This PR was squashed before being merged into the 3.4 branch (closes#27668).
Discussion
----------
[Lock] use 'r+' for fopen (fixes issue on Solaris)
| Q | A
| ------------- | ---
| Branch? | 3.4 (also applicable to _LockHandler_ in 2.8 and 3.3)
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes [1]
| Fixed tickets | -
| License | MIT
| Doc PR | -
We discovered a curious case on a specific hosting environment: the `FlockStorage` (or `LockHandler` in previous Symfony versions) could never get a file lock on **existing files**. So if you run a script for the first time and the lock file did not exist yet, the `FlockStorage` could get a lock on that file just fine. However on the second and subsequent runs, `FlockStorage` could _never_ get a file lock anymore. You can follow the discussion [here](https://github.com/contao/core-bundle/issues/1551) (if you speak German).
We have been using this little script to confirm the issue on the hosting environment:
```php
$fileName = __DIR__ . '/file.lock';
if (!$handle = @fopen($fileName, 'r')) {
$handle = fopen($fileName, 'x');
}
if (!$handle) {
echo "Could not open $fileName\n";
exit;
}
if (flock($handle, LOCK_EX | LOCK_NB)) {
echo "Got a lock on $fileName\n";
flock($handle, LOCK_UN | LOCK_NB);
} else {
echo "Could not get a lock on $fileName\n";
}
```
Whenever `file.lock` already existed prior to running the script, a lock could not be made.
After contacting the hosting provider's support on this they confirmed the issue and told us they are using **Solaris** instead of a Linux environment. And this is supposedly why it does not work. Instead you have to use `'r+'` instead of `'r'` for `fopen`.
I was able to confirm that changing from `'r'` to `'r+'` fixes the issue. However I am wondering who's actually at fault here. Is it Solaris? PHP? The compiled PHP version under Solaris? The hosting provider's operating system configuration?
### System information
```
uname -a
SunOS vlek 5.11 11.3 i86pc i386 i86pc Solaris
```
```
phpinfo:
SunOS localhost 5.10 Generic_150401-49 i86pc
```
```
php -v
PHP 7.2.5 (cli) (built: May 4 2018 12:57:43) ( NTS )
Copyright (c) 1997-2018 The PHP Group
Zend Engine v3.2.0, Copyright (c) 1998-2018 Zend Technologies
```
### Footnotes
[1] Previously one failed test on the first run and two failing tests on the second run with the proposed changes:
```
1) Symfony\Component\Lock\Tests\Store\FlockStoreTest::testSaveSanitizeName
Symfony\Component\Lock\Exception\LockStorageException: fopen(C:\Users\[…]\AppData\Local\Temp/sf.-php-echo-hello-word-.Sz2dDSf.lock)
: failed to open stream: Permission denied
```
```
2) Symfony\Component\Lock\Tests\Store\FlockStoreTest::testSaveWithDifferentKeysOnSameResources
Symfony\Component\Lock\Exception\LockStorageException: fopen(C:\Users\[…]\AppData\Local\Temp/sf.Symfony-Component-Lock-Tests-Store-
AbstractStoreTest-testSaveWithDifferentKeysOnSameResources5b2b5f00872538.64807920.2u9bH+a.lock): failed to open stream: Permission denied
```
The latter failed both on the first run and on the second run.
After the [proposed changes](https://github.com/symfony/symfony/pull/27668#discussion_r197054198) from @nicolas-grekas everything works fine 👍
Commits
-------
9c9ae7d9c9 [Lock] use 'r+' for fopen (fixes issue on Solaris)
* 2.8:
[HttpKernel] fix test compat with PHP 5.3
fix file lock on SunOS
change `evaluate()` docblock return type from string to mixed
Set serialize_precision explicitly to avoid fancy float rounding
This PR was merged into the 2.8 branch.
Discussion
----------
Set serialize_precision explicitly to avoid fancy float rounding
| Q | A
| ------------- | ---
| Branch? | 2.8
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | none
| License | MIT
| Doc PR | -
This is fixing some of the test failures I was seeing locally due to increased `serialize_precision` INI setting:
```
2) Symfony\Component\HttpFoundation\Tests\JsonResponseTest::testConstructorWithSimpleTypes
Failed asserting that two strings are identical.
--- Expected
+++ Actual
@@ @@
-'0.1'
+'0.10000000000000001'
/www/symfony/symfony/src/Symfony/Component/HttpFoundation/Tests/JsonResponseTest.php:46
3) Symfony\Component\HttpFoundation\Tests\JsonResponseTest::testStaticCreateWithSimpleTypes
Failed asserting that two strings are identical.
--- Expected
+++ Actual
@@ @@
-'0.1'
+'0.10000000000000001'
```
The test assertions should not depend on externally configured PHP configuration.
Also default value for this option was changed multiple times: http://php.net/manual/en/ini.core.php#ini.serialize-precision
For compatibility reasons (with PHP <7.x) `-1` can't be used.
https://3v4l.org/HBNsT
HHVM doesn't seem to support this though, how to handle this?
Commits
-------
b5ee7c3ccd Set serialize_precision explicitly to avoid fancy float rounding
This PR was merged into the 3.4 branch.
Discussion
----------
[Config] Fixing GlobResource when inside phar archive
| Q | A
| ------------- | ---
| Branch? | master
| Bug fix? | yes
| New feature? | no
| BC breaks? | yes if old broken behavior counts as stable
| Deprecations? | no
| Tests pass? | no tests yet
| Fixed tickets |
| License | MIT
| Doc PR | N/A
When packaging an Sf4 application as a PHAR archive using globs at various locations (`Kernel`, `services.yaml`) most glob files are not found because the `glob()` PHP method [does not support PHAR streams](https://stackoverflow.com/questions/8203188/unexpected-problems-with-php-phar).
Using the regex fallback instead when operating inside PHAR archives fixes the behavior for me.
## Examples:
`src/Kernel.php::configureContainer()`:
```php
$loader->load($confDir.'/{services}'.self::CONFIG_EXTS, 'glob');
```
Expected behavior: `config/services.yaml` inside PHAR archive is found and parsed
Actual behavior: the file will not be loaded
`config/services.yaml` (hard-coded in Kernel without using glob pattern)
```yaml
App\:
resource: '../src/*'
exclude: '../src/{Entity,Migrations,Tests,Kernel.php}'
```
Expected behavior: service classes in `src/` will be found and auto-wired
Actual behavior: services are not auto-wired because the class files are not found
Commits
-------
e336ebeecf Fixing GlobResource when inside phar archive
This PR was merged into the 3.4 branch.
Discussion
----------
[Cache] fix Memcached tests
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
Clearing a memcached pool works only when versioning is enabled. Dunno why this has not be caught before, but it's making tests fail now.
Commits
-------
09ec9e7cce [Cache] fix Memcached tests
* 2.8:
[HttpKernel] fix PHP 5.4 compat
Fix surrogate not using original request
[Finder] Update RealIteratorTestCase
[Routing] remove unneeded dev dep on doctrine/common
[Validator] Remove BOM in some xlf files
When packaging an Sf4 application as a PHAR archive using globs at various locations (`Kernel`, `services.yaml`) most glob files are not found because the `glob()` PHP method [does not support PHAR streams](https://stackoverflow.com/questions/8203188/unexpected-problems-with-php-phar).
Using the regex fallback instead when operating inside PHAR archives fixes the behavior for me.
This PR was merged into the 3.4 branch.
Discussion
----------
[Form] Fix error when rendering a DateIntervalType form with exactly 0 weeks
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #27378
| License | MIT
Fixes the issue described in #27378
Commits
-------
dae704ad2fFix#27378: Error when rendering a DateIntervalType form with exactly 0 weeks
This PR was merged into the 3.4 branch.
Discussion
----------
[HttpKernel] fix session tracking in surrogate master requests
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
Spotted while looking at ESI fragments resolved by`HttpCache`: right now when the master request starts the session, fragments are not cacheable anymore, even when they do not use the session.
Commits
-------
146e01cb44 [HttpKernel] fix session tracking in surrogate master requests
This PR was submitted for the master branch but it was squashed and merged into the 2.8 branch instead (closes#27508).
Discussion
----------
[Finder] Update RealIteratorTestCase
| Q | A
| ------------- | ---
| Branch? | 2.8
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #27480
| License | MIT
| Doc PR | n/a
Makes the entire test directory empty instead of trying to delete particular files and directories. The old method failed when trying to remove a directory which was not empty.
Commits
-------
7d0ebd41ab [Finder] Update RealIteratorTestCase
This PR was merged into the 2.8 branch.
Discussion
----------
[VarDumper] Fix dumping ArrayObject and ArrayIterator instances
| Q | A
| ------------- | ---
| Branch? | 2.8
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
Properties added on child classes of `ArrayObject` and `ArrayIterator`, or dynamic properties added on instances of them were now properly dumped. This fixes it.
![image](https://user-images.githubusercontent.com/243674/41349429-2660cbc6-6f10-11e8-8015-a3d6ad8b0c9c.png)
Commits
-------
3ecabfc36e [VarDumper] Fix dumping ArrayObject and ArrayIterator instances
This PR was squashed before being merged into the 3.4 branch (closes#27556).
Discussion
----------
Avoiding session migration for stateless firewall UsernamePasswordJsonAuthenticationListener
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | Related to #27395
| License | MIT
| Doc PR | symfony/symfony-docs#9860
This is the sister PR to #27452, which covered all the other authentication listeners.
Commits
-------
c06f3229de Avoiding session migration for stateless firewall UsernamePasswordJsonAuthenticationListener
This PR was squashed before being merged into the 2.8 branch (closes#27452).
Discussion
----------
Avoid migration on stateless firewalls
| Q | A
| ------------- | ---
| Branch? | 2.8
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | Related to #27395
| License | MIT
| Doc PR | symfony/symfony-docs#9860
This is a proof-of-concept. Once we agree / are happy, I need to add this to all of the other authentication mechanisms that recently got the session migration code & add tests.
Basically, this avoids migrating the session if the firewall is stateless. There were 2 options to do this:
A) Make the `SessionAuthenticationStrategy` aware of all stateless firewalls. **This is the current approach**
or
B) Make each individual authentication listener aware whether or not *its* firewall is stateless.
Commits
-------
cca73bb564 Avoid migration on stateless firewalls
This PR was squashed before being merged into the 3.4 branch (closes#27326).
Discussion
----------
[Serializer] deserialize from xml: Fix a collection that contains the only one element
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #27279
| License | MIT
| Doc PR |
In xml when parent node (`restaurants`) contains several children nodes with the same tag (`restaurant`) it is clear that the children form a collection:
```
restaurants = {array} [1]
restaurant = {array} [2]
0 = {array} [2]
name = "Some restaurant name"
type = "Chinese"
1 = {array} [2]
name = "Another restaurant name"
type = "Italian"
```
Afterwards the object denormalizer has no problem to create a collection of restaurants.
But when there is only one child (`restaurant`) the decoded normalized array will not contain a collection:
```
restaurants = {array} [1]
restaurant = {array} [2]
name = "Some restaurant name"
type = "Chinese"
```
In this situation the object denormalizer threw unexpected exception. This PR modifies `AbstractObjectNormalizer` that is it will fill a collection containing the sole element properly.
Commits
-------
1f346f446d [Serializer] deserialize from xml: Fix a collection that contains the only one element