Open megahall opened 10 years ago
I tried it on my OS X, perl 5.18.2 (no thread), zeromq-3.2.4, ZMQ-LibZMQ3 HEAD, and got "should be able to die at this point but it gets stuck";
Sent from my mobile device.
On March 12, 2014 2:46:58 PM PDT, lestrrat notifications@github.com wrote:
I tried it on my OS X, zeromq-3.2.4, ZMQ-LibZMQ3 HEAD, and got "should be able to die at this point but it gets stuck";
Reply to this email directly or view it on GitHub: https://github.com/lestrrat/p5-ZMQ/issues/50#issuecomment-37472904
I'm confused. It died as the message implied. it exited.
Sent from my mobile device.
On March 12, 2014 3:26:37 PM PDT, lestrrat notifications@github.com wrote:
I'm confused. It died as the message implied. it exited.
Reply to this email directly or view it on GitHub: https://github.com/lestrrat/p5-ZMQ/issues/50#issuecomment-37476961
We had some similar symptoms which were cause by indefinite linger settings. Maybe that's part of the problem? (I didn't run or check your code, so this is just a shot in the dark)
Something gets broken inside of
perl
, when certain functions fromZMQ::LibZMQ3
are added into the code. It does not matter if the functions ever actually execute. This test case should get stuck forever instead of finishing thedie
process on Mac OS X. I have seen it happen sometimes on Linux as well but not as reliably.I am guessing, referencing certain functions must cause a ZMQ event loop to be started, but there is no code to ensure the loop is ended and doesn't become infinite when
perl
tries to shut down. When tracing the code withdtruss
, etc. you can see it gets stuck in some kind ofpoll(...)
.When a custom
__DIE__
handler is registered, it seems possible to work around the issue.