Could you please revert this please.
the P4Java can't sync large number of files, crashes the slave due to memory issues and isn't as fast as the command line native one

List<TagAction> actions = run.getActions(TagAction.class);
if (actions.isEmpty()) {
return;
}

    // Grab the last action
    TagAction tagAction=actions.get(actions.size()-1);

ok, I can see why my problem is, I had job parameters that matched the special json names that the plugin used, and those was used and not the pipeline items, when I use the pipeline items then both pipeline items sync to the correct CL's

correction to above , the first sync and second sync it totally ignoring the pin value and is just syncing to the last.
Will add some debug to see why

This doesn't work.

My test is as follows, i have a wrapper
def stream_sync(sync_stream,credential_id,perforce_label)
that i use to sync and return me the P4_CHANGELIST.

This works perfectly for a single sync.

def stream_sync(sync_stream,credential_id,perforce_label)
{
echo perforce_label
def _friendly_stream = syncstream.replaceAll('/','')
def p4_local = 'C:/u-blox/gallery/perforce/p4/win_64/2015.2/bin/p4'
if (workspace.startsWith('/'))
{
p4_local = '/u-blox/gallery/perforce/p4/re6_64/2015.2/bin/p4'
}
p4sync charset: 'none', credential: credential_id,
format: _friendlystream+''+env.EXECUTOR_NUMBER+'_jenkinspipeline${SITE}-'+env.NODE_NAME,
populate: [$class: 'AutoCleanImpl', delete: true, modtime: true, parallel: [enable: true, minbytes: '1024', minfiles: '1', path: p4_local, threads: '4'], pin: perforce_label, quiet: true, replace: true], stream: sync_stream
return currentBuild.getRawBuild().getEnvironment()["P4_CHANGELIST"]
}

node(slaveRestriction) {
def workspace ='workspace/'+friendlystream+''+env.EXECUTOR_NUMBER
ws(workspace)
{
withEnv(["WORKSPACE=${pwd()}"]) {
stage ("sync code"){
def P4_CHANGELIST=stream_sync(perforceStream,p4_credential,'12345')
echo P4_CHANGELIST

P4_CHANGELIST=stream_sync(perforceStream,p4_credential,'12000')
echo P4_CHANGELIST

my example above is to sync to the same stream twice, just with a different CL
The first sync to 12345 works , no problems.

The second one has a problem before we even get to the env.
I pin to 12000 but the sync task still thinks the pin value is 12345 and the console log shows it is syncing to 12345 and not 12000
the env returned is 12345 since 12000 was not used

the getLastAction does seem to do more than get the last Action for a particular run
It seems be be written for a polling task.

If i have a pipeline where I sync 2 different streams then i would expect 2 different P4_CHANGELIST values, not do any comparisons with other items.

I will sync and build the code tomorrow, but just having a quick look it doesn't look right (unless i am missing something)

I have now got my head around on how the extension point works.
The problem was to get to currentBuild.getRawBuild().getEnvironment() without good documentation.
just using env doesn't work.
I can now proceed to pull the values out

I have been trying to make this work....

if I modify the code to:
@Extension()
public class P4EnvironmentContributor extends EnvironmentContributor {

<a href="/@Override" target="_self">@Override</a>
public void buildEnvironmentFor(Run run, EnvVars env, TaskListener listener) throws IOException, InterruptedException {
    TagAction tagAction = run.getAction(TagAction.class);
    listener.getLogger().println(&quot;buildEnvironmentFor&quot;);

Then I can see the "buildEnvironmentFor" printed BEFORE the P4 step run.

Looking at the docs, this API is called at the beginning of a well written workflow step to get the env variables.

But this doesn't set the env for main flow only for other steps that does call buildEnvironmentFor.
The sh or batch step doesn't call buildEnvironmentFor either...

I want to do
currentBuild.displayName = "#${BUILD_NUMBER}.${env.P4_CHANGELIST}"
for example...

how do I get the env back ? (without writing my own step?)

At the moment i set my variables system wide, so i expect all windows and linux slaves to have their p4 tools on the same path.
Since we mount our tools on linux and i use jenkins to push p4.exe to my windows slaves i can do this.

i could not get it to work without it.....
Since i am running on any slave , i couldn't know if the slave had anything set.

If we can set a ticket and then do the call then that would be the best way to go but i couldn't figure out how to get the auth settings into a P4TICKET for the cmd line to work.

this i couldn't reproduce...
just randomly fails.

When i looked at:
https://www.perforce.com/perforce/doc.current/user/p4javanotes.txt

I saw

1308053 (Bug #83624)

    Fixed a potential problem where busy concurrent p4 login and
    p4 logout commands could wipe out (loose) tickets in the
    P4TICKETFILE file by creating a shadow P4TICKETFILE lock
    (".lck") file.

not sure if it is related.
But at least the try catch won't kill the job, it will just continue with some env not set.
I wasn't sure how many threads you have so doing the get and then the check would be safer.

BTW i am currently using the username/pw credential but was going to see if i move over to the ticketing credential if would improve things.

Been working with Karl Wirth, have found the the use case where the p4java sync hangs.

My other shelve is about the P4_TICKET=NULL, saw a bug in the p4java change notes about the same thing.
My change will at least give some protection for the error

so if the id is a shelved cl then this command fails and the job bombs out.

if id is an shelved cl then it throws:
Unable to get current change: com.perforce.p4java.exception.RequestException: Can't use a pending changelist number for this command

i also tried the command using CLI and get the same error

using just @ and setMaxMostRecent(1) works for any type of change

ClientHelper.java getClientHead

// get last change in server
// This will returned the also shelved CLs
String latestChange = connection.getCounter("change");
int change = Integer.parseInt(latestChange);

This counter also have shelved CLs.

Inside ConnectionHelper
public IChangelistSummary getChangeSummary(int id) throws P4JavaException {

there was
List<IFileSpec> spec = FileSpecBuilder.makeFileSpecList("@=" + id);
and this command doesn't work with shelved CLs.

We are now hitting this many times a day.

let me restate how i work.
I have 1000s of projects that use the p4 plugin to sync to a custom workspace BUT they don't poll for changes.

I have a trigger job that uses the p4 plugin to sync to a custom workspace and then automatically trigger all the relevant jobs that does need rebuilding based on filter.

That way i can have a handfull of jobs polling p4 vs 1000s of jobs polling p4

I have not checked any logs of the trigger jobs that do the polling to see if there is a problem.

List<IFileSpec> synMsg could potentially be very large, correct ?
If so then wouldn't it be better to use the stream callback method and check each message as it it returned ?

That way you are guaranteed to have a zero memory impact and check for hangs by detecting that if no new callback have been called for n seconds then something must have gone wrong

The plugin seems to randomly fails after this point.
Either the server progress messages get deadlocked but we can see the actual p4 thread has finished by the plugin seems to sit here

I use EXECUTOR_NUMBER in all my custom workspaces and if i don't use a custom workspace then i use EXECUTOR_NUMBER in the generated template name.
e.g. ${SITE}-${STREAM}-${NODE_NAME}-${EXECUTOR_NUMBER}

${SITE} allows me to cover the case where the same job is going to be run on different masters. (we have our own variable called SITE that has a meaningful value for each master.

${NODE_NAME}-${EXECUTOR_NUMBER} allows me to cover all the cases of have a single job running in parallel on the same slave or across different slaves.
Using ${STREAM} allows me to have any number of jobs all sharing the same set of workspaces.
So if i have a slave with 4 executors and 100 jobs that all use the same stream then i will only have 4 workspaces. (and not the 100 workspaces+parallel ones that is jenkins default)

Hi, i use the ability to set a jobs custom workspace myself, due to various bugs in Jenkins when running parallel builds.

Please check if your code will still work if the user have specified a custom workspace

i would use 3 different jobs, jenkins doesn';t support these type of things like TeamCity or ElectricCommander does so you have to split the workload.

2 jobs that sync the different code trees to a known location and a 3rd to do the compare ?

Hi,
I have asked Jenkins/Cloudbees a similar question regarding core support for multiple SCM steps in a job.
They pointed me to a "proof of concept" plugin someone wrote for SVN.

As i understand it the current default model only handles a single P4 SCM step, do you have jobs that can actually sync,track changes,trigger etc from multiple SCM ?

Morné