Yo
Hmmm ... philisophical difference there ... python's stack traces are
pretty good, so I don't check things like string vs. int. I consider it
a plus for python that I could essentially determine the root cause of
the error just by seeing a stack trace in an email.
Too bad more languages aren't like that. VAX FORTRAN used to be the
same way.
Last year I gave a guest lecture at the University of Amsterdam and
lcg-info-dynamic-scheduler was used as an example of "deployment
realities" ... at that point, 18% of the nontrivial lines in the program
were dealing with either error checking or logging. It's probably not
enough ...
JT
Stephen Childs wrote:
> John R. Hover wrote:
>
>> 2) /opt/lcg/var/gip/plugin/lcg-info-dynamic-scheduler-wrapper (which is
>> just a wrapper performing '/opt/lcg/libexec/lcg-info-dynamic-scheduler
>> -c /opt/lcg/etc/lcg-info-dynamic-scheduler-lcgcondor.conf' ) *sometimes*
>> works ( giving real values for, e.g. GlueCEStateFreeJobSlots). But
>> other times it dumps a Python stack trace:
> As your previous post showed, there is very little error-checking in
> that first bash at lrmsinfo-condor. It's probably outputting
> incompatible data on error conditions. I will have a second look at the
> script once I get finished with some other stuff I'm doing. However, it
> also indicates that Jeff's script could do a little more error-checking
> to protect against garbage thrown its way by incompetents like me!
>
> Stephen
|