CLARIN-DSpace v9/Restore the fork health checks taken wholesale from vanilla (23 of 26 report-diff mappings have no emitter)

[milanmajchrak]

References

• Acceptance card X-08 (_sync3/cards/X-08.md); depends on BE-13, merged as ac1eeecd79
• The fork hunks come from 3dc9cecf7d "Health report, report diff fixes (Health report, report diff fixes #1254)" on dtq-dev
• Instance of guard X1 (vanilla-wholesale): ItemCheck and Report are the two largest entries in the BE X1 worklist

Description

InfoCheck, ItemCheck and UserCheck were byte-identical with vanilla dspace-9.3 on this branch
although the fork changes all three. Their setReportJson calls were therefore gone, and 23 of the 26
mappings in report-diff-fields.json addressed a value that nothing produced. report-diff printed
those columns empty and no test failed.

Instructions for Reviewers

The defect

report-diff never reads the checks. It reads the JSON that a health-report run stored and addresses each
value by a path such as /checks/[name=Item summary]/report/collectionsSizesInfo/totalSize. Nothing links
that path back to the class that is supposed to emit it, so when the v9 upgrade took the three classes from
upstream whole, the mappings kept resolving to nothing and the only symptom was empty columns.

Measured on the base (6e8185521f):

$ for c in InfoCheck ItemCheck UserCheck; do f=.../health/$c.java;
    [ "$(git rev-parse origin/dtq-dev-9-base:$f)" = "$(git rev-parse dspace-9.3:$f)" ] && echo "STILL-VANILLA $c"; done
STILL-VANILLA InfoCheck
STILL-VANILLA ItemCheck
STILL-VANILLA UserCheck

$ git grep -l setReportJson origin/dtq-dev-9-base -- 'dspace-api/src/main/java/org/dspace/health/*.java' | wc -l
3        # Check.java (the setter), LicenseCheck, MetadataCheck

Breakdown of the 23: Item summary 17, User summary 4, General Information 2.

List of changes in this PR

• InfoCheck - emits generated, fromTill, url and the directoryStats array. report-diff
  addresses that array positionally (directoryStats/0 assetstore, /1 log dir), so an entry is appended
  for every directory even when it cannot be read; dropping one would shift the other's index.
• ItemCheck - emits communities, collectionsSizesInfo (with totalSize, deletedBitstreams,
  orphanBitstreamsCount, ...), publishedItems, withdrawnItems, notPublishedItems, stagesCounts,
  waitingForApproval and the eleven entity counts. All 17 Item summary mappings.
• UserCheck - emits users, haveEmail, the camel-cased per-attribute counts (including
  selfRegistered), emptyGroups, subscribers and subscribedCollections. It also carries the fork's
  fix of the "Self registered" counter, which tested getNetid() twice and so reported a copy of
  "Have netid".
• Report.java deleted, and its healthcheck command removed from launcher.xml - see below.
• ReportDiffFieldMappingIT (new) - three tests that keep this from happening again.

Two deliberate deviations from the fork, both to avoid importing a defect:

1. ItemCheck.getObjectSizesInfo keeps the v9 direct countTotal(context) calls instead of the fork's
   wrapSql + LinkedHashMap<String, CountInfo> refactor. wrapSql converts the checked SQLException
   into a RuntimeException, which run()'s catch (SQLException) cannot see - one failing count would
   abort the entire health report instead of being recorded by error(e) and the rest of the report
   finishing. The JSON keys, their order and the report text are the fork's.
2. ReportInfo returns LocalDate on v9 and Date on the fork, so the dates are formatted directly with
   DateFormatConstants.DATE_FORMATTER rather than through Date.toInstant().atZone(...).

Minor: !s.isEmpty() kept where the fork writes s.length() > 0; instance service fields kept rather than
the fork's static final.

Why Report.java is in this PR

3dc9cecf7d deleted org/dspace/health/Report.java when the health-report DSpaceRunnable replaced it,
and removed the healthcheck command that pointed at it. The v9 base kept both: the class, and
launcher.xml:14 still declaring healthcheck with <class>org.dspace.health.Report</class>. So
dspace healthcheck on the CLI ran the pre-script class, which writes neither setReportJson output nor a
ReportResult row - report-diff would have had nothing to diff even after the three checks above were
fixed. The replacement is already wired on this branch
(dspace/config/spring/api/scripts.xml:115, bean health-report), and Report had no other reference in
the tree. It belongs to the same fork hunk and the same feature, so it goes in the same PR. The
matomo-report-generator command that sits in the same launcher.xml delta does not - that is card X-04.

How to test

mvn -o -pl dspace-api verify -DskipIntegrationTests=false -Dit.test=ReportDiffFieldMappingIT \
    -Denforcer.skip=true -Dcheckstyle.skip=true -Dlicense.skip=true -Dxml.skip=true
# Tests run: 3, Failures: 0, Errors: 0

Test	What it pins down
everyMappedFieldIsEmittedByItsCheck	runs a real health report and resolves every mapping in report-diff-fields.json against the JSON it stored
everyMappedCheckNameResolvesToACheckOnTheClasspath	every [name=...] resolves to a Check plugin registered in config/modules/healthcheck.cfg
fieldMappingsAndFieldOrderDescribeTheSameFields	ReportDiff iterates fieldOrder and only looks the label up in fieldMappings, so a field in one and not the other is silently dropped

Negative control - revert each class to its vanilla blob and the test names exactly the mappings that class
owns:

reverted to dspace-9.3	unresolved	paths named
ItemCheck	17	all [name=Item summary]
UserCheck	4	all [name=User summary]
InfoCheck	2	both [name=General Information]/report/directoryStats/N/size_bytes

17 + 4 + 2 = the 23 mappings this PR restores.

Regression gate: HealthReportIT - Tests run: 12, Failures: 0, Errors: 0.
mvn clean install -P-assembly -DskipTests - BUILD SUCCESS, 0 Checkstyle violations.

Found while working, not fixed here

dspace-api/src/test/java/org/dspace/scripts/ReportDiffIT.java exists on dtq-dev (1016 lines) and is
absent from this branch, and no acceptance card owns it. The report-diff script itself is ported here;
only its test is missing. It needs its own card.

Checklist

• My PR is created against the dtq-dev-9-base branch (fork's DSpace 9 base; not main).
• My PR is small in size (not counting the deleted dead class and the new IT).
• My PR passes Checkstyle validation.
• My PR includes Javadoc for the new and modified non-trivial methods.
• My PR passes all tests and includes a new Integration Test.
• My PR includes details on how to test it.
• My PR includes no new libraries/dependencies.
• My PR does not modify REST API endpoints.
• My PR includes no new configuration keys; it removes one dead launcher.xml command.

🤖 Generated with Claude Code