This PR was merged into the 2.2 branch.
Discussion
----------
[Process][2.2] Fix#8970 : read output once the process is finished, enable pipe tests on Windows
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8970
| License | MIT
This fix the read of file handles after `proc_close` as described in the issue. I enable some stdin=>stdout pipes tests on windows with some dedicated buffer size. Solving the issue, I finally reproduced this [PHP bug](https://bugs.php.net/bug.php?id=65650) that I first described in my PR note in #8924
Most of Windows usage should be okay, but in case of a program throws lots of output and fills the buffer, some data might be lost/corrupted. Should it be documented ? This is a Windows only known issue.
Commits
-------
1e75cf9 [Process] Fix#8970 : read output once the process is finished, enable pipe tests on Windows
This PR was squashed before being merged into the 2.2 branch (closes#8924).
Discussion
----------
[Process][2.2] Fix Process component on windows
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8836, #8799, #7078
| License | MIT
This PR fixes Process on windows (almost, see note below).
- Some unit tests were not Windows compatible
- Use a file handle for STDERR as well as STDOUT to avoid blocking
- Decouple pipes and descriptors from Process
As this move some a part of Process in a sub class, I hope merging this in 2.3 and master would not be a PITA. I'm here to make some adjustments after theses merge if needed.
**Important note** :
We are using file handles instead of streams for `proc_open` pipes as described in the code (see [PHP bug #51800](https://bugs.php.net/bug.php?id=51800)). Unfortunately, this workaround may produce corrupted output/error output in some race conditions. That's why `AbstractProcessTest::testProcessPipes` randomly fails when using file handles (on unix and windows).
Commits
-------
4a76c76 [Process][2.2] Fix Process component on windows
This PR was merged into the 2.2 branch.
Discussion
----------
[Security] fixed a leak in the ContextListener
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Trying to fix leaks when using the same Kernel to handle several requests in a row without resetting the Container or shutting down the Kernel.
Commits
-------
899f176 [Security] fixed a leak in ExceptionListener
2fd8a7a [Security] fixed a leak in the ContextListener
This PR was merged into the 2.2 branch.
Discussion
----------
Fixed some random leaks
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Commits
-------
970405f fixed some circular references
This PR was merged into the 2.2 branch.
Discussion
----------
[2.2] Return BC compatibility for `@Route` parameters and default values
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #7724
| License | MIT
Commits
-------
d400b5a Return BC compatibility for `@Route` parameters and default values
This PR was merged into the 2.2 branch.
Discussion
----------
[HttpKernel] fixed route parameters storage in the Request data collector (closes#8867)
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8867
| License | MIT
| Doc PR | n/a
Commits
-------
c931eb7 [HttpKernel] fixed route parameters storage in the Request data collector (closes#8867)
This PR was merged into the 2.2 branch.
Discussion
----------
[2.2] [BrowserKit] Pass headers when `followRedirect()` is called
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Tests pass? | yes
| Fixed tickets | #7929
| License | MIT
Commits
-------
0d07af8 [BrowserKit] Pass headers when `followRedirect()` is called
This PR was merged into the 2.2 branch.
Discussion
----------
Made fragment URLs relative instead of absolute by default
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8458
| License | MIT
| Doc PR | n/a
This fixes a regression introduced with the new fragment system where fragment URLs were generated as absolute URLs. As per the ESI spec, there is no need to generate an absolute URL and this can even be problematic when dealing with internal sub-requests in a more "complex" hosting environment.
Commits
-------
91234cd [HttpKernel] changed fragment URLs to be relative by default (closes#8458)
This PR was merged into the 2.2 branch.
Discussion
----------
[2.2][Form] Dereferenced Form::&all() since it is not supported for PHP <= 5.3.3
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
see 00bc2708bc
Commits
-------
ea480bd [Form] Fixed Form::all() signature for PHP 5.3.3
This PR was merged into the 2.2 branch.
Discussion
----------
[Translation] Fixed regression: When only one rule is passed to transChoice(), this rule should be used
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
There was one regression introduced by #8789, namely when exactly one message is passed to `transChoice()`. Before, the following code was perfectly valid:
```php
$translator = new IdentityTranslator(new MessageSelector());
echo $translator->transChoice('There are two apples', 2, array('%count%' => 2));
```
This didn't fail, because internally, the locale `null` was passed to `MessageSelector`, which always returned the index `0` in that case.
Now, the user's locale is passed, and if the selector returns anything but `0`, an exception is thrown.
I removed this exception for the case that exactly one standard message (without explicit interval) is passed. In that case, this message is returned (just like from `trans()`). If the message key contains more than one message, or if the message has explicit intervals, the exception will still be thrown. See the tests for clarification.
A use case for why no exception should be thrown is validation:
```php
/**
* @Assert\Length(min=2, minMessage="Please enter at least two characters")
*/
```
Commits
-------
2b3dcb9 [Translation] Grammar fix
0951b8d [Translation] Fixed regression: When only one rule is passed to transChoice(), this rule should be used
This PR was merged into the 2.2 branch.
Discussion
----------
fixed request format when forwarding a request
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #6932
| License | MIT
| Doc PR | n/a
When calling forward() in a controller, the format is not the same as for the master request.
Commits
-------
7e87eb1 fixed request format when forwarding a request
This PR was merged into the 2.2 branch.
Discussion
----------
fixed locale management in sub-requests
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #6932, #5532, #8604
| License | MIT
| Doc PR | n/a
Same fix as #8821 but for `_locale`, and a rebase of #8604.
Commits
-------
7117328 [HttpKernel] added a comment to warn about possible inconsistencies
c4636e1 added a functional test for locale handling in sub-requests
05fdb12 Fixed issue #6932 - Inconsistent locale handling in subrequests
b3c3159 fixed locale of sub-requests when explicitely set by the developer (refs #8821)
This PR was merged into the 2.2 branch.
Discussion
----------
[Form][2.2] Fixed Form::submit() to react to dynamic form modifications
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
ref #3767, #3768, #4548, #8748
cc @Burgov
This PR ensures that fields that are added during the submission process of a form are submitted as well. For example:
```php
$builder = $this->createFormBuilder()
->add('country', 'choice', ...)
->getForm();
$builder->get('country')->addEventListener(FormEvents::POST_SUBMIT, function (FormEvent $event) {
$form = $event->getForm()->getParent();
$country = $event->getForm()->getData();
$form->add('province', 'choice', /* ... something with $country ... */);
});
```
Currently, the field "province" will not be submitted, because the submission loop uses `foreach`, which ignores changes in the underyling array.
Additionally, this PR ensures that `setData()` is called on *every* element of a form (except those inheriting their parent data) when `setData()` is called on the root element (i.e., during initialization). Currently, when some of the intermediate nodes (e.g. embedded forms) are submitted with an empty value, `setData()` won't be called on the nodes below (i.e. the fields of the embedded form) until `get*Data()` is called on them. If `getData()` is *not* called before `createView()`, any effects of `*_DATA` event listeners attached to those fields will not be visible in the view.
Commits
-------
cd27e1f [Form] Extracted ReferencingArrayIterator out of VirtualFormAwareIterator
ccaaedf [Form] PropertyPathMapper::mapDataToForms() *always* calls setData() on every child to ensure that all *_DATA events were fired when the initialization phase is over (except for virtual forms)
19b483f [Form] Removed superfluous reset() call
00bc270 [Form] Fixed: submit() reacts to dynamic modifications of the form children
This PR was merged into the 2.2 branch.
Discussion
----------
Sets _format attribute only if it wasn't set previously by the user
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8787, #8791
| License | MIT
| Doc PR |
Commits
-------
9bb7a3d fixed request format of sub-requests when explicitely set by the developer (closes#8787)
fa35597 Sets _format attribute only if it wasn't set previously by the user.
This PR was merged into the 2.2 branch.
Discussion
----------
[HttpKernel] Fix request _format duplication for HttpKernel's ExceptionListener
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8777, #8778
| License | MIT
This is a better fix for #8777.
Commits
-------
f946108 fixed the format of the request used to render an exception
This PR was merged into the 2.2 branch.
Discussion
----------
[Process] fix for Process:isSuccessful()
| Q | A
| ------------- | ---
| Bug fix? | [yes]
| New feature? | [no]
| BC breaks? | [no]
| Deprecations? | [no]
| Tests pass? | [yes]
| Fixed tickets | []
| License | MIT
| Doc PR | []
This is a rebase of https://github.com/symfony/symfony/pull/8801
If you call isSuccessful() on a running process you would get true because of 0 == null. I think that is not the expected behavior and that is not what phpdoc @return block said so.
Commits
-------
262879d fix for Process:isSuccessful()
This PR was merged into the 2.2 branch.
Discussion
----------
[2.2][Process] Use a consistent way to reset data of the process latest run
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets |
| License | MIT
It is actually useful when cloning or running again a process.
Commits
-------
0723c10 [Process] Use a consistent way to reset data of the process latest run
Per https://bugs.php.net/bug.php?id=61470, and in fixing #7380, the following error occurs when using a Memcache or Redis session store w/ Symfony security: "Authentication exception occurred; redirecting to authentication entry point (A Token was not found in the SecurityContext.)". This patch applies the first fix only if the session store is "files"
{HttpFoundation] [Session] fixed session compatibility with memcached/redis session storage
Per https://bugs.php.net/bug.php?id=61470, and in fixing #7380, the following error occurs when using a Memcache or Redis session store w/ Symfony security: "Authentication exception occurred; redirecting to authentication entry point (A Token was not found in the SecurityContext.)". This patch applies the first fix only if the session store is "files"
This PR was merged into the 2.2 branch.
Discussion
----------
[Process][2.2] Fix#8739
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8739
| License | MIT
This adds a fix to #8739. Whenever a call is done to to any non-blocking methods (`Process::isRunning`, `Process::isStopped`, `Process::isTerminated`, `Process::getStatus`, `Process::getPid`...), buffers are read, and callback executed.
Such code will now work :
```
$process->start(function ($type, $data) {
echo $data;
});
while ($process->isRunning()) {
// some stuff
// callback is executed
}
```
Commits
-------
fa769a2 [Process] Add more precision to Process::stop timeout
57d4159 [Process] Avoid zombie process in case of unit tests failure
3ef517b [Process] Fix#87397716fb2 [Process] Add failing test for #8739bff6f3c [Process] Fix CS
This PR was merged into the 2.2 branch.
Discussion
----------
[Form] Removed the "disabled" attribute from the placeholder option in select fields due to problems with the BlackBerry 10 browser
| Q | A
| ------------- | ---
| Bug fix? | yes (together with #8623)
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #7678#8478#8526
| License | MIT
| Doc PR | -
Commits
-------
97cbb19 [Form] Removed the "disabled" attribute from the placeholder option in select fields due to problems with the BlackBerry 10 browser
This PR was merged into the 2.2 branch.
Discussion
----------
[Routing] add ability for apache matcher to handle array values
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8567
Commits
-------
c138304 [routing] added ability for apache matcher to handle array values
This PR was merged into the 2.2 branch.
Discussion
----------
removed dead code and fixed CS
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Commits
-------
7d58147 removed dead code and fixed CS
This PR was squashed before being merged into the 2.2 branch (closes#8421).
Discussion
----------
[Validator] fixed ConstraintViolation::$propertyPath incorrect when nested
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8351
| License | MIT
In nested constraints, the property path will be overwritten in the context (as there would be a recursive call to CollectionValidotor when nested). Reason is, in ConstraintValidatorFactory object is loaded from memory if exists and context is initialized with the new context. So, other constraints after the nested constraints PropertyPath would be wrong.
So I think better create a new object for CollectionValidator always.
see this https://gist.github.com/alexkappa/5851274
It shows [name][email] even though the email is not under the name node.
Commits
-------
28e0709 [Validator] fixed ConstraintViolation:: incorrect when nested
This PR was merged into the 2.2 branch.
Discussion
----------
[Validator] Fixed groups argument misplace for validateValue method from validator class
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
Signature of validateValue method in ExecutionContext Class is this.
```
public function validateValue($value, $constraints, $subPath = '', $groups = null)
```
But this was called wrongly in Validator Class.
Commits
-------
d3eb9b7 [Validator] Fixed groups argument misplace for validateValue method from validator class
This PR was merged into the 2.2 branch.
Discussion
----------
[DependencyInjection] Add exception for service name not dumpable in PHP
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8485#8030
| License | MIT
| Doc PR | n/a
Throws an exception when the DIC is dumped to PHP, before generating invalid PHP.
The regex comes from the PHP doc: http://www.php.net/manual/en/language.oop5.basic.php
Commits
-------
242b318 [DependencyInjection] Add exception for service name not dumpable in PHP
This PR was merged into the 2.2 branch.
Discussion
----------
[Security] fixed issue where x509 authentication clears unrelated tokens
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8226
| License | MIT
| Doc PR | symfony/symfony-docs#2825
| Notes | Replaces PR #8283
TODO:
- [x] Feedback on change to make sure security is not affected
- [x] Fix other authentication listeners (they suffer the same problem)
- [x] Write unit tests for bug and maybe a few listener classes as well
This pull request is the summary of the problem mentioned in the ticket above.
It only fixes the "disappearing token" problem for one authentication provider, not all. If acceptable, the change needs to be applied to all authentication listeners since they always clear all tokens from the security context.
Commits
-------
2317443 [Security] fixed issue where authentication listeners clear unrelated tokens
This commit fixes an issue where authentication listeners clear all security tokens in case of authentication failure.
This behavior makes it impossible to combine certain authentication mechanisms, notably x509 with form-based login.
This PR was merged into the 2.2 branch.
Discussion
----------
[Console] Show completed percentage, not a rounded value.
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
This pull request has 2 minor changes:
- 199/200 => 99% instead of 100% (there was no test case and I added one, everything else passes) In the past the value was rounded, the progress bar ended up being 100% for 0.5% of the total execution time. That means 18 seconds on a 1 hour process.
- Fixed the comments to match the inner workings of RedrawFrequency.
First Symfony PR and second overall so hope I just did this the right way.
PR #8419 Rebased for 2.2
Commits
-------
91bb757 ProgressHelper shows percentage complete.
40591b9 Comment fixed: RedrawFrequency is measured in steps.
In the past the value was rounded, the progressbar ended up being 100% for 0.5%
of the total execution time. That means 18 seconds on a 1 hour process.
