Hi Paul,

indeed, some of these stats are not consistent. I will go through them. General idea is that the old stats framework will be replaced by a new one (present already in the code) that does not need any kind of locking when updating the counters, thus is faster.

If you have specific examples of current stats that go nuts in 64b, please report them here so I will start with them then I will go through the rest of sources.

Thanks,
Daniel

On 3/29/11 12:04 PM, Paul Pankhurst wrote:
 
I have noticed that 'kamctl monitor' displays corrupt statistics when kamailio is built for a 64 bit platform.

Further investigation has revealed that the reason is the underlying datatype is a long, whilst some of the wrapper functions (e.g. get_stat_val) return unsigned int. On a 32bit machine where a long and int are both the same size it is not a problem, but on a 64 bit machine they are different and hence the problem.

Some code uses the wrapper function, other does not, in the case of mi_get_stats it uses both, so depending on paramaters passed to the MI fifo it may work or produce garbage results.

 
Paul
_______________________________________________ sr-dev mailing list sr-dev@lists.sip-router.org http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev

-- 
Daniel-Constantin Mierla
http://www.asipto.com