Norbert Sendetzky wrote: > > On Monday 10 March 2008 23:22:38 Mathew Hennessy wrote: >> OK, I've compiled powerdns on a 32-bit Linux system, and it appears to do >> notifies correctly (from Linux master running pdns to Solaris slave >> running >> bind). This has got to be some linuxisms somewhere that are not portable >> across to solaris... > > I guess you hit some big endian vs. little endian problem. The big > question is > now where is the buggy code located and what is the right representation > of > the ID. >
I see that in dns.hh endianness is detected by arch, and __sun or __sparc (or some variant) should be detected to configure the dnsheader struct appropriately. Also, the id (query identification number) is the first word regardless of endian in that file. Unless that's incorrect, my suspicions are more along the lines of miscast references to the message id (which should be uint16_t everywhere AFAIK?) or bad alignment assumptions. -- View this message in context: http://www.nabble.com/pdns_control-notify-to-bind-9.4-slave%3A-%22failed-while-receiving-responses%3A-unexpected-message-id%22-tp15859705p15978134.html Sent from the PowerDNS mailing list archive at Nabble.com. _______________________________________________ Pdns-users mailing list [email protected] http://mailman.powerdns.com/mailman/listinfo/pdns-users
