This might also fix#22448.

Not to completely nerd-snipe this issue, but I had wondered some time ago (basically when thinking about this gc.collect call, although I certainly didn't go through all the analysis you did!) whether CPython could have an API likegc.try_collect_at_frame_exit(obj) which says, at the end of this frame, after all decrefs for frame locals have been done, check whetherobj could be gc'ed (i.e. whether it is only kept alive by an isolated refcycle, and if so, gc that cycle).

whether CPython could have an API like...

I'm also not sure. I do not think there is currently enough information to do that but am not clear on how much more you would have to track to get there. My instinct is that this would add more overhead in cases where it is not used than you would ever get back in the cases when it is used.

I think a Python exposed function for "run the full collection, but only if you would have otherwise" is what we really want here.

The other thing that just connected in my brain is that dictionaries (which I assume includes instance dictionaries) are only swept in the oldest generation (which is why our objects all survived thegc.collect(1).

I had a thought about being aggressive about callingfig.cla() when we destroy the manager, but I think that would break the workflow of

fig,ax=plt.subplots()ax.plot(...)plt.show(block=True)fig.savefig(...)

Similarly, any way I can think of putting in weak references is going to break someone terribly.

Perhaps we could callcla() (and recursively detach artists from the parent figure) whenplt.close(fig) is called, but let pressing-on-the-X-to-close just unregister the manager and not bother to callcla()? I think a user that programatically closes the figure is more likely to be running a loop where the memleak may matter, whereas interactive users are more likely to be OK with a delayed GC?

That is reasonable. It will be a bunch of work, but something like

• 3.7
  • add a "and recursively destroy" flag toplt.close
  • inplt.closeif the flag is not set by the user mark theFigure with afig._zombie flag
  • wrap all of the figure methods to warn if the zombie flag is set
• 3.8 flip the flag to default to True and remove the zombie warnings

might be a deprecation path to that?

That said, it leaves a weird inconsistency between doingplt.close() and hitting the 'x' if you have doneplt.ion() and are working interactively.

Doesn't the inline backend auto-close figures? I'm pretty sure many people useplt.savefig after that, so we can't really break that.