This PR was merged into the 2.3 branch.
Discussion
----------
[DependencyInjection] Fixed YamlFileLoader imports path
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10516
| License | MIT
| Doc PR | http://symfony.com/doc/current/book/service_container.html#importing-configuration-with-imports
YamlFileLoader used the resource name as the current directory during import, which can cause a failed import when using a relative path. Using the $path variable output from the locator is consistent with other loaders and fixes the bug.
Commits
-------
fd1d48b Fixed YamlFileLoader imports path
YamlFileLoader used the resource name as the current directory during import, which can cause a failed import when using a relative path. Using the $path variable output from the locator is consistent with other loaders and fixes the bug.
This PR was merged into the 2.3 branch.
Discussion
----------
[Debug] Check headers sent before sending PHP response
If the response contents has been sent before an error occurs, PHP triggers the warning "Cannot modify header information - headers already sent".
This change ensure that the useful error message is echoed, while it's impossible to change the HTTP status code and headers.
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Commits
-------
97591c1 Check headers sent before sending PHP response
If the response contents has been sent before an error occurs, PHP
triggers the warning "Cannot modify header information - headers already sent"
This change ensure that the error message is echoed, while it's impossible
to change the HTTP status code and headers.
This PR was merged into the 2.3 branch.
Discussion
----------
[Validator] Fixed ACE domain checks on UrlValidator (#10031)
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10031
| License | MIT
| Doc PR |
I added some checks to allow internationalised domain names as per reported on #10031 and some extra testcases to cover ACE domain names.
Commits
-------
7a2f154 Fixed ACE domain checks on UrlValidator (#10031)
This PR was submitted for the 2.4 branch but it was merged into the 2.3 branch instead (closes#10123).
Discussion
----------
handle array root element
An array to string conversion notice was thrown when the root element of the thing being validated is an array.
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets |
| License | MIT
Commits
-------
17ed8bf handle array root element
This PR was merged into the 2.3 branch.
Discussion
----------
Fixed regression when using Symfony on filesystems without chmod support
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #8205
| License | MIT
| Doc PR | n/a
Commits
-------
cefb67aFix#8205 : Deprecate file mode update when calling dumpFile
This PR was merged into the 2.3 branch.
Discussion
----------
[HttpKernel] Fix#10437: Catch exceptions when reloading a no-cache request
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10437
| License | MIT
Commits
-------
8e3a395Fix#10437: Catch exceptions when reloading a no-cache request
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3] Fix libxml_use_internal_errors and libxml_disable_entity_loader usage
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #9731, #9483
| License | MIT
This should fix#9731 and #9483 that seems to be triggered when `libxml_disable_entity_loader` has been called with `true` (see https://bugs.php.net/bug.php?id=62577)
As `libxml_disable_entity_loader` is non thread safe (see https://bugs.php.net/bug.php?id=64938) and as we have some calls that might leave the setting to `true`, I think the bug should be fixed.
I've checked the use of both `libxml_use_internal_errors` and `libxml_disable_entity_loader` among symfony code.
You can see I prefered to skip DomDocument::loadXML warnings using the `@` instead of using `LIBXML_NOERROR | LIBXML_NO_WARNING` because we can log these errors whereas libxml errors would be furtives.
- [x] Check calls to DOMDocument::load
- [x] Check calls to DOMDocument::loadXML
- [x] Check calls to DOMDocument::loadHTML
- [x] Check calls to DOMDocument::loadHTMLFile
- [x] Add more tests
Commits
-------
489b8ae Fix libxml_use_internal_errors and libxml_disable_entity_loader usage
This PR was submitted for the master branch but it was merged into the 2.3 branch instead (closes#9784).
Discussion
----------
[HttpFoundation] Removed ini check to make Uploadedfile work on Google App Engine
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | yes
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
related to #9238
Commits
-------
ecb6481 removed ini check to make uploadedfile work on gae
This PR was submitted for the 2.3-dev branch but it was merged into the 2.3 branch instead (closes#10528).
Discussion
----------
Update OptionsResolver.php
Deleting an extra space in the exception.
Commits
-------
7507fbb Update OptionsResolver.php
This PR was submitted for the 2.3-dev branch but it was merged into the 2.3 branch instead (closes#10525).
Discussion
----------
[HttpKernel] Clean KernelInterface docblocks
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets |
| License | MIT
| Doc PR |
- Fixed questionable grammar
- Added missing periods to make it consistent
Commits
-------
4b1f050 Clean KernelInterface docblocks
This PR was submitted for the 2.3-dev branch but it was merged into the 2.3 branch instead (closes#10416).
Discussion
----------
[Form] Allow options to be grouped by objects
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets |
| License | MIT
| Doc PR |
Commits
-------
ae3f02f Cast the group name as a string
This PR was submitted for the 2.3-dev branch but it was merged into the 2.3 branch instead (closes#10497).
Discussion
----------
[SecurityBundle] Fixed doc of InitAclCommand
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | no
| License | MIT
| Doc PR | n/a
Use {@inheritdoc}. Consistency with https://github.com/symfony/symfony/pull/9990#discussion-diff-10219606.
Commits
-------
aa49009 [SecurityBundle] Fixed doc of InitAclCommand
This PR was squashed before being merged into the 2.3 branch (closes#10410).
Discussion
----------
[Form] Fix "Array was modified outside object" in ResizeFormListener.
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10405
| License | MIT
| Doc PR |
This is the second pull request for this issue. The history of this is as follows:
Original Fix was added Feb 11th under Pull Request #10232.
Users began complaining of Doctrine ArrayCollection not being updated in forms.
Revert was added Feb 15th under Pull Request #10269.
Issue #10405 was opened on 7th Mar for the original bug.
Pull Request #10269 has a failing test that illustrates users concerns.
I have added failing tests to this pull request to illustrate the problems described in #10405.
All tests now pass, and all forms of $data are now supported, including arrays, DoctrineCollection, ArrayObject, and other IteratorAggregates.
__Details as follows:__
The onSubmit() method of the ResizeFormListener class is assuming the data is an array, and calling unset directly inside a foreach. This works fine in most scenarios, but if data is an instance of IteratorAggregate that is iterating over $data by reference, it breaks with the following error:
Symfony\Component\Form\Extension\Core\EventListener\ResizeFormListener::onSubmit(): ArrayIterator::next(): Array was modified outside object and internal position is no longer valid in ./vendor/symfony/symfony/src/Symfony/Component/Form/Extension/Core/EventListener/ResizeFormListener.php line 142
This is because the foreach loop is using an Iterator in the background, but the ResizeFormListener has unset the underlying data directly, causing the value that the Iterators internal pointer is pointing to to change.
The Iterator provided by IteratorAggregate may or may not have the ability to modify it's underlying data. So it is not possible to rely on the Iterator to modify the data. Instead, it is simpler to avoid modifying $data at all while we are iterating over it. So instead, we simply iterate over $data once to determine the keys we need to delete, and store them. Then we iterate over the keys and delete them from $data.
Commits
-------
aa63fae [Form] Fix "Array was modified outside object" in ResizeFormListener.
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3][Validator] Minor fix in IBAN validator
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10481, #10489
| License | MIT
Added more values to unit tests of IBAN validator to make clear it doesn't accept lower case letters.
> Permitted IBAN characters are the digits 0 to 9 and the 26 upper-case Latin alphabetic characters A to Z.
http://en.wikipedia.org/wiki/International_Bank_Account_Number
Also made little adjustment to code which meant to validate lowercase letters but actually was useless.
Commits
-------
3eeb306 Fix IBAN validator
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3][Process] Remove unreachable code + avoid skipping tests in sigchild environment
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| License | MIT
As mentioned by @Tobion in https://github.com/symfony/symfony/pull/10480#issuecomment-38002910, I removed the dead code. I also fixed/updated the test suite on PHP compiled with `--enable-sigchild`.
Commits
-------
d52dd32 [Process] Remove unreachable code + avoid skipping tests in sigchild environment
This PR was submitted for the 2.4 branch but it was merged into the 2.3 branch instead (closes#10491).
Discussion
----------
Fixed bug that incorrectly causes the "required" attribute to be omitted from select even though it contains the "multiple" attribute
According to the HTML5 specification the required attribute should be allowed in this case: "A select element with a required attribute and **without** a multiple attribute, and whose size is “1”, must have a child option element."
Related PR: https://github.com/symfony/symfony/pull/9030
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| License | MIT
Commits
-------
f9aa676 Fixed bug that incorrectly causes the "required" attribute to be omitted from select even though it contains the "multiple" attribute
This PR was submitted for the 2.3-dev branch but it was merged into the 2.3 branch instead (closes#10142).
Discussion
----------
[Tests] Added travis_retry to .travis.yml
| Q | A
| ------------- | ---
| Bug fix? | -
| New feature? | -
| BC breaks? | -
| Deprecations? | -
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
As stated on their [blog](http://blog.travis-ci.com/2013-05-20-network-timeouts-build-retries/), this is the recommended way.
Example of a failed build because of it:
https://travis-ci.org/symfony/symfony/jobs/17664775
Commits
-------
702432a Added travis_retry to .travis.yml
This PR was merged into the 2.3 branch.
Discussion
----------
[Process] fix some typos and refactor some code
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
Commits
-------
b422613 [Process] fix some typos and refactor some code
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3][Process] Fix escaping on Windows
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| License | MIT
Windows escaping is broken since the last merges.
After digging more on Windows escaping, I realised some things:
- We forbid environment variable expansion by escaping `%APPDATA%` to `^%"APPDATA"^%`
- We explicitly ask for variable expansion at runtime (running the command line with the [`/V:ON`](https://github.com/symfony/symfony/blob/2.3/src/Symfony/Component/Process/Process.php#L235) flag). Running a command containing `!APPDATA!` will be escaped and expanded (our previous rule is easily overriden)
- On platform that are not windows, we use strong escaping that prevents any variable expansion (`$PATH` will be escaped to `'$PATH'` that is not interpreted as the current PATH)
We have three possibilities:
- Keep this behavior as this.
- Prefer a consistent API and use a strong escaping strategy everywhere, but it would result in a BC break (see #8975).
- Allow environment variable expansion and escape `%APPDATA%` to `"%APPDATA%"`
Any thoughts about this ?
Commits
-------
0f65f90 [Process] Fix escaping on Windows
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3][Process] Fix unit tests in sigchild disabled environment
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| License | MIT
We've been a bit fast when merging #10480 and tests were broken in the last update I did.
Commits
-------
5f6ee12 [Process] Fix unit tests in sigchild disabled environment
This PR was merged into the 2.3 branch.
Discussion
----------
[Process] Fixed fatal errors in getOutput and getErrorOutput when process was not started
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #10022
| License | MIT
This PR replaces #9452 and address the latest changes.
Side note : I've not updated `getExitCode`, `getExitCodeText` and `isSuccessful` as they were explicitly tested to return null in case the process is not started or terminated. I think it would be a BC break to do that.
Commits
-------
449fe01 [Process] Trow exceptions in case a Process method is supposed to be called after termination
0ae6858 [Process] fixed fatal errors in getOutput and getErrorOutput when process was not started
This PR was merged into the 2.3 branch.
Discussion
----------
[2.3][Process] Make Process::start non-blocking on Windows platform
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #9755
| License | MIT
This PR is a proposition that solves issue #9755.
Let me sum-up what problems we have on Windows platform:
- We can not use pipes with `proc_open` on Windows because `stream_set_blocking` does not work on this platform, resulting in blocking pipes and a blocking `Process::start` (see https://bugs.php.net/bug.php?id=51800, https://bugs.php.net/bug.php?id=47918 and https://bugs.php.net/bug.php?id=50856).
- We can not use file handles for both `STDOUT` and `STDERR` as it might produce corrupted content (see https://bugs.php.net/bug.php?id=65650).
- We currently use file handles for `STDOUT` and pipe for `STDERR` but it makes `Process::start` blocking.
The solution I propose here is to use native redirections, write `STDERR` / `STDOUT` in temporary files, read these files, finally cleanup.
It works pretty well, all tests pass on Windows. Better : [previous tests that were specific to this platform](https://github.com/symfony/symfony/blob/master/src/Symfony/Component/Process/Tests/AbstractProcessTest.php#L720-725) disappear as unix one are now okay :).
The drawback of this is the need of using `taskkill`: When stopping a process (`Process::stop` or in case of a timeout reached) `proc_terminate` does not kill the underlying process properly, only the `cmd` parent. The underlying process run by Process still runs after the `proc_terminate` call, having a lock on the temporary files we're using for getting the output, avoiding PHP to remove them (or any user on system) until the underlying process has finish to run).
So I use the `taskkill` Windows command to terminate the underlying process. This works well but I've to admit it's not pretty. If we do not use this hack, let's face that the underlying process is not stopped.
Last but not least, let's deal with the case we're running on Windows platform with PHP having --enable-sigchild environment (I don't know if it can really happen)
In this case:
- we can not retrieve the `pid`
- we can not `taskkill` the underlying process
- it runs
- locks remain on temporary files, we might not be able to remove them
We can't really deal with this.
Feedback more than welcome
Commits
-------
1f5bf32 [Process] Make Process::start non-blocking on Windows platform
