when either using the polyfill,or future php8 versions.
In every call where the argument may not be a string, this commit enforces the string type of the argument (with transtyping)
--- For reviewers
* I generally don't like transtyping, but found it was the more "secure" way (on a non-BC point of view) here.
Specially in Console/Helper/Table.php, where $cell can be an object (there are 2 "$cell instanceof ... tests)
However, where the argument can already be either null or string (and not anything else), there may a beter approach ?
* It's the first time I send a PR on symfony, so don't hesitate pointing me to thinks I've forgotten to done.
Commits
-------
ac45be2580 In calls to mb_ functions, silently transform arg into string
This PR was submitted for the 5.x branch but it was merged into the 4.4 branch instead.
Discussion
----------
[Cache] Fix Redis TLS scheme `rediss` for Redis connection
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR | https://github.com/symfony/symfony-docs/pull/14728
Like https://github.com/symfony/symfony/pull/35503 on Symfony Messenger, this will enable TLS support for Redis adapter.
The implementation just prefix the host with `tls://` as described here: https://github.com/phpredis/phpredis#connect-open
I don't know how to test it because I guess I need a TLS Redis in `src/Symfony/Component/Cache/Tests/Adapter/RedisAdapterTest.php`.
Commits
-------
3288897e0f [Cache] Fix Redis TLS scheme `rediss` for Redis connection
In PHP8, a number of functions who were accepting null arguments will only accept
string ones.
In the polyfill, mb_* functions are declared with a trict type checking of "string".
Therefore, we deprecate the use of non string arguments, so that it won't break when either using the polyfill,
or future php8 versions.
* 4.4:
Switched to non-null defaults in exception constructors
[Routing] fix conflict with param named class in attribute
[Cache] fix setting items' metadata on commit()
This PR was merged into the 4.4 branch.
Discussion
----------
Switched to non-null defaults in exception constructors
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | N/A
| License | MIT
| Doc PR | N/A
PHP 8.1 will trigger a deprecation warning if we pass `null` as `$message` or `$code` to the constructor of `\Exception`. However, many of our own exception accept `null` for those parameters and even use them as default.
This is unfortunate because code like the following snippet would trigger that deprecation although the code itself is perfectly fine:
```php
throw new NotFoundHttpException();
```
With this PR, I'd like to change our defaults to `''` and `0` while still allowing to pass `null` for BC. In a follow-up PR for the 5.x branch, I'd like to deprecate passing `null`, matching the future behavior of PHP.
This PR also adjust various PHPDoc blocks with inaccurate types.
Commits
-------
f8e10094a4 Switched to non-null defaults in exception constructors
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[Routing] fix conflict with param named class in attribute
| Q | A
| ------------- | ---
| Branch | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40225
| License | MIT
| Doc PR | -
Fix conflict with AnnotationFileLoader and class PHP8 Attribute with param named "class"
Commits
-------
27bba684d8 [Routing] fix conflict with param named class in attribute
This PR was merged into the 5.3-dev branch.
Discussion
----------
[FrameworkBundle] allow container/routing configurators to vary by env
| Q | A
| ------------- | ---
| Branch? | 5.x
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | #40215
| License | MIT
| Doc PR | -
Inspired by https://github.com/symfony/webpack-encore/pull/900 and by a chat on Slack with @weaverryan
This aims at allowing conditional configuration, which would allow merging config files in one.
Using the PHP-DSL:
```php
$container
->when(env: 'prod')
->services()
->set(Foo::class)
//...
```
In Yaml:
```yaml
framework:
secret: '%env(APP_SECRET)%'
when@dev:
services:
App\FooForDev: ~
when@test:
framework:
test: true
session:
storage_factory_id: session.storage.mock_file
```
In XML (omitting namespaces):
```xml
<when env="test">
<framework test="true">
<!-- ... -->
</framework>
</when>
```
A similar syntax is also provided for routes, with support for annotations:
`@Route(env="prod")` defines a route that is enabled only on the "prod" env.
Commits
-------
108375b068 [FrameworkBundle] allow container/routing configurators to vary by env
* 5.2:
[Form] Ignoring invalid forms from delete_empty behavior in CollectionType
Add Symfony Armenian Translations
cs fix
Add different header notations to tests
* 4.4:
[Form] Ignoring invalid forms from delete_empty behavior in CollectionType
Add Symfony Armenian Translations
cs fix
Add different header notations to tests
This PR was squashed before being merged into the 4.4 branch.
Discussion
----------
[Form] Ignoring invalid forms from delete_empty behavior in CollectionType
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#37660
| License | MIT
| Doc PR | -
As expained on linked issue, empty forms cannot be removed if they are invalid upon submitting (e.g. transformation failures). Thus, form errors can be displayed properly to the end user.
Commits
-------
e4911554e3 [Form] Ignoring invalid forms from delete_empty behavior in CollectionType
This PR was merged into the 4.4 branch.
Discussion
----------
Add Symfony Armenian Translations
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | Fix#40204
| License | MIT
| Doc PR |
Added missing armenian translations for Form components
Commits
-------
d1c3e21eff Add Symfony Armenian Translations
This PR was merged into the 4.4 branch.
Discussion
----------
[BrowserKit] Add different header notations to tests
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | Fix#29736
| License | MIT
| Doc PR | -
In the past there was a issue #29736 that the header wasn't processed correctly. To reproduce and prevent this behaviour i extended the tests for this. At the end everything is fine, also in 5.x branch. See bf03d6889d.
Commits
-------
251ce8f924 Add different header notations to tests