How to replay the di log file using the “trace” option of the CDI.

CDI can run in ‘trace mode’, which takes a ‘di’ log filefrom a

previous run as its only input, and will run readingserver messages

and Asterisk packets from the log file and will produce anew log file

essentially identical to the input log file.It is very useful when

trying to figure out why cdi did something bad, like geta corefault

or a weird error message or any other kind of error.The idea is, run

cdi again with the log file and with additional loggingturned on, or

run it from the debugger and put a breakpoint at the spotwhere things

go astray.

It can also be useful to add additional logging or othercode to the

cdi source, and then recompile and run with the logfilefrom the

problem run to get a detailed information about what wentwrong.

If the version of cdi is the same as the one whichproduced the

logfile, the new logfile will be very nearly identical tothe original

except timing in seconds might be off by a second becausethe second

rolls over to the next second at a slightly differentpoint in the

run.Aside fromthat, any differences betweenthe logfile before and

the logfile after should be examined and understood.

If the cdi version is different from the one whichproduced the

original logfile, it might still run just fine but thelogfiles might

differ.However,even if they are not the same, the logfile lines

beginning with “ZZ” (connects) or”ERROR” (errors) should match


If the cdi version is very different than the one whichmade the

original file, the trace mode run might not run to theend.

To make cdi produce different logging than the originalrun, find the

last logfile line which begins “LOGGING” andadd or remove logging

classes from it.The must obvious of this is to add “DBG” to the

logging and the trace run will reproduce the original logfile with

all the debugging logging added to it.It can also be useful to add

logging to cdi in class “NEW”, for example, andthen add “NEW” to the

logging line in the original logfile.

Running cdi in trace mode, you must have a logfile (or aset of

contiguous log files) which begins with “LOG#1”, the logfile with the

“initial start-up”.If you are running through more than onelogfile,

they must be sequential.The timestamp at the end of one will be

the same as the timestamp at the beginning of the nextone or very

rarely it will be one second earlier.

For doing tricky timing analysis, you can also addmillisecond or

microsecond resolution to the logfile produced in thetrace run by

setting the appropriate dump switches in the initialfile.Running in

trace mode, cdi should set the dump bits according to thelogging

lines “DUMPBITS” but right now it doesn’t; ituses whatever dump

switches are set in the initial file or files it picks upthe same as

any program.

To run cdi in trace mode, say :

“cdi <logfile>[<logfile>] [<logfile>] trace”

  • celeus-trace.doc