Do you actually have '::' in the client name? I don't think that's valid.

Yes, I actually have '::' in the client name. We have that in (virtually) all our client names, and Perforce seems quite happy with that

It doesn't seem very happy to me. =O

There weren't any changes made to the tek42 stuff in 1.1.9, so I can't see how the upgrade might have caused this to break.

Any suggestions for me, then?
I know you've fixed some stuff concerning names of remote clients - I was thinking this might be related, perhaps vaguely?
I've only seen this problem happen for builds on a slave (remote clients), but not all of those have used parameter substitution on their names. One of them did, the other 3-4 didn't.

I think I actually deleted the extracted plugin subdirs this time before/during the upgrade, so there shouldn't be any conflicts with leftover code.

I would downgrade temporarily to see if the issue persists. It definitely looks like an environment issue to me, likely something to do with your connection to the slaves or to the perforce server.

I've also got a CI build at http://www.drikanis.org/hudson/target/perforce.hpi that has some unrelated fixes you could try, but as with this last release, the code dealing with the tek42 interface has not been changed.

I'll check the servers tomorrow. The slave machine did have a crash the other day, so perhaps it's not entirely well.

I don't suppose anyone who can reproduce this problem has the ability to build the perforce plugin? We may have to bisect the changes between 1.1.8 and 1.1.9 to see what is causing this. If not, I can provide a set of builds that you can test.