Will this also solve doctrine/DoctrineBundle#562?

@antanas-arvasevicius looking at your comment here, I think we can close this PR, isn't it?

Hi, no this PR fixes the issue, I'll add some test cases to this PR and
hope this will be merged into release. But why do you think that this PR
should be closed?

why do you think that this PR should be closed?

Because reading your comment, you say "it's not due lazy->lazy" :)
And also because "lazy" is an optional feature that requires proxy-manager.
Or should this be done conditionally when proxy-manager is used?

Yes, that doctrine bug is not due lazy->lazy. It's due this bug which was
fixed with this PR. It's due a fact that hasReference() also scans lazy
service dependencies to find out circular dependencies, why is that so
because lazy is indirectly referenced and instantiated lazily. And that
Doctrine lazy listeners has worked only by miracle because it used
Configuration as inlined service definition and thats why hasReference
hadn't detected any circular dep.

So, what about:

Or should this be done conditionally when proxy-manager is used?

It would be reasonable to do it conditionally because if proxy-manager
doesn't exists then everything is instantiated eagerly and we need
detection of circular references in this case. If proxy-manager is
available then we need to "skip reference search in lazy service's
arguments". What do you think?

I agree :)

Anything to be done before merging this one?

On Nov 10, 2016 2:26 AM, "Fabien Potencier" notifications@github.com
wrote:

Anything to be done before merging this one?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#19902 (comment), or
mute the thread
https://github.com/notifications/unsubscribe-auth/AEAcnFDb7RZgYeYoDImXkxBprTf1OUu2ks5q8mTOgaJpZM4J5yIT
.

Hi, needed to add unit test with DI setup which shows correct behavior (no
circular reference error while injecting argument in lazy service which is
dependent of that service itself) e.g. graph: new A(new B(A)). If B is
marked as lazy we should allow this configuration.

On Nov 10, 2016 7:32 AM, "Antanas Arvasevicius" <
antanas.arvasevicius@gmail.com> wrote:

On Nov 10, 2016 2:26 AM, "Fabien Potencier" notifications@github.com
wrote:

Anything to be done before merging this one?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#19902 (comment), or
mute the thread
https://github.com/notifications/unsubscribe-auth/AEAcnFDb7RZgYeYoDImXkxBprTf1OUu2ks5q8mTOgaJpZM4J5yIT
.

if B is lazy and we can actually make it lazy (if we cannot generate the lazy proxy, it is the same than not being lazy)

Right, but what's the point to allow lazy:true without proxy mngr? Are
there any use cases for that?
Now, if you need a lazy for avoiding circular refs and proxy manager is not
there you'll get circular reference error and not an error which tells you
that you must enable proxy manager for this.

On Nov 10, 2016 12:11 PM, "Christophe Coevoet" notifications@github.com
wrote:

if B is lazy and we can actually make it lazy (if we cannot generate
the lazy proxy, it is the same than not being lazy)

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#19902 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AEAcnF7eP7HvJbyqeC04_6KkyTVKGC3hks5q8u3rgaJpZM4J5yIT
.

At least for Open-Source bundles, people might tag a service as lazy as an optimization if the end user has the proxy manager and as a noop if it's not installed.

Added unit tests, And also had found out that CheckCircularReferencesPass detects circular references. I've applied patch for that too. Please review.

@fabpot could this be merged now? Or there are any problems?

rebase needed

Rebase onto what? Haven't thought that bug fix could take so long to be applied... I really do not understand how your bug fix apply policy works, but it sucks. Bye.

A rebase on the target branch of this PR which is 2.7, since some other commits have changed some files modified in this PR, you need to resolve some conflicts so maintainers are able to merge it automatically. Thank you for contributing.

Thanks for explanation. Now rebased my branch. Is this fix will be available in 2.8 .. 3.1 also?

Yes it will ;)

👍

Status: Reviewed

👍 (the $isLazy temp var could be removed)

Thank you @antanas-arvasevicius.

Thank you @antanas-arvasevicius.
@fabpot do you know in which version this fix will be available for 7.2.* ?

@gxolin it's always the next release. So 2.7.22

no prob :)