That looks fine to me. Could you also contribute a test (since you seem to have reproduced the issue)?
CC'ing Richard, but I can't imagine he'd object. :-) On Thu, Jan 9, 2014 at 8:23 AM, Victor Stinner <[email protected]> wrote: > (Oops, I forgot to specify an object to my email. Sorry. Gmail didn't warn > me.) > > The proactor event loop has a similar issue with a pending overlapped > operation/future. The "self reading future" used by > _loop_self_reading() is not cancelled in close(). Because of that, > unit tests log many warnings. > > I fixed this issue in Trollius in changeset: > > "BaseProactorEventLoop: close() cancels the reading future" > https://bitbucket.org/haypo/trollius/commits/a3d3439e91130d7f1d89e24fd6b58d35a0e145c9 > > If you think that the change looks correct, I will open an issue in > Tulip with a patch. > > Victor -- --Guido van Rossum (python.org/~guido)
