https://pulumi.com logo
Title
h

high-translator-22614

08/06/2019, 5:54 PM
funny you should say that...
Just upgraded, seeing errors from asyncio saying loop closed
m

microscopic-florist-22719

08/06/2019, 5:56 PM
yikes
That's definitely a new one
h

high-translator-22614

08/06/2019, 5:56 PM
m

microscopic-florist-22719

08/06/2019, 5:57 PM
Certainly very possible
h

high-translator-22614

08/06/2019, 5:57 PM
but all I know for sure is if I downgrade the pulumi python package from 0.17.28 to 0.17.26, it goes away
m

microscopic-florist-22719

08/06/2019, 5:58 PM
Where is the invoke in that code? Nothing jumps out at me...
Are there multiple loops involved?
h

high-translator-22614

08/06/2019, 5:58 PM
it's the inner mechanism for a lambda package builder
m

microscopic-florist-22719

08/06/2019, 5:59 PM
handle = self._ready.popleft()
    IndexError: pop from an empty deque
    error: an unhandled error occurred: Program exited with non-zero exit code: 1
This seems bad.
h

high-translator-22614

08/06/2019, 5:59 PM
yeah
unfortunately, I haven't spent enough time in the inner guts of
asyncio
to understand it
m

microscopic-florist-22719

08/06/2019, 6:00 PM
Is there any way to turn this into a straightforward local repro?
I'd love to dig into it
h

high-translator-22614

08/06/2019, 6:01 PM
🤔 no idea, but i'll try?
my first guess is long-running tasks/futures?
it might also have to do with all the machinery that goes into process handling?
m

microscopic-florist-22719

08/06/2019, 6:07 PM
Yeah, I wondered about that
h

high-translator-22614

08/06/2019, 6:15 PM
aside, is
pulumi.export("value", my_async_func())
a good test stand?
m

microscopic-florist-22719

08/06/2019, 6:24 PM
Not sure what you mean by test stand
But I would expect that to work
h

high-translator-22614

08/06/2019, 6:29 PM
like, will it invoke the same pertinent code paths as if i embedded the process into a resource?
m

microscopic-florist-22719

08/06/2019, 6:33 PM
Ah
Yes, I believe so
h

high-translator-22614

08/06/2019, 6:38 PM
ok, so far trivial repros aren't working
git bisect
puts the blame on Make
pulumi.runtime.invoke
synchronous. (#3019)
so i guess that theory is lightly confirmed
(63eb7abb5949e562b20bf9fe43bac1868e730774)
I filed https://github.com/pulumi/pulumi/issues/3038 so that there's a note where other people can find that this bug exists
m

microscopic-florist-22719

08/06/2019, 7:11 PM
Thank you!
h

high-translator-22614

08/06/2019, 7:22 PM
fun fact: i don't actually have to invoke deplumi to trigger the bug, just import it
those are the best bugs