[na] Share the resource-report overlay between pipeline paths

[arniber21]

Summary

overlay_target_report is the shared report-overlay stage for the bare NA pipeline and the hybrid QEC schedule. pipeline::finish_pipeline and qec_schedule::schedule_expanded both call it after build_resource_report.

One function applies:

• aware-search status, when the caller has it
• schedule-optimality label, when the caller has it
• routing-agnostic placer mechanism
• analytic error budget, when the target carries error_model
• end-to-end fidelity estimate (fidelity is mandatory)
• atom-loss budget, when the target carries atom_loss_model

Those writes touch disjoint fields, so this order matches both previous inline chains. QEC workload counts (memory_rounds, t_count, tdag_count, ccz_count, magic_state_demand) stay in schedule_expanded. The bare pipeline has no workload to size them from.

Issue 468.

Behavior

Resource reports are unchanged. Hybrid QEC report snapshots and NaStats snapshots stay on the existing goldens.

Tests

• overlay_target_report_matches_both_previous_chains locks equality with both previous inline orders, including preserved QEC counts.
• overlay_target_report_omits_optional_sections_without_models checks that fidelity still applies when the optional models are absent.

Out of scope

• zoned.rs cost-model selection
• emit-time --emit-resource-report hard-require of error_model in quonc

  

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
quon	Ready	Preview	Oct 1, 2026 5:35am UTC

[cursor]

Approve at d88b934.

overlay_target_report is the one production overlay. pipeline::finish_pipeline and qec_schedule::schedule_expanded both call it after build_resource_report. Neither path still inlines fidelity, atom-loss, error budget, or the diagnostic labels. QEC workload counts (memory_rounds, t_count, tdag_count, ccz_count, magic_state_demand) stay on schedule_expanded. The bare path has no workload to size them from.

The shared order is the old bare-pipeline order. The hybrid path used to apply the budget and fidelity first and the diagnostic labels after. Those writes touch disjoint fields: ErrorBudgetContributions::from_schedule_and_model reads only the schedule counts, fidelity and atom-loss rescan the layers, and the diagnostic setters write only their own fields. overlay_target_report_matches_both_previous_chains locks equality with both previous orders and keeps the QEC counts. overlay_target_report_omits_optional_sections_without_models covers a target with neither optional model, where fidelity still applies.

.expect in the new tests is inside #[cfg(test)], which the library unwrap rule excludes. overlay_target_report itself returns Result. The commit and the PR body say "Issue 468" and do not use a closing keyword. closingIssuesReferences is empty, so merging this does not close 468.

I would merge this. Formal gh pr review --approve was rejected by the GitHub App (addPullRequestReview is not available to this integration), so this comment is the approval.