…cations (chalasr)This PR was merged into the 2.8 branch.Discussion----------[DI] Prevent AutowirePass from triggering irrelevant deprecations| Q             | A| ------------- | ---| Branch?       | 2.8| Bug fix?      | no (just a failing test case atm)| New feature?  | no| BC breaks?    | no| Deprecations? | no| Tests pass?   | no| Fixed tickets | n/a| License       | MIT| Doc PR        | n/aFor populating available types the AutowirePass iterates over `$container->getDefinitions()` trying to instantiate a reflection class for each definition.Problem is that if one of these classes is deprecated, a notice is triggered due to the reflection, even if the service is actually never used.~~Right now this only reproduces the issue with a failing test case~~, this bug (if we agree it's a bug) breaks the test suite of a bundle I maintain (seehttps://travis-ci.org/lexik/LexikJWTAuthenticationBundle/jobs/218275650#L262)Solutions I can think about for now:- ~~Skip deprecated definitions from type registering, meaning that if a service is deprecated a day, all autowired services that rely on it will suddenly be broken, also the bug would remain if the definition is not deprecated but relies on a deprecated class, not good I think~~- Register an error handler ignoring deprecations during the type registering process (`AutowirePass#populateAvailableType()`), ~~works but makes my test suite say `THE ERROR HANDLER HAS CHANGED`. I'll push my try as a 2nd commit.~~Thoughts?Commits-------77927f4 [DI] Prevent AutowirePass from triggering irrelevant deprecations

…g reflection against all existing services (nicolas-grekas)This PR was merged into the 3.3-dev branch.Discussion----------[BC BREAK][DI] Always autowire "by id" instead of using reflection against all existing services| Q             | A| ------------- | ---| Branch?       | 3.3| Bug fix?      | no| New feature?  | yes| BC breaks?    | yes - compile time only| Deprecations? | no| Tests pass?   | yes| Fixed tickets | -| License       | MIT| Doc PR        | -(patch best reviewed [ignoring whitespaces](https://github.com/symfony/symfony/pull/22295/files?w=1).)"By-id" autowiring, as introduced in#22060 is free from all the issues that "by-type" autowiring has:- it has no magic and requires explicit type<>id matching (*vs* using reflection on all services to cherry-pick *the* one that matches some type-hint *at that time*, which is fragile)- it is free from any ambiguities (*vs* the Damocles' sword of breaking config just by enabling some unrelated bundle)- it is easily introspected: just look at DI config files (*vs* inspecting the type-hierarchy of all services + their type-hints)- ~~it is side-effect free, thus plain predictable (*vs* auto-registration of discovered types as services)~~- it plays nice with deprecated services or classes (see#22282)- *etc.*Another consideration is that any "by-type" autowired configuration is either broken (because of future ambiguities) - or equivalent to a "by-id" configuration (because resolving ambiguities *means* adding explicit type<>id mappings.) For theoreticians, we could say that "by-id" autowiring is the asymptotic limit of "by-type" autowiring :-)For all these reasons - and also because it reduces the complexity of the code base - I propose to change the behavior and only support "by-id" autowiring in 3.3. This will break some configurations relying on "by-type" autowiring. Yet the break will only happen at compile-time, which means this won't *silently* break any apps. For all core Symfony services, they will work out of the box thanks to#22098 *et al.* For the remaining services, fixing ones config should be pretty straightforward: just follow the suggestions provided by the exception messages. If they are fine to you, you'll end up with the exact same config in the end. And maybe you'll spot issues that were hidden previously.Commits-------cc5e582 [BC BREAK][DI] Always autowire "by id" instead of using reflection against all existing services