14:11 <sb0> mithro, "please sign in to view this video"

17:42 <GitHub86> [artiq] sbourdeauducq pushed 3 new commits to master: https://git.io/vgIvg

17:42 <GitHub86> artiq/master 55b3124 Sebastien Bourdeauducq: remove stale dependency on pyelftools

17:42 <GitHub86> artiq/master c639860 Sebastien Bourdeauducq: setup.py: remove stale import

17:42 <GitHub86> artiq/master b7de92e Sebastien Bourdeauducq: remove pxi6733 support (now lives at https://github.com/m-labs/aq_ni6733

19:46 <sb0> rjo, maybe you specified "--bind ::1" even though it was already bound to localhost?

19:47 <sb0> this is somewhat annoying to resolve as there can be name aliases to localhost, and the name resolving happens inside python

20:05 <rjo> probably. it was "host": "::1"

20:10 <sb0> yes, i can reproduce the problem with your ctlmgr unittest if i remove --no-bind-localhost

20:10 <sb0> fixing that in python should be a 1-line patch, hopefully that can get merged fast... mh

20:57 <GitHub153> [artiq] sbourdeauducq created newtesting (+3 new commits): https://git.io/vgI9u

20:57 <GitHub153> artiq/newtesting ca8a075 Sebastien Bourdeauducq: test/ctlmgr: use sys.executable, use default localhost binding

20:57 <GitHub153> artiq/newtesting 9ca8b48 Sebastien Bourdeauducq: assume 'import artiq' works before running tests

20:57 <GitHub153> artiq/newtesting 89f68f3 Sebastien Bourdeauducq: style

21:14 <sb0> rjo, https://github.com/python/asyncio/issues/315

21:52 <rjo> sb0: ack.

21:54 <rjo> sb0: can i expose socket.settimeout in Client? AsyncClient and constantly having to run the loop seems a bit of overkill just to get timeouts (currently in the hardware_testbench).

21:56 <sb0> rjo, you want RPC calls that time out, right?

21:57 <rjo> specifically terminate() in the controller shutdown code path of the unittest

21:59 <rjo> well it's not "constantly having to run it" but i would need to use AsyncClient and pull in asyncio just for that.

22:00 <sb0> ok, adding timeout to Client sounds fine

22:00 <sb0> just beware of bugs... does BestEffortClient use any of the Client code? i don't remember

22:00 <rjo> constructor argument or method?

22:01 <rjo> they are separate

22:01 <sb0> constructor argument, unless it gets crowded there

22:02 <rjo> ack

22:07 <rjo> it would probably be nasty if some experiment code does socket.setdefaulttimeout(0).

22:33 <GitHub108> [artiq] jordens force-pushed subprocess-termination from 34403b4 to f7df393: https://git.io/vgLqg

22:33 <GitHub108> artiq/subprocess-termination 5500611 Robert Jordens: subprocesses: unify termination logic

22:33 <GitHub108> artiq/subprocess-termination c105949 Robert Jordens: pc_rpc.Client: support socket timeouts...

22:33 <GitHub108> artiq/subprocess-termination 7636952 Robert Jordens: ctlmgr: fix import

22:43 <GitHub129> [artiq] jordens pushed 1 new commit to subprocess-termination: https://git.io/vgLY7

22:43 <GitHub129> artiq/subprocess-termination c28b938 Robert Jordens: Client: add note about timeout sideeffects