We are facing the issue with the p4 software plugin in the case where we have parent/child jobs running. When we triggered the parent's job it captures the P4_CHANGELIST to pass down to the child job. However, in between running the parent job and the child job, if there are changelists submitted to perforce, then the value for P4_CHANGELIST is changed for the child job. This is not good for us since we need to make sure that both the parent and child jobs are syncing down to the same P4_CHANGELIST. Is there anyway that the p4 plugin will allow us to set the P4_CHANGELIST and not use the latest changelist in order for us to ensure that the same P4_CHANGELIST are used by both parent and child jobs?

The "Use the upstream project revision" appears to be a solution to resolve, and ask this be prioritized.

One workaround is to use parameterized jobs and the following plugin:

https://wiki.jenkins-ci.org/display/JENKINS/Parameterized+Trigger+Plugin

Create a downstream job with a string parameter 'P4CL' then pin the downstream job at the Perforce label '${P4CL}'. When the upstream job calls the downstream job set 'P4CL= ${P4_CHANGELIST)'.

In this was the downstream jobs runs with the same changelist number that was used by the upstream job. In testing this also sets 'P4_CHANGELIST' to '${P4CL}'

Code changed in jenkins
User: Paul Allen
Path:
SETUP.md
http://jenkins-ci.org/commit/p4-plugin/0a667176ad6dfb38ac755cc8e556875562dcd9ce
Log:
Document downstream changelists.

Workaround for JENKINS-33163.

Yes, we are already using that workaround using the pinned downstream build.
However, as I said in the description of the issue that this is not good as you then have to list all downsteam jobs in parameterized trigger section of the upstream job, rather than in the downstream job (which is the usual method to link jobs in jenkins)

1.4.9 (documentation)

Fixed in 1.4.9

I'm seeing behaviour similar to that reported by Jim Poje above.

I'm following the documented pattern, but the change list reported by $P4_CHANGELIST (when captured by the parent job and passed to as a parameter to trigger the child job) reflects more recent commits.

I suspect there may have been a regression in the plugin. I'm going to downgrade (1.10.10 -> 1.9.6) and monitor over the next few days.

Hi philipa - Has this ever worked for you? From this thread, it looks like we didnt put in a code fix and only provided a workaround where you avoided using P4_CHANGELIST.

Hi philipa - Yes that's the one. I'll do some basic tests here to see if I can repro this.

Hi philipa - I just tried the following test with 1.10.10 and it seemed to work:

(1) Add CL 123 to downstream job directory.

(2) Add CL 124 to upstream job directory.

(3) Trigger upstream job that has a 20 second sleep as a build step.

(4) Add 2 more CLs to each directory.

(5) Downstream job is triggered with CL124 as the paramater and jobs is pinned to that CL. P4-Plugin looks for most recent CL in downstream depot path and syncs CL 123.

My setup is - Upstream job...

Downstream job...

and:

Thanks for investigating Karl. Yes, that configuration corresponds to mine. I guess the one difference might be that my jobs run on nodes on different machines, with different workspaces. Also, the later commit will trigger another run of the upstream job (on a different machine).

Its not recurred since I've downgraded, so I'm going to run with this for 1.9.6 for next week, then I'll try upgrading again.

Hi philipa - Thanks and happy I can help.

I ran for a week 1.9.6 with no issues.

I switched back to 1.10.10, and saw the problem again. It's a production config, so I've reverted back to 1.9.6.

Hi philipa - Thanks. I'll test on WIndows and Linux build slaves and get back to you.

Hi philipa

Finally got a chance to test this properly and still no luck in reproducing it. Can you please send an email to 'support@perforce.com' for my attention and send the full console logs for the two jobs when th. I'll try and figure out what I'm missing.

Thanks in advance,

Karl

Thanks p4karl. I've upgraded the plugin again; I'll send you the logs when I see the next failure.

Hi philipa - Thanks.