Add ClassInfo.initializer to object.di

[nemanja-boric-sociomantic]

This method was added to object_.d and it should be present
in the accompanying di file.

[leandro-lucarella-sociomantic]

Is there any easy way to reproduce the problem and add an unittest for it to avoid some inconsistency like this to happen again?

[nemanja-boric-sociomantic]

It's super easy to test if you have a binary that was built against this di file but linked against the library with the different object.d. I'm not sure how to pull that off with the unittest (I don't think it's possible).

I could add turtle test and make makd link in the built library, but I think that would be overkill, as it would still protect you just from this particular field discrepancy. I think the safer way is just to put libtangort in support-only mode for the rest of the transition time.

[leandro-lucarella-sociomantic]

Let's merge as is then, tangort will hopefully die soon enough anyway!

[mathias-baumann-sociomantic]

Well, maybe not as unittest, but it should be simple enough with some makefile additions?

[nemanja-boric-sociomantic]

Maybe. We had to push this ASAP when we released v1.9.0 (it was causing a segfault in all turtle tests). Still, I think it's better to stop putting new features here (we had this one in, what, two years?)