v2_11 doesn't fix provenance from v2_10
In this issue
a provenance problem was found in art v2_10 and fixed in v2_11.
Now we are testing art v2_11_01 and still see a form
of the problem. However we are not sure if the fact that
the input files were created in v2_10 means that they will
always be corrupt, or they should be read cleanly by v2_11.
Below is the how to reproduce the current error. Is this
expected? Is there a way to repair files or suppress the error?
In process 1:
(sets up art v2_10_04)
mu2e -n 10 -c JobConfig/common/artcat.fcl -s $IN -o copy_old.art
In process 2
(sets up art v2_11_01)
mu2e -s copy_old.art -c Print/fcl/print.fcl
produces error: Product reported as not present, but is pointed to nonetheless!
Note that if the steps in process 1 are done in process2, creating
copy_new.art, the result will be read cleanly by v2_11. If copy_old.art
is read and written by v2_11, then the provenance is read in v2_11, it
produces the error, so v2_11 isn't repairing the bad file. Is this expected?
The releases can be created locally by
git clone http://cdcvs.fnal.gov/projects/mu2eofflinesoftwaremu2eoffline/Offline.git
git checkout v7_1_2 (or v7_1_2a)
source Offline/setup.sh (instead of cvmfs)
scons -j 20
#5 Updated by Kyle Knoepfel about 1 year ago
- Category set to Infrastructure
- Status changed from Assigned to Feedback
- % Done changed from 0 to 100
- Estimated time set to 4.00 h
The problem is understood. Although the changes made for art 2.11.02 addressed the persistency issues when using the
RootOutput module, the workflow here uses a
ProvenanceDumper-based module, which is independent of
RootOutput and was not updated for art 2.11.02. In addition there were some mistaken assumptions regarding how art calculates the availability of the product.
The following commits resolve these problems:
Does Mu2e request a bug-fix release (e.g. 2.11.05), or shall we just include the fix for the next planned release of art, which is version 3.02.00? I suspect the former.