This PR was squashed before being merged into the 3.4 branch (closes#26532).
Discussion
----------
[HttpKernel] Correctly merging cache directives in HttpCache/ResponseCacheStrategy
| Q | A
| ------------- | ---
| Branch? | 3.4
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #26245, #26352, #28872
| License | MIT
| Doc PR | -
This PR is a first draft to fix the incorrect merging of private and other cache-related headers that are not meant for the shared cache but the browser (see mentioned issues).
The existing implementation of `HttpFoundation\Response` is very much tailored to the `HttpCache`, for example `isCacheable` returns `false` if the response is `private`, which is not true for a browser cache. That is why my implementation does not longer use much of the response methods. They are however still used by the `HttpCache` and we should keep them as-is. FYI, the `ResponseCacheStrategy` does **not** affect the stored data of `HttpCache` but is only applied to the result of multiple merged subrequests/ESI responses.
I did read up a lot on RFC2616 as a reference. [Section 13.4](https://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html#sec13.4) gives an overall view of when a response MAY be cached. [Section 14.9.1](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.9.1) has more insight into the `Cache-Control` directives.
Here's a summary of the relevant information I applied to the implementation:
- > Unless specifically constrained by a cache-control (section 14.9) directive, a caching system MAY always store a successful response (see section 13.8) as a cache entry, MAY return it without validation if it is fresh, and MAY return it after successful validation.
A response without cache control headers is totally fine, and it's up to the cache (shared or private) to decide what to do with it. That is why the implementation does not longer set `no-cache` if no `Cache-Control` headers are present.
- > A response received with a status code of 200, 203, 206, 300, 301 or 410 MAY be stored […] unless a cache-control directive prohibits caching.
> A response received with any other status code (e.g. status codes 302 and 307) MUST NOT be returned […] unless there are cache-control directives or another header(s) that explicitly allow it.
This is what `ResponseCacheStrategy::isUncacheable` implements to decide whether a response is not cacheable at all. It differs from `Response::isCacheable` which only returns true if there are actual `Cache-Control` headers.
- > [Section 13.2.3](https://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html#sec13.2.3): When a response is generated from a cache entry, the cache MUST include a single Age header field in the response with a value equal to the cache entry's current_age.
That's why the implementation **always** adds the `Age` header. It takes the oldest age of any of the responses as common denominator for the content.
- > [Section 14.9.3](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.9.3): If a response includes an s-maxage directive, then for a shared cache (but not for a private cache), the maximum age specified by this directive overrides the maximum age specified by either the max-age directive or the Expires header.
This effectively means that `max-age`, `s-maxage` and `Expires` must all be kept on the response. My implementation assumes that we can only do that if they exist in **all** of the responses, and then takes the lowest value of any of them. Be aware the implementation might look confusing at first. Due to the fact that the `Age` header might come from another subresponse than the lowest expiration value, the values are stored relative to the current response date and then re-calculated based on the age header.
The Symfony implementation did not and still does not implement the full RFC. As an example, some of the `Cache-Control` headers (like `private` and `no-cache`) MAY actually have a string value, but the implementation only supports boolean. Also, [Custom `Cache-Control` headers](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.9.6) are currently not merged into the final response.
**ToDo/Questions:**
1. [Section 13.5.2](https://www.w3.org/Protocols/rfc2616/rfc2616-sec13.html#sec13.5.2) specifies that we must add a [`Warning 214 Transformation applied`](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.46) if we modify the response headers.
2. Should we add an `Expires` headers based on `max-age` if none is explicitly set in the responses? This would essentially provide the same information as `max-age` but with support for HTTP/1.0 proxies/clients.
3. I'm not sure about the implemented handling of the `private` directive. The directive is currently only added to the final response if it is present in all of the subresponses. This can effectively result in no cache-control directive, which does not tell a shared cache that the response must not be cached. However, adding a `private` might also tell a browser to actually cache it, even though non of the other responses asked for that.
4. > [Section 14.9.2](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.9.2): The purpose of the `no-store` directive is to prevent the inadvertent release or retention of sensitive information […]. The `no-store` directive applies to the entire message, and MAY be sent either in a response or in a request. If sent in a request, a cache MUST NOT store any part of either this request or any response to it. If sent in a response, a cache MUST NOT store any part of either this response or the request that elicited it.
I have not (yet) validated whether the `HttpCache` implementation respects any of this.
5. As far as I understand, the current implementation of [`ResponseHeaderBag::computeCacheControlValue`](https://github.com/symfony/symfony/blob/master/src/Symfony/Component/HttpFoundation/ResponseHeaderBag.php#L313) is incorrect. `no-cache` means a response [must not be cached by a shared or private cache](https://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.9.1), which overrides `private` automatically.
5. The unit tests are still very limited and I want to add plenty more to test and sort-of describe the implementation or assumptions on the RFC.
/cc @nicolas-grekas
#SymfonyConHackday2018
Commits
-------
893118f978 [HttpKernel] Correctly merging cache directives in HttpCache/ResponseCacheStrategy
This PR was merged into the 3.4 branch.
Discussion
----------
[HttpKernel] Correctly Render Signed URIs Containing Fragments
| Q | A
| ------------- | ---
| Branch? | `3.4`
| Bug fix? | yes
| New feature? | no
| BC breaks? | no?
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
- Rebuild the URL with the computed hash instead of appending it onto the end of the fragment.
- Update unit tests, and add new unit test to cover URIs that include fragments.
Commits
-------
b9ece6bde7 [HttpKernel] Correctly Render Signed URIs Containing Fragments
Rebuild the URL with the computed hash instead of appending it onto the end of the URI, preventing incorrect formatting when dealing with URIs containing fragments.
Since https://github.com/symfony/symfony/pull/25733 the Kernel attempts to unlink the legacy container while being built.
This throws an error if the file did not exist, for example on a clean install, on the build, which is then silenced.
That's fine on production systems, but on our build we have enabled "xdebug.scream" in order to visualise every errors, which basically un-silences the errors. I believe there should not be a need to silence anything on a usual, clean usage of the system.
Making this `unlink` conditional fixes it.
Could you please approve and merge this PR?
Thanks
Some attributes being used in the phpunit configuration files, namely
failOnRisky and failOnWarning were introduced in phpunit 5.2.0. The
Composer configuration shows that tests should run with old versions of
phpunit, but phpunit only validates the configuration against the XSD
since phpunit 7.2.0.
These changes can be tested as follows:
wget http://schema.phpunit.de/5.2/phpunit.xsd
xargs xmllint --schema phpunit.xsd 1>/dev/null
find src -name phpunit.xml.dist| xargs xmllint --schema phpunit.xsd 1>/dev/null
See 7e06a82806
See 46e3745a03/composer.json (L98)
This PR was merged into the 3.4 branch.
Discussion
----------
[VarDumper] fix dump of closures created from callables
| 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 | -
We are missing displaying full information about closures created using `ReflectionMethod::getClosure()` or `Closure::fromCallable()`.
This PR fixes it. For VarDumper but also other places where we have logic to display them.
Commits
-------
1c1818b876 [VarDumper] fix dump of closures created from callables
