[Python-Dev] Death to WITH_CYCLE_GC

[Tim Peters]

[Tim Peters]
> but since we never test without it the ability to compile it out
> isn't much of "a feature".

[Martijn Faassen]
> If the problem is you don't have time to test it,

That's not a problem for me.  In context, it was just one more reason why
keeping WITH_CYCLE_GC has become a poor idea at best.

> what about talking to the Snake Farm people of the Python Business
> Forum?
>
> http://www.lysator.liu.se/~sfarmer/
>
> They may be able and willing to help.

This would be a good idea for an "optional feature" somebody actually wants.
For example, is HAVE_UNICODE actually turned off out in the world?  If that
possibility is important to the PBF, then they should arrange to test it.
We certainly don't.  We shouldn't be the ones telling the PBF what's
important to them, either -- they need to figure that out following their
own lights.

We don't test any variations beyond debug vs release build, and it appears
that the debug build isn't tested much except on Windows (although I expect
the debugging memory allocator in 2.3 will suck more Linux developers into
running debug builds).