[zeromq-dev] Assertion failed: get_load () == 0 (poller_base.cpp:31)

Pieter Hintjens ph at imatix.com
Wed Oct 2 09:38:03 CEST 2013


Hi Renato,

What version of the library are you using?

The asserts in the library do not handle error conditions but
impossible situations caused by remaining bugs in the code, or unknown
states. In these situations there is no way to recover. If the process
continued, the results could be data sent wrongly, corrupted, etc. The
only sane strategy in such cases is to assert.

Removing assertions from such code makes it more fragile, more likely
to break in obscure ways, not stronger.

To your specific case, the only way to fix this is to reproduce it, so
anything you can do to help that would be good.

-Pieter


On Wed, Oct 2, 2013 at 9:22 AM, Renato Samperio
<mintaka_alnitak at hotmail.com> wrote:
> Hi,
>
> While running an extensive test sending messages from PUB/SUB pattern, we
> had an assertion failure that caused a kill of a process.
>
> The message says:
>
> Assertion failed: get_load () == 0 (poller_base.cpp:31)
>
> This comes from a PUB process and after the socket was closed. Also, this
> assertion happened before or during the context destruction.
>
> We run this code for long periods of time and we see an assertion failure
> from ZMQ, which is quite disappointing because we would expect an error back
> from the library not a kill. We think the assertions should not be in the
> library because they will kill our processes.
>
> We could provide a minimal test case (as requested in a similar issue by
> Pieter) but our automated tests were done in a very long run and most of the
> times it runs successfully. Also, we were not attending the system while it
> crashed.
>
> Does any one knows:
>  - What does it mean?
>  - What happened? In our tests we send plain strings continuously from a
> PUBlisher to  a SUBscriber.
>  - How could we avoid this problem?
>
>
> We have also documented an older issue related to Python but in our opinion
> it is not useful. Any feedback is very welcome.
>
> Regards,
>
> Renato.
>
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>



More information about the zeromq-dev mailing list