* 2.1:
Fix several instances of doubled words
[Finder] Fix iteration fails with non-rewindable streams
[Finder] Fix unexpected duplicate sub path related AppendIterator issue
Added type of return value in VoterInterface.
Fixed two bugs in HttpCache
Conflicts:
src/Symfony/Component/Finder/Tests/FinderTest.php
1. 304 responses always send "Content-Type: text/html; charset=UTF-8"
header
I discovered that the HttpCache::handle method calls Response::prepare
after calling Response::isModified. Response::isModified removes the
Content-Type header as it should, but Response::handle adds in the
default Content-Type header when none is set. If the default
Content-Type is not the correct Content-Type, then the Content-Type in
the cache gets clobered. I solved this problem by moving the
Response::isModified call after the Response::prepare call. I updated
the testRespondsWith304WhenIfModifiedSinceMatchesLastModified and
testRespondsWith304WhenIfNoneMatchMatchesETag tests to verify that the
Content-Type header was not being sent for 304 responses.
2. Failure to invalidate cached entities referred to by the Location
header
I discovered that the Store::invalidate method was looking for Location
and Content-Location headers to invalidate, but it was looking in the
request headers instead of the response headers. Because the
Store::invalidate method doesn't take a response, I decided it was
better to move this logic to the HttpCache::invalidate method instead.
I updated the testInvalidatesCachedResponsesOnPost test to verify that
Location headers are getting invalidated correctly.
* 2.1:
remove validation related headers when needed
use while loop for iterating
[Filesystem] copy() is not working when open_basedir is set
Conflicts:
src/Symfony/Component/HttpKernel/Profiler/FileProfilerStorage.php
This PR was squashed before being merged into the 2.2 branch (closes#7253).
Discussion
----------
[2.2] Pass ESI header to subrequests
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #7252
| License | MIT
Commits
-------
af819a7 [2.2] Pass ESI header to subrequests
* 2.1:
Add a public modifier to an interface method
[HttpRequest] fixes Request::getLanguages() bug
[HttpCache] added a test (cached content should be kept after purging)
[DoctrineBridge] Fixed non-utf-8 recognition
[Security] fixed HttpUtils class tests
This PR was merged into the 2.2 branch.
Commits
-------
54d7d25 [HttpKernel] hinclude fragment renderer must escape URIs properly to return valid html
Discussion
----------
[HttpKernel] hinclude fragment renderer must escape URIs properly to return valid html
| Q | A
| ------------- | ---
| Bug fix? | [yes]
| New feature? | [no]
| BC breaks? | [no]
| Deprecations? | [no]
| Tests pass? | [yes]
| Fixed tickets | [-]
| License | MIT
| Doc PR | [-]
Since rendering of hinclude fragments returns html/xml, it is marked as safe. So it's not auto-escaped of course. But that means it must properly escape it's input (the URI) when outputting in html context.
Btw, this does not need to be done for esi because esi tags are processed in middleware which do not go to the client/browser.
---------------------------------------------------------------------------
by Koc at 2013-02-15T22:59:05Z
Will it works correct when `arg_separator.output="&"`?
---------------------------------------------------------------------------
by stof at 2013-02-15T23:04:01Z
if your url comes form the routing, yes. It [does not rely on the default separator](https://github.com/symfony/Routing/blob/master/Generator/UrlGenerator.php#L265) to avoid issues when the separator is configured to ``&`` as it would have been escaped again in Twig templates for instance.
---------------------------------------------------------------------------
by fabpot at 2013-02-16T07:26:19Z
Can you include the proper PR header in the description? Thanks.
---------------------------------------------------------------------------
by Tobion at 2013-02-16T12:28:18Z
Added.
* 2.1:
[FrameworkBundle] tweaked reference dumper command (see #7093)
[HttpKernel] added some tests for previous merge
Fix REMOTE_ADDR for cached subrequests
[Process] Warn user with a useful message when tmpfile() failed
Conflicts:
src/Symfony/Bundle/FrameworkBundle/Command/ConfigDumpReferenceCommand.php
* 2.1:
added support for the X-Forwarded-For header (closes#6982, closes#7000)
fixed the IP address in HttpCache when calling the backend
[EventDispatcher] Added assertion.
[EventDispathcer] Fix removeListener
[DependencyInjection] Add clone for resources which were introduced in 2.1
[DependencyInjection] Allow frozen containers to be dumped to graphviz
Fix 'undefined index' error, when entering scope recursively
[Security] fixed session creation on login (closes#7011)
Add dot character `.` to legal mime subtype regular expression
[HttpFoundation] fixed the creation of sub-requests under some circumstancies (closes#6923, closes#6936)
HttpContentRenderer has been renamed to FragmentHandler.
The RendererStrategy subnamespace has been renamed to Fragment.
The strategy classes now have Fragment in their names.
ProxyRouterListener has been renamed to FragmentListener
The router_proxy configuration entry has been renamed to fragments.
The previous code allowed to pass null as a Request but that does not
really make sense as rendering a sub-request can only happen from a
master request. This was done to ease testing but that was a mistake.
This PR was merged into the master branch.
Commits
-------
76fefe3 updated CHANGELOG and UPGRADE files
f7da1f0 added some unit tests (and fixed some bugs)
f17f586 moved the container aware HTTP kernel to the HttpKernel component
2eea768 moved the deprecation logic calls outside the new HttpContentRenderer class
bd102c5 made the content renderer work even when ESI is disabled or when no templating engine is available (the latter being mostly useful when testing)
a8ea4e4 [FrameworkBundle] deprecated HttpKernel::forward() (it is only used once now and not part of any interface anyway)
1240690 [HttpKernel] made the strategy a regular parameter in HttpContentRenderer::render()
adc067e [FrameworkBundle] made some services private
1f1392d [HttpKernel] simplified and enhanced code managing the hinclude strategy
403bb06 [HttpKernel] added missing phpdoc and tweaked existing ones
892f00f [HttpKernel] added a URL signer mechanism for hincludes
a0c49c3 [TwigBridge] added a render_* function to ease usage of custom rendering strategies
9aaceb1 moved the logic from HttpKernel in FrameworkBundle to the HttpKernel component
Discussion
----------
[WIP] Kernel refactor
Currently, the handling of sub-requests (including ESI and hinclude) is mostly done in FrameworkBundle. It makes these important features harder to implement for people using only HttpKernel (like Drupal and Silex for instance).
This PR moves the code to HttpKernel instead. The code has also been refactored to allow easier integration of other rendering strategies (refs #6108).
The internal route has been re-introduced but it can only be used for trusted IPs (so for the internal rendering which is managed by Symfony itself, or by a trusted reverse proxy like Varnish for ESI handling). For the hinclude strategy, when using a controller, the URL is automatically signed (see #6463).
The usage of a listener instead of a controller to handle internal sub-requests speeds up things quite a lot as it saves one sub-request handling. In Symfony 2.0 and 2.1, the handling of a sub-request actually creates two sub-requests.
Rendering a sub-request from a controller can be done with the following code:
```jinja
{# default strategy #}
{{ render(path("partial")) }}
{{ render(controller("SomeBundle:Controller:partial")) }}
{# ESI strategy #}
{{ render(path("partial"), { strategy: 'esi' }) }}
{{ render(controller("SomeBundle:Controller:partial"), { strategy: 'esi' }) }}
{# hinclude strategy #}
{{ render(path("default1"), { strategy: 'hinclude' }) }}
```
The second commit allows to simplify the calls a little bit thanks to some nice syntactic sugar:
```jinja
{# default strategy #}
{{ render(path("partial")) }}
{{ render(controller("SomeBundle:Controller:partial")) }}
{# ESI strategy #}
{{ render_esi(path("partial")) }}
{{ render_esi(controller("SomeBundle:Controller:partial")) }}
{# hinclude strategy #}
{{ render_hinclude(path("default1")) }}
```
---------------------------------------------------------------------------
by fabpot at 2013-01-03T17:58:49Z
I've just pushed a new version of the code that actually works in my browser (but I've not yet written any unit tests). I've updated the PR description accordingly.
All comments welcome!
---------------------------------------------------------------------------
by Koc at 2013-01-03T20:11:43Z
what about `render(controller="SomeBundle:Controller:partial", strategy="esi")`?
---------------------------------------------------------------------------
by stof at 2013-01-04T09:01:01Z
shouldn't we have interfaces for the UriSigner and the HttpContentRenderer ?
---------------------------------------------------------------------------
by lsmith77 at 2013-01-04T19:28:09Z
btw .. as mentioned in #6213 i think it would make sense to refactor the HttpCache to use a cache layer to allow more flexibility in where to cache the data (including clustering) and better invalidation. as such if you are refactoring HttpKernel .. it might also make sense to explore splitting off HttpCache.
---------------------------------------------------------------------------
by fabpot at 2013-01-04T19:30:07Z
@lsmith77 This is a totally different topic. This PR is just about moving things from FrameworkBundle to HttpKernel to make them more reusable outside of the full-stack framework.
---------------------------------------------------------------------------
by fabpot at 2013-01-05T09:39:52Z
I think this PR is almost ready now. I still need to update the docs and add some unit tests. Any other comments on the whole approach? The class names? The `controller` function thingy? The URI signer mechanism? The proxy protection for the internal controller? The proxy to handle internal routes?
---------------------------------------------------------------------------
by sstok at 2013-01-05T10:08:25Z
Looks good to me 👍
---------------------------------------------------------------------------
by sdboyer at 2013-01-07T18:17:08Z
@Crell asked me to weigh in, since i'm one of the Drupal folks who's likely to work most with this.
i think i've grokked about 60% of the big picture here, and i'm generally happy with what i see. the assumption that the HInclude strategy makes about working with templates probably isn't one that we'll be able to use (and so, would need to write our own), but that's not a big deal since the whole goal here is to make strategies pluggable.
so, yeah. +1.
---------------------------------------------------------------------------
by winzou at 2013-01-09T20:21:44Z
Just for my information: will this PR be merged for 2.2 version? Thanks.
---------------------------------------------------------------------------
by stof at 2013-01-09T20:41:04Z
@winzou according to the blog post announcing the beta 1 release, yes. It is explicitly listed as being one of the reason to make it a beta instead of the first RC.
---------------------------------------------------------------------------
by winzou at 2013-01-09T20:49:36Z
OK thanks, I've totally skipped this blog post.
---------------------------------------------------------------------------
by fabpot at 2013-01-10T15:26:15Z
I've just added a bunch of unit tests and fix some bugs I found while writing the tests.
