-
Notifications
You must be signed in to change notification settings - Fork 106
ci: test Python 3.6 on AppVeyor #235
Conversation
Hey there! The failure on Travis seems like a Travis issue, but the AppVeyor one could somehow be legit. |
Oh yes, there is really a regression in 3.6+ (see Travis). Any idea? |
Please see: https://bugs.python.org/issue26685 I guess we should just remove the |
Not sure about the UDP test case though. There is a different behavior on Linux and Windows. Is pyuv or libuv closing the underlying socket on bind error on Linux? |
Oh yes. libuv closes the underlying socket on bind error on Linux (https://github.com/libuv/libuv/blob/d898a1f6f87ba30e7803fe402121464553b2ffaf/src/unix/udp.c#L343) but not on Windows (https://github.com/libuv/libuv/blob/87df1448a48fb64c2b9ebe37e3344f7e5b81dd88/src/win/udp.c#L261). |
Reported upstream: libuv/libuv#1336 |
Thanks for digging though it! I think the current workaround looks ok! |
Are you planning a new release any time soon? If not we could create a temporary branch based on the pyuv-1.3.0 tag with these changes so AppVeyor is building the Python 3.6 wheels for 1.3.0. :) |
@saghul Ping :) |
@saghul Should I squash this in one commit? |
New proposal (see commits) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, thanks!
BTW, setuptools has broken the Travis builds :/ |
No description provided.