Author vstinner
Recipients asvetlov, lukasz.langa, vstinner, yselivanov
Date 2019-06-04.11:50:37
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1559649037.91.0.578697527832.issue34037@roundup.psfhosted.org>
In-reply-to
Content
Łukasz Langa reverted my change (PR 13786). I pushed my change without any review to try to fix https://bugs.python.org/issue37137 which blocked the Python 3.8 beta1. Sadly, my change didn't fix this issue. The root cause was differently (and it's now fixed).

Andrew Svetlov asked to make some changes on my API. Yury Selivanov added "I don't like this last minute api decision. Please don't merge this.". See PR 13786 for the discussion. In short, my change was reverted because of the API.

But I still consider that calling executor.shutdown(wait=False) in loop.close() in wrong: we must wait until it finishes. Otherwise, the code "leaks" threads which continue to run and so can cause troubles. I would prefer to wait by default.

Andrew, Yury: Would you be interested to rewrite my change in the "asyncio way"?

IMHO this change can wait Python 3.9, it's not a critical bug.
History
Date User Action Args
2019-06-04 11:50:37vstinnersetrecipients: + vstinner, asvetlov, lukasz.langa, yselivanov
2019-06-04 11:50:37vstinnersetmessageid: <1559649037.91.0.578697527832.issue34037@roundup.psfhosted.org>
2019-06-04 11:50:37vstinnerlinkissue34037 messages
2019-06-04 11:50:37vstinnercreate