This one would be very hard -- it's an .so hanging off a custom web server.
However, it has the same flavor as the earlier null packet problem. My general protocol is: X ::= 'A'..'Z' data ::= byte* marker_msg ::= [[ X ]] arg_msg ::= [[ X data ]] data_msg ::= [[ data ]] Ø ::= [[ ]] msg_list ::= marker_msg data_msg* Ø transmission := (arg_msg | msg_list )* so a typical message might be: [['R']] [[ data ]] [[ data ]] Ø This transaction can end with a zero length (empty) packet with the more bit cleared. It also turns out that empty data packets were being passed (and are now valid), so I've been getting all sorts of Øs interspersed with "live" packets. (The protocol will be repaired of course :-) ) Anyway, the empty packets interspersed with regular packets seem to trigger this bug. I will try to reproduce this in a more controlled environment, but it's about tenth on the list :-) Thank you, Best, Matt On Jul 21, 2010, at 3:03 AM, Martin Sustrik wrote: > Matt, > >> I've started running into this error over the last day or so >> (actually, it was concealed by an overly complex environment). >> >> I'm using the git://gitbhub.com/zeromq/zeromq2.git for my source. >> >> My recollection is that this was fixed? I'm using zero length >> packets to delimit record lists... > > It looks like the reply received by REQ socket is ill-formed for some > reason. > > Can you provide the code to reproduce the problem? > > Martin > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
