[kepler-dev] [Bug 4559] - local repositories not showing up in components tree

[bugzilla-daemon at ecoinformatics.org]

http://bugzilla.ecoinformatics.org/show_bug.cgi?id=4559


soong at nceas.ucsb.edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |4512




------- Comment #10 from soong at nceas.ucsb.edu  2009-11-20 13:15 -------
I think this bug is partly due to bug 4512, partly due to the new hiding empty
repositories "feature", partly due to changes in the KAR structure, and partly
due to an ordering problem.  There's also another slightly separate problem
with local repository content not being added properly (differences between
adding repositories from within Kepler and rebuilding the cache on Kepler
startup), so that I have to close and restart Kepler before it finds any
components.

So to explain.  The tpc01 report archive contains an entity "ROML" and
"tpc01-buffalo-tb" (the MOML).  The tpc01 run archive contains those two
entities as well as "RIO" and "WorkflowRun".  Because of bug 4512, if the run
archive is added to the cache first, then there is no new content in the report
archive, and so the new functionality hides it.

If the report archive is added first, then the run archive adds some new
content (but no new MOML or ROML) and is shown.  I'm not sure why that triggers
bug 4563.

There seems to have been a change in the way ROML is stored.  If I generate a
new pair of report and run archives, the report archive contains "test" (the
MOML) and "test_ROML".  On the other hand, the run archive contains "RIO",
"WorkflowRun", and "ROML" in addition to "test" but no "test_ROML".  The LSIDs
for test_ROML and ROML also differ, even though the reports themselves never
changed.

I think this is all due to the contents of MANIFEST.MF in the KARs and the
names coming out of it.  As far as I can tell, the names that show up in the
components tree are the names in MANIFEST.MF truncated to the first dot (.). 
The name for the ROML used to be identical between the run and report archives,
but now it differs.  Also, the tpc01 run archive seems to contain 2 entries for
the ROML with different names but the same LSID.  The test run archive seems
not to have that problem.