Closed wolfv closed 6 years ago
Yep, looks like the problem is solved -- thanks!
Are you planning on putting out a 0.19.1 release for this? I think this is a pretty serious issue -- my jobs were leaking tens of gigabytes per hour and couldn't run to completion, and I spent over a day narrowing down the issue, thinking the whole time I was using shared_ptr wrong or something.
Thanks for the reminder, you're right we should have made a release a week ago -- will do it now!
Release is out on conda-forge! BTW if you ever have quick questions or remarks, we also try to be very active on our gitter channel: gitter.im/QuantStack/Lobby would be cool to hear what you're doing with xtensor :)
This should fix #156 (the leak was in the copy constructor of pytensor, it's fine with std::move into the tuple).
Thanks for the great bug report @daniel-ziegler – would be cool if you could confirm that this solves your issue.