fix(reporting): resolve #283 review blockers (Print Schedules — connection deadlock + bookmarks/comments/perf)

[gpascucci]

Why

#283 (Story 20.2, combined Print Schedules PDF) was merged while the review was still CHANGES_REQUESTED, so these five review fixes — including a blocker — did not make it into main. This is the follow-up.

Fixes

1. 🔴 Connection-pool deadlock (blocker). PrintService.render was @Transactional(readOnly=true) — the tx manager held one pooled connection for the whole method while ReportService.fillSchedule9 checked out a second raw connection. So every /print including Schedule 9 pinned 2 connections on a pool of 5; ~5 concurrent prints (year-end close) deadlock until the 30s timeout. render() is six independent read-only reads — dropped the annotation (each service manages its own tx; the Sch 9 fill borrows/releases its own connection).
2. Single-schedule bookmark. The sectionCount > 1 gate meant a one-schedule /print had no outline (violates BR-08). JR 7.0.8 batch bookmarks only emit for >1 doc, so switched to a per-template outline anchor (bookmarkLevel=1) threaded via a bookmarkTitle fill param; standalone GET /schedule9 stays bookmark-less. New IT: single Schedule 6 → exactly one bookmark.
3. Comments-only keeps the row identifier. With printScheduleInformation=false, printComments=true, p_do_print_body had been hiding the identifier columns. Un-gated areaLabel (Sch 6), type (Sch 7B), location (Sch 11) — matching legacy (which left them ungated) and Sch 5. New IT asserts the identifier survives.
4. Deferred-selection logging now fires on the allSchedules path too.
5. Perf: Schedule 9 empty pre-check uses a COUNT query instead of materializing the full record list then re-querying in the template.

Verification

421 ITs green (PrintScheduleIT incl. the two new bookmark/identity cases), checkstyle clean — verified in WSL against Testcontainers Oracle. Cherry-picked cleanly onto current main.

Resolves the review comments from @SScholefield and @paulushcgcj on #283.

🤖 Generated with Claude Code

---

Thanks for the PR!

Deployments, as required, will be available below:

• Frontend
• Backend

Please create PRs in draft mode. Mark as ready to enable:

• Analysis Workflow

After merge, new images are deployed in:

• Merge Workflow

[gpascucci]

Self-CR (adversarial pass over the five fixes)

Reviewed each fix against its intent; all sound, and I caught + fixed one gap in my own change.

• 🔴 Connection deadlock — @Transactional removed from PrintService.render; correct — it's six independent read-only reads (each *Service runs its own tx, the Sch 9 fill borrows/releases its own connection), so no method-wide tx that would pin a second connection. Rationale documented inline.
• Bookmarks — the batch-mode approach genuinely can't bookmark a single-doc PDF in JR 7.0.8 (JRPdfExporter gates addBookmark on items.size() > 1); the per-template outline anchor (bookmarkLevel=1, anchorNameExpression=$P{bookmarkTitle}, gated on non-null) is the right fix. Standalone GET /schedule9 passes null → no anchor. Single-schedule + combined bookmark cases both have ITs.
• Comments-only identity — un-gated location (Sch 11), areaLabel (Sch 6), type (Sch 7B) from p_do_print_body, verified against legacy (which left those ungated) and matching Sch 5. IT added.
• Sch 9 perf — countRecords mirrors findRecords' WHERE (category '9', mill/year) with no joins/projection; empty→404 preserved.
• Deferred-selection log (fixed a gap) — the first pass only checked the individual deferred flags, so a pure {allSchedules:true} request still wouldn't log the skipped 1/2/3/4/8/10 — the exact case @paulushcgcj raised. Fixed in 91866dc (include allSchedules in the condition).

Verified: 421 ITs green + checkstyle clean; the log-only follow-up compiles clean (no test impact). Ready to merge — this brings main up to the reviewed 20.2 state.