Received: from [10.1.2.163] (root@raven.ravenbrook.com [193.82.131.18]) by raven.ravenbrook.com (8.9.3/8.9.3) with ESMTP id AAA21510 for ; Wed, 25 Oct 2000 00:14:19 +0100 (BST) Mime-Version: 1.0 X-Sender: gdr@pop3 Message-Id: Date: Tue, 24 Oct 2000 17:13:17 -0600 To: p4dti-staff@ravenbrook.com From: Gareth Rees Subject: TeamShare PSG alpha test report, 2000-10-23 Content-Type: text/plain; charset="us-ascii" ; format="flowed" This is a report on testing P4DTI release 0.3.2 at TeamShare PSG with Derek George. 1. The integration won't work with SourceBridge. We need to tell people to uninstall it. We also need to explain in the AG and/or UG how to achieve whatever SourceBridge achieved. 2. The prerequisites section of the AG needs to give more explicit instructions. More whitespace would be nice, too. It's very hard to follow. 3. The installation guide needs to give explicit instructions for which machine to install each prerequisite on. 4. "See figure 2" in AG is wrong. (Should be figure 1.) 5. You can skip step 3 of installing if you're using an existing TeamTrack database. Point this out in the AG. 6. The "Administrator" group might not exist at some sites. We need to figure out the recommended set of privileges for the replicator and give thre replicator those privileges explicitly. Derek will help us figure this out. 7. The "Upgrade" installation option didn't seem to work when we installed 4402. We couldn't connect to the web server. So we tried the "Re-install" installation option and it worked. This is probably a bug in the 4402 installer that needs to be reported to TeamShare. 8. The P4DTI- fields in the Perforce jobspec need to be at the bottom of the job so that they're out of the way. So number them with large numbers. 9. We need to give suggested types, lengths and persistences for different kinds of fields. Installers may not be jobspec experts. When is it sensible to use "optional" and when "once"? 10. Perforce insists on there being job statuses "open" and "closed". (New jobs get "open" status by default; fixed jobs get "closed" status by default). Tell people to put these in the Values-Status field in addition to the statuses they replicate from TeamTrack. 11. Tell people to add comments to the jobspec explaining what the replicated fields are. We could provide a comment for the P4DTI- fields. 12. The AG needs to have examples of field mappings. 13. It's a bad idea to have two state maps (one for each direction). If you need this kind of complexity you're probably in all sorts of other trouble already. Have just one map and read it in both directions. 14. Most of the time the state map will be very simple. Could we provide a default map that just converts from spaces to underscores? Derek points out that state names can be very long in TeamTrack and people will want to have short forms of state names in Perforce, so we need to preserve the option of having an explicit map. 15. Cases may have no OWNER. So what do you replicate to the User field in Perforce? 16. Various errors can repeat themselves, because the replicator discovers a conflict and tries to set the P4DTI_ACTION field. This can also fail for the same reason. Missing required fields in Perforce is a typical situation. We need a way of recovering from this. 17. Supposing that the initial replication goes wrong. Then the case is marked "wait" but there's no corresponding job to set to "keep" so you can't fix the problem without going into the database directly. 18. Bug in Perforce: we saw Perforce change a field's disposition in the jobspec from "optional" to "once". 19. How do you tell if the configuration is correct? There should be automatic checking of the configuration. See e-mail "Checking the configuration automatically" for some ideas. 20. Should you add a newline to a field when it's blank (in text translation from TeamTrack to Perforce)? It looks to me as though this doesn't do the right thing. Blank text fields are really blank in Perforce. So 21. If a field in a job is optional, it may have no value. This means that when you do a p4 job -o the field simply doesn't show up. That means that the replicator crashes, because it's assuming that any field to be replicated must appear in the job.