This PR was merged into the 3.0-dev branch.
Discussion
----------
changed deps from 2.7 to 2.8 in all composer.json
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Commits
-------
4dedb97 bumped deps on 2.8 in all composer.json
This PR was merged into the 2.7 branch.
Discussion
----------
removed 3.0 constraints from 2.7 composer files
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | n/a
| License | MIT
| Doc PR | n/a
Those 3.0 constraints will be kept in the new 2.8.
Commits
-------
5503804 removed 3.0 constraints from 2.7 composer files
* 2.8:
[Translation] avoid freshness check based on content *inside* the cache.
[Translator] Cache does not take fallback locales into consideration
[Translator] Cache does not take fallback locales into consideration
[VarDumper] Fix call site detection
[Form] Cleanup deprecation notices
[Console] Made output docopt compatible
[Process] Fix volatile test
Remove some useless @group annotations
Removed useless strtolower call
[Validator] Use strict comparisons in loaders
CS: Use "self" keyword instead of class name if possible
Conflicts:
.travis.yml
src/Symfony/Component/Console/Tests/Input/InputDefinitionTest.php
src/Symfony/Component/Form/FormTypeExtensionInterface.php
src/Symfony/Component/Form/FormTypeInterface.php
* 2.6:
[Translator] Cache does not take fallback locales into consideration
[VarDumper] Fix call site detection
[Process] Fix volatile test
Remove some useless @group annotations
Removed useless strtolower call
[Validator] Use strict comparisons in loaders
CS: Use "self" keyword instead of class name if possible
Conflicts:
.travis.yml
src/Symfony/Component/Translation/Translator.php
This PR was merged into the 2.6 branch.
Discussion
----------
[VarDumper] Fix call site detection
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
`$trace[$i]['file']` is not always defined and `call_user_func_array` should also be filtered out.
Commits
-------
d8fb6a4 [VarDumper] Fix call site detection
This PR was merged into the 2.7 branch.
Discussion
----------
[Form] Cleanup deprecation notices
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
This uses the same technique used for FlattenException in HttpKernel that prevents repeated notices triggered by the constructor.
Commits
-------
3f58862 [Form] Cleanup deprecation notices
This PR was squashed before being merged into the 2.7 branch (closes#13220).
Discussion
----------
[Console] Made output docopt compatible
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | yes
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #6329
| License | MIT
| Doc PR | https://github.com/symfony/symfony-docs/issues/5016
This was harder than I thought. To sum up:
* The output now follows the [docopt](http://docopt.org/) specification
* There is a new `addUsage` method to add more usage patterns
* The handling of spaces in the descriptors is refactored to make it easier to understand and to make it render better (using sprintf's features only made it worse imo)
Todo
---
* [x] Add test for `addUsage` and friends
* [x] Add test for multiline descriptions of arguments
* <s>Convert long descriptions to multiline automatically</s>
* [ ] Submit a doc PR for `addUsage`
Question
---
The docopt specification suggests we should add these usage patterns:
%command.name% -h | --help
%command.name% --version
I didn't do that yet, as I think it'll only makes the output more verbose and it's already pretty obvious.
I've taken some decisions which I don't think everybody agrees with. I'm willing to change it, so feel free to comment :)
/cc @Seldaek
Commits
-------
3910940 [Console] Made output docopt compatible
* 2.3:
Remove some useless @group annotations
Removed useless strtolower call
[Validator] Use strict comparisons in loaders
CS: Use "self" keyword instead of class name if possible
* 2.7:
[Config] Delegate creation of ConfigCache instances to a factory.
[Form] Fixed DateType/TimeType that were broken since 849fb29984 and 1c4a75a00a
[Console] Fixed SymfonyStyle when get terminal width fails
[SecurityBundle] UserPasswordEncoderCommand: Improve & simplify the command usage
Fixed the line length of the new Symfony Styles
[Debug] Skip unsilencing test on PHP7
[Profiler][Logger] fixed cycle odd/even.
Update logger.html.twig
Conflicts:
src/Symfony/Bundle/SecurityBundle/composer.json
src/Symfony/Component/Config/CHANGELOG.md
src/Symfony/Component/HttpKernel/composer.json
src/Symfony/Component/Routing/composer.json
src/Symfony/Component/Translation/composer.json
* 2.6:
[Form] Fixed DateType/TimeType that were broken since 849fb29984 and 1c4a75a00a
[Debug] Skip unsilencing test on PHP7
[Profiler][Logger] fixed cycle odd/even.
This PR was merged into the 2.7 branch.
Discussion
----------
[Console] Fixed SymfonyStyle when get terminal width fails
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | -
| License | MIT
| Doc PR | -
Commits
-------
a52b658 [Console] Fixed SymfonyStyle when get terminal width fails
This PR was squashed before being merged into the 2.7 branch (closes#14178).
Discussion
----------
[Config] Delegate creation of ConfigCache instances to a factory.
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | yes (refactoring, new flex point)
| BC breaks? | no
| Deprecations? | yes
| Tests pass? | we'll see :-)
| Fixed tickets | n/a
| License | MIT
| Doc PR | symfony/symfony-docs#5136
In the Routing/Router and Translation/Translator, delegate creation of ConfigCache instances to a factory. The factory can be setter-injected but will default to a BC implementation.
The ```ConfigCacheFactoryInterface``` is designed in a way that captures the common ```$cache = new ...; if (!$cache->isFresh()) { ... do sth }``` pattern. But more importantly, this design allows factory implementations to take additional measures to avoid race conditions before actually filling the cache.
By using an exchangeable ConfigCache factory it becomes possible to implement different resource (freshness) checking strategies, especially service-based ones.
The goal is to be able to validate Translators and Routers generated by database-based loaders. It might also help with symfony/AsseticBundle#168. This PR only contains the minimum changes needed, so the rest could be implemented in a bundle outside the core (at least for the beginning).
Component/HttpKernel/Kernel::initializeContainer still uses the ConfigCache implementation directly as there is no sensible way of getting/injecting a factory service (chicken-egg).
This is a pick off #7230. It replaces #7781 which was against the master branch. Also see #7781 for additional comments/explanations.
## Todo
* [ ] Allow `symfony/config` `~3.0.0` in `composer.json` for the HttpKernel and Translator component as well as TwigBundle once this PR has been merged into the master branch (fail deps=high tests for the time being).
Commits
-------
6fbe9b1 [Config] Delegate creation of ConfigCache instances to a factory.
This PR was merged into the 2.6 branch.
Discussion
----------
[Debug] Skip unsilencing test on PHP7
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #14227
| License | MIT
| Doc PR | -
Commits
-------
6a87ad3 [Debug] Skip unsilencing test on PHP7
This PR was merged into the 3.0-dev branch.
Discussion
----------
[3.0][Console][OutputStyle] Implements verbosity levels methods
| Q | A
| ------------- | ---
| Bug fix? | yes
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | N/A - Travis build for #14032
| License | MIT
Implements the verbosity level methods from `OutputInterface` for the new `OutputStyle` abstract class in 3.0 branch.
Fixes Travis builds with `deps=high` env var for #14032 PR and future console style updates.
=========
Related to: #14032, #14057, #14138, #14132, #14172, ...
Commits
-------
a30f507 [Console] [OutputStyle] Implements verbosity methods
This PR was squashed before being merged into the 2.7 branch (closes#14032).
Discussion
----------
[SecurityBundle] UserPasswordEncoderCommand: Improve & simplify the command usage
| Q | A
| ------------- | ---
| Bug fix? | no
| New feature? | no
| BC breaks? | no
| Deprecations? | no
| Tests pass? | yes
| Fixed tickets | #13988
| License | MIT
Overlaps #14017 (might replace or follow it) . Sorry if it is considered as a duplicate, but the debate has evolved, and I think the arguments ordering isn't the best nor single way to improve this command usage anymore.
Thank you @saro0h for having considered the mentioned issues and spent time on it.
# Salt option & salt generation
Thanks to @ircmaxell & @inanimatt, we came to the conclusion that the use-cases for the `salt` option/argument are pretty arguable. So I suggest to get rid of it: **a salt will always be generated by the command.**
### Generated salt
The generated salt is now in the Table output:
`security:encode-password test -n`
The "Generated salt" row and the last comment about the salt aren't present if the new `empty-salt` option is provided (see below).
# New empty-salt option
As some encoders might generate their own built-in salts (like the `BCryptPasswordEncoder`) and some custom encoders could do the same (or not require a salt at all), I suggest a new option: `empty-salt`.
This option will not provide any salt to the configured encoder, which will generate its own, then.
With the interactive way, the user will always be asked confirmation for the salt generation if the `empty-salt` option isn't set:
`security:encode-password password`
### bcrypt encoder
As the `BCryptPasswordEncoder` is shipped with the security component, and listening to @inanimatt valuable comments in #13988, I introduced a second commit (0cdb546) making an exception for the `bcrypt` encoder, and always set the `empty-salt` option with it.
We're aware that's not ideal from a OO design perspective, but far better from a DX one. If not desired, I will revert it.
Anyway I think https://github.com/symfony/symfony/issues/13988#issuecomment-85068382 and other comments about the `PasswordEncoder` API and salt generation should be considered for 3.0.
A note is added when bcrypt encoder is detected without the `empty-salt` option:
`security:encode-password password "Custom\Class\Bcrypt\User" -n`
# Default user-class
The default `user-class` is set to `Symfony\Component\Security\Core\User\User`.
I think this makes sense, as in the previous version of the command, the Symfony\Component\Security\Core\User\User was configured in the setAutocompleterValues and set the $value to Symfony\Component\Security\Core\User\User if null.
Asking the question to the user with the interactive command would have been useful only if the user was able to pick one encoder from a choice list.
# Arguments order / make arguments options
When we had both `salt` and `user-class` as arguments, the command looked like:
```sh
security:encode-password [password] [user-class] [salt]
```
All arguments were optional, for the interactive command purpose (The command asked the user for missing arguments).
But, as they were arguments, we had to provide each of them in the proper order. That meant we couldn't provide a salt without defining the optional user-class.
So I suggested using options instead of arguments for both user-class & salt.
But as the `salt` option/argument is removed, now I don't feel the need for the `user-class` to be an option.
Indeed, the new command short version will look like the following:
```sh
#Default user-class: Symfony\Component\Security\Core\User\User
security:encode-password password
#Another user-class:
security:encode-password password "AppBundle\Model\User"
```
Making the user-class an option IMO isn't a necessity anymore, and will only lengthen the command:
```sh
security:encode-password password --user-class="AppBundle\Model\User"
```
## Bonus:
- [The new command documentation](https://cloud.githubusercontent.com/assets/2211145/6845201/48a66382-d3b2-11e4-8227-b799215a2783.PNG). Thanks to @javiereguiluz.
- [Full interactivity output](https://cloud.githubusercontent.com/assets/2211145/6906381/d7753ce4-d72e-11e4-8547-2ef35c6257e9.PNG)
Commits
-------
b3f6340 [SecurityBundle] UserPasswordEncoderCommand: Improve & simplify the command usage
Harmonize TranslationDebugCommand and TranslationUpdateCommand to
expect an optional bundle name or a directory and fall back to kernel
root dir if none of them is given.
