* 5.0:
[travis] fix CI (ter)
Revert "[travis][appveyor] don't cache .phpunit"
silence E_NOTICE triggered since PHP 7.4
[Form] Removed legacy check in `ValidationListener`
[HttpClient] fix HTTP/2 support on non-SSL connections - CurlHttpClient only
Force ping after transport Exception
do not merge constraints within interfaces
[Validator] Fixed default group for nested composite constraints
* 4.4:
[travis] fix CI (ter)
Revert "[travis][appveyor] don't cache .phpunit"
silence E_NOTICE triggered since PHP 7.4
[Form] Removed legacy check in `ValidationListener`
[HttpClient] fix HTTP/2 support on non-SSL connections - CurlHttpClient only
Force ping after transport Exception
do not merge constraints within interfaces
[Validator] Fixed default group for nested composite constraints
* 3.4:
Revert "[travis][appveyor] don't cache .phpunit"
silence E_NOTICE triggered since PHP 7.4
[Form] Removed legacy check in `ValidationListener`
do not merge constraints within interfaces
[Validator] Fixed default group for nested composite constraints
This PR was merged into the 3.4 branch.
Discussion
----------
[Validator] Fixed default group for nested composite constraints
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#33986
| License | MIT
| Doc PR | ~
Take a breath: when composite constraints are nested in a parent composite constraint without having non composite nested constraints (i.e empty), then the default group is not added, making the validator failing to validate in any group (including default), because there is no group at all, which should never happen.
Commits
-------
117ee34698 [Validator] Fixed default group for nested composite constraints
This PR was merged into the 4.4 branch.
Discussion
----------
[HttpClient] fix HTTP/2 support on non-SSL connections - CurlHttpClient only
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#36419
| License | MIT
| Doc PR | -
Commits
-------
a5b884cd94 [HttpClient] fix HTTP/2 support on non-SSL connections - CurlHttpClient only
This PR was merged into the 4.4 branch.
Discussion
----------
Force ping after transport exception
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#36301
| License | MIT
| Doc PR |
SMTP transport fails for long running processes after tranport exception, if stream is closed all messages will throw a transport exception until $lastMessageTime exceds $pingThreshold.
With this PR, after transport expception the transport will ping the server to check if the connection is still alive.
Commits
-------
7ccbef62f6 Force ping after transport Exception
This PR was merged into the 3.4 branch.
Discussion
----------
[Validator] do not merge constraints within interfaces
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | Fix#22538
| License | MIT
| Doc PR |
This fix disables merge constraints within interfaces.
There is no reason to merge constraints from one interface to another because each class merges the constraints of all its interfaces. Only one check is needed is to eliminate all interfaces that comes from parent class to avoid duplication.
Commits
-------
67f336b808 do not merge constraints within interfaces
This PR was merged into the 3.4 branch.
Discussion
----------
[Form] Removed legacy check in `ValidationListener`
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | ~
| License | MIT
| Doc PR | ~
A left over of #13198, should have been removed in 3.0. The tests don't use `null` anymore, no update needed here, this is just about removing dead code.
Commits
-------
e479e51f7c [Form] Removed legacy check in `ValidationListener`
This PR was merged into the 5.1-dev branch.
Discussion
----------
[Serializer] UnwrappingDenormalizer
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | no
| New feature? | yes
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
UnwrappingDenormalizer, registered with very high priority. Unwrapping the data if UNWRAP_PATH is provided.
Very often some APIs give nested responses in which we need only the child object. With UnwrappingDenormalizer we can get the needed object without creating unnecessary Model class that we don't really need.
Regarding to https://github.com/symfony/symfony/pull/28887 and https://github.com/symfony/symfony/pull/30894
Usage:
`$serialiser->deserialize('{"baz": {"foo": "bar", "inner": {"title": "value", "numbers": [5,3]}}}', Object::class, ['UnwrappingDenormalizer::UNWRAP_PATH' => '[baz][inner]'])`
Commits
-------
00d103d5f7 UnwrappingDenormalizer
This PR was merged into the 5.1-dev branch.
Discussion
----------
[ErrorHandler] Remove trigger_deprecation frame from trace
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
This accounts for the new `trigger_deprecation` function from [symfony/deprecation-contracts](https://github.com/symfony/deprecation-contracts).
Replaces #36329
Commits
-------
d4eb4a4bd7 [ErrorHandler] Remove trigger_deprecation frame from trace (add tests)
c293aee9ab [ErrorHandler] Remove trigger_deprecation frame from trace
This PR was merged into the 5.1-dev branch.
Discussion
----------
[DI] remove restriction and allow mixing "parent" and instanceof-conditionals/defaults/bindings
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
We added these limitations because we didn't know how to merge all those dimensions. Yet, the code is perfectly able to handle all of them together.
This PR makes `parent` definitions first-class citizens again - no need to split them in a dedicated file anymore.
Commits
-------
622925300f [DI] remove restriction and allow mixing "parent" and instanceof-conditionals/defaults/bindings
This PR was merged into the 5.1-dev branch.
Discussion
----------
[DI] deprecate the `inline()` function from the PHP-DSL in favor of `service()`
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | no
| New feature? | no
| Deprecations? | yes
| Tickets | -
| License | MIT
| Doc PR | -
In the PHP-DSL, the `inline()` function helps declaring ... inline services.
I'm proposing to rename it to `service()`, for consistency with yaml tags. All other helpers but this one have the same name as the yaml tag.
Let's do this while "nobody" uses this format yet :)
Commits
-------
647d971ae4 [DI] deprecate the `inline()` function from the PHP-DSL in favor of `service()`
This PR was merged into the 5.1-dev branch.
Discussion
----------
[DI] allow decorators to reference their decorated service using the special `.inner` id
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | no
| New feature? | yes
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
Right now, when one wants to decorate a service, one needs to reference the decorated service using `foo.inner`, where `foo` is the id of the decorating service.
There are two issues with this IMHO:
- it's weird to have to repeat the name of declaring service inside of it;
- it doesn't play well with child definitions.
In this PR, I propose to use `.inner` to fix both issues.
Before:
```yaml
services:
decorating_service:
arguments: [@decorating_service.inner]
decorates: decorated_service
```
After:
```yaml
services:
decorating_service:
arguments: [@.inner]
decorates: decorated_service
```
Commits
-------
7b6f767fbe [DI] allow decorators to reference their decorated service using the special `.inner` id
This PR was merged into the 4.4 branch.
Discussion
----------
[travis] fix CI
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | no
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
Commits
-------
f6410cbab2 [travis] fix CI
* 5.0:
[appveyor] bump cache
[Twig][Mime] Removed extra quotes in missing package exception message
[DI] µfix
Allowing empty secrets to be set
[DI] add missing property declarations in InlineServiceConfigurator
[DI] fix detecting short service syntax in yaml
Supress error from fread when reading a unix pipe
[HttpClient] Fix scoped client without query option configuration
[Workflow] Use a strict comparison when retrieving raw marking in MarkingStore
[Workflow] Use a strict comparison when retrieving raw markin in MarkingStore
* 4.4:
[appveyor] bump cache
[Twig][Mime] Removed extra quotes in missing package exception message
[DI] µfix
Allowing empty secrets to be set
[DI] add missing property declarations in InlineServiceConfigurator
[DI] fix detecting short service syntax in yaml
Supress error from fread when reading a unix pipe
[HttpClient] Fix scoped client without query option configuration
[Workflow] Use a strict comparison when retrieving raw marking in MarkingStore
[Workflow] Use a strict comparison when retrieving raw markin in MarkingStore
This PR was merged into the 4.4 branch.
Discussion
----------
[HttpClient] Fix scoped client without query option configuration
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets |
| License | MIT
| Doc PR |
The `query` key default value is an [empty array](https://github.com/symfony/symfony/blob/v4.4.7/src/Symfony/Component/Config/Definition/PrototypedArrayNode.php#L30) and because of that it is always set. Processing a configuration for a scoped HTTP client (which has a `scope` and does not have a `base_uri`) results in the configuration being invalid. The error message says that query parameters cannot be aplied to the base URI since it is not defined (which doesn't make sense since the query parameters don't exist because they are empty).
Commits
-------
a07578dba3 [HttpClient] Fix scoped client without query option configuration
This PR was merged into the 4.4 branch.
Discussion
----------
[DI] fix detecting short service syntax in yaml
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
We do allow interfaces/classes as keys in arguments, yet the short syntax fails to know about those.
This fixes it, allowing one to use:
```
services:
App\Foo:
App\BarInterface: '@App\BarClass'
```
As a reminder, by-name is also allowed:
```
services:
App\Foo: {
$bar: '@App\BarClass'
}
```
Commits
-------
bf17165fb1 [DI] fix detecting short service syntax in yaml
This PR was merged into the 4.4 branch.
Discussion
----------
[DI] add missing property declarations in InlineServiceConfigurator
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | -
| License | MIT
| Doc PR | -
These are accessed by the traits used by the class.
Commits
-------
a6a4442cd9 [DI] add missing property declarations in InlineServiceConfigurator
This PR was merged into the 4.4 branch.
Discussion
----------
Allowing empty secrets to be set
| Q | A
| ------------- | ---
| Branch? | 4.4
| Bug fix? | yes
| New feature? | no
| Deprecations? | no
| Tickets | None
| License | MIT
| Doc PR | Not needed
Setting secrets to an empty string is a valid value - you can already do this by consuming from an empty file or `stdin` with no input. But it's *not* currently possible to use the interactive prompt to set a secret to an empty string. This fixes that.
Commits
-------
c9bf0c8683 Allowing empty secrets to be set