Commits
-------
e81c710 Increased the priority of the profiler request listener
Discussion
----------
Increased the priority of the profiler request listener
If a request listener returns a response before calling the profiler
listener, the request will not be added in the stack leading to an error
during the handling of the kernel.response event. The profiler listener
should ideally be run first.
The same issue will occur if a previous listener throws an exception btw.
---------------------------------------------------------------------------
by schmittjoh at 2011/10/20 08:21:15 -0700
Can you add a test?
These interdependencies between listeners are really easy to mess up, and no-one will remember why listener X has to be run before listener Y, but after listener Z.
---------------------------------------------------------------------------
by stof at 2011/10/20 08:27:51 -0700
Well, in fact, the only rule is that the listener **must** run for the request event to avoid failing during the response event of an eventual subrequest. So if another listener triggers a subrequest before the profiler is called, it will fail because the parent request is not in the stack.
A better fix would probably make the listener work even when the propagation of the request event has been stopped (or has not reached this listener yet) to avoid issues (someone doing a subrequest in its own listener will still be able to use a higher priority for instance) but I don't have time to refactor @fabpot's refactoring now.
---------------------------------------------------------------------------
by schmittjoh at 2011/10/20 08:36:40 -0700
Another idea would be to add a ProfilerHttpKernel as it basically needs to wrap the request/response cycle, and people might have priorities beyond 1000 (see @Seldaek's SecurityBundle).
Using an extra kernel would be the only way to really guarantee that it is working. Note that I'm not implying you should do it, but as a "real" fix this would probably be better.
---------------------------------------------------------------------------
by stof at 2011/10/20 08:43:07 -0700
The issue by wrapping the HttpKernel is that the kernel passed in the event is the one dispatching the event. So we would have to implement it by using inheritance, not composition (otherwise the listener would create subrequests using the inner kernel and they would not be profiled).
But FrameworkBundle already extend the HttpKernel to handle the creation of the request scope (which must be done using inheritance for the same reason).
So it would require having a ContainerAwareHttpKernel (handling the scope), a ProfilerHttpKernel (handling the profiler) both extending the base HttpKernel (the Profiler one should be usable when using only the component, and the ContainerAware one is needed in prod when disabling the profiler) but also a ContainerAwareProfilerHttpKernel extending one of them and duplicating the logic of the other.
---------------------------------------------------------------------------
by schmittjoh at 2011/10/20 08:52:13 -0700
So maybe, we need to revisit which kernel is passed in the event as well.
But doing things more explicitly is better than making assumptions about priorities that no-one can remember in a few weeks.
---------------------------------------------------------------------------
by stof at 2011/10/20 09:53:45 -0700
@schmittjoh Issue is, using composition would require to let the inner kernel know about the wrapper to be able to pass it in the event. This looks ugly.
But I agree it would be better to do things explicitly
---------------------------------------------------------------------------
by Seldaek at 2011/10/21 07:39:23 -0700
Didn't read everything, busy, but regarding the mention of our security bundle, it is highlighting a big issue with priorities imo, in that we're missing a one-stop page that lists all the existing priorities, and possibly some more details like "if you do this, don't go above this priority or it'll blow up". Otherwise it'll just be a game of escalation :)
Commits
-------
6b872cf Check if cache_warmer service is available before doing the actual cache warmup
40fb76d [Framework] removed wrong listener
Discussion
----------
fix cache warump exception when service is not available
Bug fix: yes
Feature addition: no
Backwards compatibility break: no
Symfony2 tests pass: yes
Fixes the following tickets: N/A
fixes [Symfony\Component\DependencyInjection\Exception\ServiceNotFoundException]
You have requested a non-existent service "cache_warmer". in console when FrameworkBundle is removed from kernel.
Commits
-------
c915f48 Added another "before" possibility to retrieve the locale from a Twig template.
Discussion
----------
Added {{ app.session.locale }} to UPGRADE-2.1.md
Added another "before" possibility to retrieve the locale from a Twig template.
Commits
-------
96235a6 cs fix
46a69f1 define a WarmableInterface and only warm the cache if it implements warmable to allow replacing the core router. this fixes#2422. combining routers will only really work when #2450 is merged too.
Discussion
----------
define a WarmableInterface and only warm the cache if it implements warmable
define a WarmableInterface and only warm the cache if it implements warmable to allow replacing the core router. this fixes#2422. combining routers will only really work when #2450 is merged too.
Bug fix: yes
Feature addition: yes
Backwards compatibility break: no
Symfony2 tests pass: yes*
Fixes the following tickets: #2422
(*) tests:
success for
phpunit src/Symfony/Bundle/FrameworkBundle/Tests/
and
phpunit tests/Symfony/Tests/Component/Routing/
but when running all tests, i have to put gc_disable() into autoload to avoid segmentation fault and then get (after 3000 successful tests):
PHP Fatal error: Call to undefined method Symfony\Tests\Component\HttpKernel\Debug\StopwatchTest::assertCount() in /home/david/liip/symfony-cmf/cmf-sandbox/vendor/symfony/tests/Symfony/Tests/Component/HttpKernel/Debug/StopwatchTest.php on line 84
---------------------------------------------------------------------------
by stof at 2011/10/22 08:46:31 -0700
@dbu assertCount is new in PHP 3.6. It seems like the test of the new 2.1 feature has been written using 3.6-RC.
---------------------------------------------------------------------------
by dbu at 2011/10/22 09:40:07 -0700
@stof: ah, thanks for the hint. i hope you mean php 3.4 and not 3.6? but i assume symfony 2.1 should be able to run on 3.3, right?
anyways, the tests run through if i disable the StopWatch, so i guess we can consider the tests succeeding.
---------------------------------------------------------------------------
by stof at 2011/10/22 09:41:28 -0700
this is a method of PHPUnit TestCase class. I was talking about the PHPUnit version
Commits
-------
6b02ffb fix a typo in the routing dependency injection configuration. the request_context service is named router, not routing
Discussion
----------
fix a typo in the routing dependency injection configuration
fix a typo in the routing dependency injection configuration. the request_context service is named router.request_context , not routing.request_context
Bug fix: yes
Feature addition: no
Backwards compatibility break: ? - should not unless somebody relied on the bug
Symfony2 tests pass: yes
Fixes the following tickets: no ticket
i ran
phpunit src/Symfony/Bundle/FrameworkBundle/Tests/
and
phpunit tests/Symfony/Tests/Component/Routing/
If a request listener returns a response before calling the profiler
listener, the request will not be added in the stack leading to an error
during the handling of the kernel.response event. The profiler listener
should ideally be run first.
Commits
-------
79638d3 Removed abstract as these services will not appear anyway
312b20f [FrameworkBundle] Added more info to debug command output
Discussion
----------
[FrameworkBundle] Added more info to debug command output
---------------------------------------------------------------------------
by stof at 2011/10/18 05:57:59 -0700
This commands uses the optimized container so abstract services will never appear there.
That allows projects that only use HttpFoundation and not HttpKernel to be able to
enforce the HTTP specification "rules".
$request = Request::createFromGlobals();
$response = new Response();
// do whatever you want with the Respons
// enforce HTTP spec
$response->prepare($request);
$response->send();
Within Symfony2, the prepare method is automatically called by the ResponseListener.
Commits
-------
ae342c7 unified toolbar.css
a7e4e70 unified profiler.css
09fe09e unified and corrected exception_layout.css
3a1674b unified and corrected exception.css
Discussion
----------
Unified and corrected CSS markup
Unified (spaces, braces, quotes, indention) and corrected (missing semicolon) the CSS markup.
Did not change any semantic, only markup!
@fabpot: New pull based on symfony:2.0 and changed formatting style as agreed in #2405
Commits
-------
1768544 Changed debug output to say Service Id rather than name for the service ids
Discussion
----------
Changed debug output to say Service Id rather than name for the service i
Changed debug output to say Service Id rather than name for the service ids
Commits
-------
9546ee6 Removed the IndexedReader
ea2cf73 Refactored the validator initializer
c6063ec [DoctrineBundle] Updated the code to use the new registry
a1784c2 [DoctrineBridge] Updated the code to use the new registry
Discussion
----------
Doctrine registry
This updates the Doctrine bridge and DoctrineBundle to use the new ManagerRegistry interface and its methods instead of using the old methods which are marked as deprecated.
Commits
-------
18cca22 cleaned up the use statements in the Registry class and RegistryInterface
Discussion
----------
cleaned up the use statements in the Registry class and RegistryInterface
Bug fix: no
Feature addition: no
Backwards compatibility break: no
Symfony2 tests pass: yes
Fixes the following tickets: -
just cosmetics ..
Commits
-------
63d2ce2 [FrameworkBundle] Fixed the ckeck for the router class
Discussion
----------
[FrameworkBundle] Fixed the ckeck for the router class
The getRouteCollection method is now part of the RouterInterface so the
command should accept any implementation of the interface instead of just
the implementations extending the core one.
