Traceback (most recent call last):
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\threading.py", line 917, in _bootstrap_inner
    self.run()
  File "D:\Proj\antwork\code\antwork_backend\server\data_handlers\order\playback_hd.py", line 147, in run
    send_to_ws(channel=self.room_name, data=json.loads(msg))
  File "D:\Proj\antwork\code\antwork_backend\server\data_exchange.py", line 26, in send_to_ws
    asyncio.run(channel_layer.group_send(channel, send_data))
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\asyncio\runners.py", line 46, in run
    _cancel_all_tasks(loop)
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\asyncio\runners.py", line 54, in _cancel_all_tasks
    to_cancel = tasks.all_tasks(loop)
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\asyncio\tasks.py", line 38, in all_tasks
    return {t for t in _all_tasks
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\asyncio\tasks.py", line 38, in <setcomp>
    return {t for t in _all_tasks
  File "C:\Users\max\AppData\Local\Programs\Python\Python37\lib\_weakrefset.py", line 61, in __iter__
    for itemref in self.data:
RuntimeError: Set changed size during iteration


https://github.com/python/cpython/commit/416c1ebd9896b394790dcb4f9f035b1a44ebe9ff#commitcomment-30887909

The issue is hard to reproduce, the problem happens on an implicit garbage collector run during iteration over `_all_tasks` weakset.

Ned, would be nice to have this trivial bugfix in 3.7.1.

The problem happens if wakes up on weakset iteration.
The chance is very low, it leads to a flaky scary bug.

3.7.1rc2 is about to be released so to add this would require retagging and remanufacturing release bits.  If the chances of it happening are "very low", I would prefer to not further delay rc2.  If you and Yury agree that the risk is low and the benefit is high, we could consider risking cherry-picking it to 3.7.1 final.  Sound OK?

Let's wait for Yuri opinion.

>> we could consider risking cherry-picking it to 3.7.1 final.  Sound OK?
> Let's wait for Yuri opinion.

I agree it's a pretty serious bug; basically a time bomb that can crash a perfectly fine asyncio application.  The PR itself seems to be safe to cherry-pick to 3.7.1 even after rc2.  I suggest to merge it asap though to have some time for buildbots to test it.

New changeset 97cf0828727ac2a269c89c5aa09570a69a22c83c by Miss Islington (bot) (Andrew Svetlov) in branch 'master':
bpo-34970: Protect tasks weak set manipulation in asyncio.all_tasks() (GH-9837)
https://github.com/python/cpython/commit/97cf0828727ac2a269c89c5aa09570a69a22c83c

New changeset 5dbb1b7df1d9ecaa6b5344028cd1777502cf5c73 by Andrew Svetlov (Miss Islington (bot)) in branch '3.7':
bpo-34970: Protect tasks weak set manipulation in asyncio.all_tasks() (GH-9837) (GH-9849)
https://github.com/python/cpython/commit/5dbb1b7df1d9ecaa6b5344028cd1777502cf5c73

Ned, the fix has landed on both master and 3.7

Your turn, please.

Thanks, Andrew and Yury!  Just leave the issue open as a "release blocker" for now and I will take care of it for 3.7.1 final.

Cool!
Thanks, Ned

New changeset 60c663c0f76c790afbed4d673c3ce264dd226b8c by Ned Deily (Miss Islington (bot)) in branch '3.7':
bpo-34970: Protect tasks weak set manipulation in asyncio.all_tasks() (GH-9837) (GH-9849)
https://github.com/python/cpython/commit/60c663c0f76c790afbed4d673c3ce264dd226b8c

Released in 3.7.1