[zeromq-dev] ZMQ 2.1.3 w/OpenPGM Assertion on pgm_rate_check2()

Steven McCoy steven.mccoy at miru.hk
Tue Mar 22 16:32:01 CET 2011


On 22 March 2011 22:02, Steven McCoy <steven.mccoy at miru.hk> wrote:

> On 22 March 2011 21:20, Bob Beaty <drbobbeaty at gmail.com> wrote:
>
>> Steve-O,
>>  I received this assertion this morning: https://gist.github.com/881176and it's pointing to the fact that data_size == 0, and that seems to be
>> causing the assertion failure. My question is this: How could this happen?
>>
>
> Ooh, the packets are so small it causes tpdu_length to be less than the IP
> header length causing the pre-adjustment to the length passed to pgm_rate_check2
> to fail because it takes an unsigned length.
>
>
So I'll get onto a 5.1.115 release for tomorrow with appropriate fixes that
are in trunk already.  I haven't finished working through enhancements for
nodename/NSS/DNS handling yet and wish to get them done first:  remaining is
to update pgm_getnodeaddr() to return a linked list of addresses like
getaddrinfo()  and filter the list just like DNS results.

Currently in trunk on a dual-stack host a null network parameter will
default to IPv6 including the loopback interface "::1".  By specifying the
network parameter "<*hostname*>" it will discard the loopback interface and
use a IPv4 address.  The complication occurs because the node address family
is used to detect the preferred interface, the default parameter ";" is not
equivalent to "<*hostname*>;" because this would break ";<IPv4 group
address>" if the *hostname *resolved to an IPv6 interface.

-- 
Steve-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110322/f64ea9ff/attachment.htm>


More information about the zeromq-dev mailing list