Commits
-------
d664f5f Changed the wording of the exception thrown when Bundle::getAlias() returns something different to the Container::underscore version, fixes#1768
Discussion
----------
Changed exception wording
Changed the wording of the exception thrown when Bundle::getAlias() returns something different to the Container::underscore version, fixes#1768
Commits
-------
9d8e6f6 [FrameworkBundle] Changed TraceableEventDispatcher to log calls to event listeners _before_ actually calling them
Discussion
----------
[FrameworkBundle] Log calls to event listereners _before_ calling them
The current implementation of `TraceableEventDispatcher` logs calls to event listeners _after_ actually calling them. This leads to strange logs when an event listener triggers another event.
For example, if I attach some `LoginListener` to the `security.interactive_login`-event, the log will look something like this:
<pre>
...
User "myusername" has been authenticated successfully
Notified event "security.interactive_login" to listener "MyVendor\MyBundle\EventListener\LoginListener::onSecurityInteractiveLogin".
Notified event "kernel.request" to listener "Symfony\Component\Security\Http\Firewall::onKernelRequest".
...
</pre>
From the logs it looks like the `kernel.request` event was fired after the user was authenticated, whereas it was actually the listener to `kernel.request` that caused the user to be authenticated.
By logging the call to the event listener _before_ calling it, the logs will look like this:
<pre>
...
Notified event "kernel.request" to listener "Symfony\Component\Security\Http\Firewall::onKernelRequest".
Notified event "security.interactive_login" to listener "MyVendor\MyBundle\EventListener\LoginListener::onSecurityInteractiveLogin".
User "myusername" has been authenticated successfully
...
</pre>
In my opinion this makes the causal relationship between the events clearer.
---------------------------------------------------------------------------
by stof at 2011/07/24 11:18:48 -0700
👍 for this.
Commits
-------
5219f81 Using the $status parameter instead of fixed value when creating a RedirectResponse.
Discussion
----------
Using the $status parameter instead of fixed value
I checked the usages and the optional `$status` parameter is never used, so maybe another option would be to remove the parameter completely...
---------------------------------------------------------------------------
by jaugustin at 2011/07/25 03:11:00 -0700
maybe you could test that $status is a valid redirect code
---------------------------------------------------------------------------
by stloyd at 2011/07/25 04:40:21 -0700
@jaugustin This check is already included in `RedirectResponse` class.
Commits
-------
266e60e Don't tell a lie to every WebServers
Discussion
----------
Please don't tell a lie to every WebServers
Fake Useragent name should be only in test case .
Commits
-------
0b3ff39 fix supported cookie date formats
Discussion
----------
fix supported cookie date formats
---------------------------------------------------------------------------
by stof at 2011/07/24 13:16:25 -0700
👍 as the current behavior breaks the crawling of some sites using Goutte (``bit.ly`` for instance)
Commits
-------
0635548 [Process] Fixed tests on Windows by using escapeshellarg to escape argument
Discussion
----------
[Process] Changed tests to use escapeshellarg for escaping arguments
To make the `ProcessTest` work on Windows, the code to pass to `php` needs to be escaped using `escapeshellarg` rather than using hardcoded single quotes. Also, since `escapeshellarg` escapes double quotes incorrectly on Windows, I changed the quotes in the code-to-be-executed to single quotes only.
I have checked that the test still runs correctly on Unix (Ubuntu 10.04).
Commits
-------
bd9af2b [BrowserKit] Fixed incorrect temp dir when running insulated functional tests under Windows
Discussion
----------
[BrowserKit] Fixed incorrect temp dir when running insulated functional tests in Windows
The `Client` class in the `BrowserKit`-component does not pass on a temporary directory environment variable recognized by Windows. This can lead to an incorrect temp dir being used when using an insulated `Client` in a functional test. A fix for Mac was already present, I extended this to also work on Windows.
Commits
-------
86aeb04 [Form] Added skip for tests that require a higher version of the ICU library (needed by intl) than installed
Discussion
----------
[Form] Skip tests that are incompatible with the installed ICU (intl) library
Two tests for the `DateTimeToLocalizedStringTransformer` depend on a datetime-format that has only been added in version 3.8 of the ICU library (used by `intl`), see http://bugs.icu-project.org/trac/changeset/21815#file60. Some PHP-versions still ship with ICU 3.6 however.
This PR skips these tests when an incompatible version of the ICU library is detected. Unfortunately I had to copy-paste some code from `Symfony\Tests\Component\Locale\TestCase`, but I don't see any other way without creating a strange dependency between the tests for the Form and Locale components.
---------------------------------------------------------------------------
by Abhoryo at 2011/07/22 16:42:40 -0700
PHP 5.3.5 = ICU 3.6
PHP 5.3.6 = ICU 3.8
PHP 5.3.7rc4 = ICU 4.6
Since the RC3 of Symfony, config/check script recommends ICU 4.0+
---------------------------------------------------------------------------
by aboks at 2011/07/23 00:33:12 -0700
I noticed, but since it's not a hard requirement (and difficult for some people to install ICU 4.0+ at the moment) this PR would at least prevent false positives when people unable to upgrade run the test suite.
Commits
-------
04d0b6c [Console] Forced undecorated output in tests that rely on Application/Command-output being equal to a fixed value
Discussion
----------
[Console] Fixed tests that rely on undecorated Application/Command-output
Some tests compare the output of an `Application`/`Command` run to a fixed string. The output is decorated however (at least) when run on a Windows box with Ansicon installed.
This PR fixes these failing tests by forcing the `Application`/`Command`-output to be undecorated.
