I think that the CVS sensor runs 'outside' the daily build basically
because it needs to run on the same machine (or with access to the
filesystem) where the CVS *,v files are. Since the daily build's build.xml
should be able to be invoked on any machine, we keep it out of the daily
build build.xml and invoke it as a scheduled task on the CVS server
machines. (Note that the CVS sensor runs on both hackydev and katrina.)
If the Jira sensor can be invoked from any machine, then I see no reason
why it couldn't run inside of the daily build build.xml.
Cedric, what's your opinion?
Cheers,
Philip
--On Monday, April 11, 2005 1:08 AM -1000 Aaron Kagawa <[EMAIL PROTECTED]>
wrote:
Philip, Cedric, and Burt,
Ok. I just realized something.
I've been advised (a while ago) by Cedric that we should run the Jira
sensor like the CVS sensor, which is ran as schedule task. See the email
below.
My main issues with doing this are:
- (Pro) Like the CVS sensor, the Jira sensor will run every day regardless
if there is a build or not.
- (Con) If the Jira Sensor fails then we don't have a mechanism to let
anyone know about the failure. Does the CVS sensor have that
functionality?
So, the question is, should we move the Jira Sensor to a scheduled task?
Or keep it in the daily build? What is more natural?
I don't mind keeping it in the Daily Build, but the only issue I have is
that the daily build doesn't run on a daily basis. Rather it only runs if
there are changes detected in CVS.
Philip: Which would you rather have; Daily build or scheduled task? Burt
is scheduled to change the Jira sensor to run via a schedule task soon.
Cedric: How hard would it be to run the daily build everyday?
thanks, aaron
Cedric wrote:
http://hackystat.ics.hawaii.edu/hackystat/docbook/apas07.html
CVS sensor runs by invoking command line: "ant cvs-sensor-task", and this
command line invocation is scheduled using unix cron or windows
scheduler.
If I am correct, Jira sensor runs by taking a snapshot to jira database.,
and it cannot be run retrospectively. Since there is no off-line storage
for multi-account, be prepared to handle missing data or patial data in
the analysis. (CVS sensor has encountered such problem, Jira sensor will
too).
Cheers,
Cedric
Aaron Kagawa wrote:
It just came to me:
Why you need to put jira sensor in build.xml so that it runs when
Hackystat-ALL runs?
It seems to me that Jira sensor works the same way as CVS sensor. Why
cann't you take the CVS sensor approach?
How does the CVS sensor do it?
thanks, aaron
