Details
-
Bug
-
Status: Closed (View Workflow)
-
Trivial
-
Resolution: Fixed
-
None
-
Platform: PC, OS: Windows XP
Description
If master is in the time zone that is behind the slave machine. and the time
difference is more than twice than SCM polling frequency, one file version is
included in more than one build. If there is only one checkin in the deration
same build is getting build more than once.
I think it will show unusual behaviour if master time zone is ahead of slave.
In this it may skip few of the file while showing the change list
Attachments
Activity
I believe the underlying problem here is that the previous build time is taken
straight from the master and used in the cleartool lshistory arguments. So, for
example, in my particular case, I've got the master in PST and the slave in EST,
so if the last build ran at 1pm PST on 2/20/2009, the ct lshistory -since
argument on the slave comes out as "20-feb.13:00:00". There's actually a couple
things wrong with that - it's not smart to not have the year there, and
lshistory accepts UTC timezone information in its dates. So
"20-feb-09.13:00:00UTC-0800" would be a valid value, and would be the correct
value for the last build time, regardless of what the timezone we're running the
lshistory in is.
I'm compiling a local build of clearcase.hpi now, with a change to ClearToolExec
to fit this. I'll test it out and update here.
Created an attachment (id=576)
Diffs to ClearToolExec.java and ClearToolExecTest.java to use UTC timezone and year in lshistory -since argument
Hi again -
Yup, tested it out and it worked like a charm. The only problem I ran into was
that ClearToolExecTest's lshistory test was comparing against a static string,
so I had to change ClearToolExecTest to use the same SimpleDateFormat/UTC
timezone behavior.
I've never actually submitted anything to an OSS project before, so I'm just
going to attach my diffs and wait for y'all to review this and/or let me know if
you want me to commit it. =)
sunfire
added a comment - Patch applied.
Thanks for the fix! Code changed in hudson
User: : sunfire
Path:
trunk/hudson/plugins/clearcase/src/main/java/hudson/plugins/clearcase/ClearToolExec.java
trunk/hudson/plugins/clearcase/src/test/java/hudson/plugins/clearcase/ClearToolExecTest.java
http://fisheye4.cenqua.com/changelog/hudson/?cs=15764
Log:
JENKINS-2493 Thanks Ajay KUMAR for the patch
I think if we manage to do following two things the problem can be solved
1) add the slave time difference while polling clearcase.
2) poll at the host where the actual build is being performed master.