GH-148726: Add heap_size to generational GC

[sergey-miryanov]

• Issue: Forward-port generational GC #148726

[nascheme]

LGTM.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @hugovk for commit 2281499 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F149195%2Fmerge

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

[sergey-miryanov]

Good catch by @DinoV in #149241. I believe we should cherry-pick those commits here.

[hugovk]

Good catch by @DinoV in #149241. I believe we should cherry-pick those commits here.

Good idea.

[sergey-miryanov]

@zanieb @nascheme Could you please take another look? I've ported the 3.14 heap_size changes and included heap_size in get_gc_stats.

@DinoV FYI, the layout of _gc_runtime_state will change since we moved generation_stats to the heap.

[nascheme]

I'd say for 3.15 we don't need the two dummy members and we could move generation0 back to it's old position. Preserving the member offsets should not be required for 3.15, right?

[sergey-miryanov]

Since we're moving the stats to the heap, all subsequent offsets will shift anyway. There's no need to keep the dummy members.

I thought this would simplify backports if needed.

I'm happy to remove the dummy members and restore generation0 if you think that's cleaner.

[nascheme]

I'm happy to remove the dummy members and restore generation0 if you think that's cleaner.

I think it's better to remove the dummy members. You don't have to restore generation0 since the code looks fine without that member. Other than that, this PR looks fine to me.

[sergey-miryanov]

@nascheme Thanks! Done.

[pablogsal]

Great job @sergey-miryanov

[sergey-miryanov]

Thanks all!