Kepler

Regarding the part about the ROML being lost (second described symptom), it reminds me a lot of bug 4592, but I think now the MOML needs to be changed between the saves to trigger it.

So:
1. New workflow, File->Save Archive, Change MOML, File->Save Archive: error
2. Saved KAR, Change MOML, File->Save Archive: no error
3. Saved KAR, Change MOML, File->Save Archive, Change MOML, File->Save Archive: error
4. Saved KAR, Change MOML, File->Save Archive, close & re-open, Change MOML, File->Save Archive: no error
5. Saved KAR, Change MOML, File->Save Archive, close & re-open, Change MOML, File->Save Archive, Change MOML, File->Save
Archive: error

Still not sure why ROML->RIO doesn't seem to link up (first described symptom)

This is observed on both r22183 and r22155. Still trying to find a build that won't lose my data...

I should also point out that bug 4571 shows up in cases 4 and 5.

4. Saved KAR, Change MOML, File->Save Archive, close & re-open (report layout is blank, but ROML exists), Change MOML, File->Save Archive: no error (report layout displays again)
5. Saved KAR, Change MOML, File->Save Archive, close & re-open (report layout is blank, but ROML exists), Change MOML, File->Save Archive, Change MOML, File->Save Archive: error (no report layout and no ROML).

I think I can hopefully clarify the example and maybe help show why this is incredibly confusing:

1. New workflow, File->Save Archive (kar1), Run. The report designer shows the report, the report viewer is a blank page, and kar1 has workflow_ROML.xml.
2. Bump LSID (I usually delete & undo), File->Save Archive (kar2), Run. workflow_ROML.xml is missing from kar2.
3. Close & reopen kar1, Run. The report designer is blank and the report viewer shows the report. I think this is bug 4571.
4. Bump LSID, File->Save Archive (kar3), Run. The report designer shows the report, the report viewer is a blank page, and kar3 has workflow_ROML.xml.
5. Now, if we export all the runs, the first is missing workflow_ROML.xml, the second has a minimal workflow_ROML.xml, the third has a minimal workflow_ROML.xml but the full ROML.#.xml, and the fourth has what I think is RIO.#.xml content in workflow_ROML.xml but a minimal ROML.#.XML and a minimal RIO.#.xml.

There is another test sequence:

1. New workflow, File->Save Archive (kar1), Run, Export run (run1). The report designer shows the report, the report viewer is a blank page, kar1 has workflow_ROML.xml, and run1 is missing workflow_ROML.xml.
2. Bump LSID (I usually delete & undo), File->Save Archive (kar2), Run, Export run (run2). kar2 has a minimal workflow_ROML.xml, and run2 is missing workflow_ROML.xml.
3. Close & reopen kar1, Run, Export run (run3). The report designer is blank and the report viewer shows the report. Bug 4571 again, but run3 has the full workflow_ROML.xml.
4. Bump LSID, File->Save Archive (kar3), Run, Export run (run4). The report designer shows the report, the report viewer is a blank page, kar3 has workflow_ROML.xml (with RIO.#.xml contents), and run4 is missing workflow_ROML.xml.

Right now, I think these examples show off some of the biggest problems in the way of being able to repeatedly change, save, and run, which is pretty central to building reports.

I've at least addressed some of this in r22198.
I concentrated on editing wf/layout->saving kar->editing->saving.... So that the layout is not lost and always shows up in the designer view after you save/reopen.
I don't believe running the workflow between saves should have much bearing on how the archive is written.

This actually looks great right now. It seems to clear all the test cases.

Original Bugzilla ID was 4632