[evla-sw-discuss] CMP timesetting code review

[Ken Sowinski]

                    Walkthrough of CMP Timesetting Code
                            May 31, 2006

Author:   H. ben Frej
Chairman: B. Clark
Scribe:   K. Sowinski
Lector:   W. Sahr
Members:  excused/absent

There was a discussion of the overall organization and how the
vxWorks and Linux saystems interact.  I wondered if the approach
taken for antenna MIBs was needed in the control building for a
system connected the the synchronous SLC.

General Comments
Replacing IP numbers with hostnames would better insulate against
future changes.  Using CVS more vigourously could replace the need
for confusing #ifdef's.  It was noted that _OLD_TIME_ was undefined
at compile time for this module.  There are comments in several places,
cmpTime() is an example, which are more related to the MIB than CMP
implementation; these should be recast.


Hichem promised to explain to us why the sense of the comparison
to determine the best time in timeMsg() is the way it is.

The calculation of time done at the end of cmpTIME could be considerably
simplified.  Much of the code before that could also be simplified
knowing that the heartbeat for the SLC is always 19.2 Hz.  The code is
already sufficiently far from the MIB implementation that such a 
specialization can be justified.

setTimeStamp() has only one line that does any work and might as well
be subsumed into cmpTime() reducing the complexity from three threads
to two.


Corrections and additions are welcome.