[3.14] GH-148726: Add heap_size to _gc_runtime_state

[sergey-miryanov]

• Issue: Forward-port generational GC #148726

[nascheme]

LGTM. However, given there are no public APIs to use this heap_size value, I'm wondering if this is worth the small cost to track it. Since it was in the 3.14.0, maybe its best to restore it. Does someone have a concrete use case for it? Maybe your enhanced GC stats?

[sergey-miryanov]

Yeah, we unfortunately don't have a public API for heap_size. Even my gc-stats module is only available for 3.15, because many _Py_DebugOffsets changes weren't backported.

But tools like py-spy can "reconstruct" the _gc_runtime_state layout for a specific Python version and get access to this field.

But,... now might be a good time to add 'heap_size' to the callback? :) @hugovk

[hugovk]

This PR looks okay as-is, can you update the ABI file and then summon the buildbots?

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @sergey-miryanov for commit 9a67a99 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F149025%2Fmerge

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

[sergey-miryanov]

Buildbots failures are unrelated - the same as in previous PR