Message326807
Wow. Thanks so much for figuring this out, I know how hard it is to debug issues like this.
Now I see it clearly: _asyncgen_finalizer_hook should be using loop.call_soon_threadsafe. Interestingly, I used _write_to_self there, so I knew about the issue, but figured that using call_soon() + _write_to_self is safe enough; evidently I was wrong.
Anyways, here's the diff that fixed it for me: https://gist.github.com/1st1/c1c9fc853cac1fadb7102ccc6201fb70
Could you please create a PR (ideally with a unittest)? |
|
Date |
User |
Action |
Args |
2018-10-01 15:43:53 | yselivanov | set | recipients:
+ yselivanov, asvetlov, twisteroid ambassador |
2018-10-01 15:43:53 | yselivanov | set | messageid: <1538408633.28.0.545547206417.issue34769@psf.upfronthosting.co.za> |
2018-10-01 15:43:53 | yselivanov | link | issue34769 messages |
2018-10-01 15:43:53 | yselivanov | create | |
|