[5.6] __set_state exists in Carbon

[kylekatarnls]

__set_state should not be overridden

[svenluijten]

If it shouldn't be overridden it should've been final or private to begin with. And a change in public API would normally warrant a major version bump (2.0.0) according to semver.

[olafnorge]

May we just fix it and discuss later?

[svenluijten]

@olafnorge sure, once the tests pass again.

[kylekatarnls]

Yes quick-revert here: #23466

As lowest setup cannot pass without __set_state(), the __set_state() removing must be done after with also a Carbon minimal version bumping.

[kylekatarnls]

With Carbon version update the tests pass and so __set_state could be simply removed. So this PR is a better resolution in my opinion.

[torrancemiller]

Need this now. This brought down my dev environment. :(

[antonioribeiro]

@kylekatarnls is the usage of __set_state being tested?

[kylekatarnls]

__set_state in Carbon is exactly the same code as in this override and yes we tested it here: https://github.com/briannesbitt/Carbon/blob/master/tests/Carbon/InstanceTest.php#L61

[LukeTowers]

@kylekatarnls is that PR going to be backported to 5.5 LTS?

[kylekatarnls]

@LukeTowers > @taylorotwell did it.

What is the workflow for master/5.7, should it be this commit merged or should I submit an other PR?

[LukeTowers]

Awesome, thanks for the update!

[Kyslik]

This might have break something; see this SO question.

[GrahamCampbell]

Always submit PRs to the earliest branch on the framework. We regularly merge 5.5 -> 5.6 -> master. :)

[kylekatarnls]

@Kyslik 32bit problem, fixed in 1.24.2