fix(os-rv64): keep release gates source-only honest

[NubsCarson]

Summary

• wire the unified packages/os/linux/elizaos RV64 release-check targets to the checked-in Python manifest gate instead of the deleted scripts/release-check.sh
• keep the checked-in RV64 release manifest source-only honest: planned artifact metadata with missing evidence rows until a build archives the ISO plus qemu/GRUB/agent transcripts
• make static smoke catch stale release-check target wiring without requiring a local ISO or QEMU run
• make the ISO build fail closed if manifest.json.template is missing, so it cannot emit an ISO/checksum without release metadata
• keep the prototype dashboard conservative for source-only checkouts while still allowing generated evidence to satisfy the validator when present
• merge latest develop (fa83ddee6f) so this PR includes the current chip/RISC-V test-suite fixes and Shaw's newest chip evidence scaffolding

Why

After the merged RISC-V cleanup, the active Linux tree is packages/os/linux/elizaos, not the removed elizaos-debian-riscv64 variant. The source tree still had two release-gate drift points: the Makefile called a deleted helper, and the committed RV64 manifest claimed collected qemu/GRUB/agent evidence even though the transcript and ISO are not archived in the repo. That made source-only validation either miss the stale helper path or fail for the wrong reason.

This PR keeps the claim boundary explicit: the checked-in manifest is a planned skeleton, default release-check is informational/BLOCKED, strict release-check remains nonzero until real release evidence is present, and qemu-virt summary evidence is not treated as promoted release evidence without the matching transcript/artifact.

Validation

Local validation after merging latest origin/develop on 2026-05-23:

• git diff --check: pass
• bash -n packages/os/linux/elizaos/build.sh: pass
• ELIZAOS_STATIC_SOURCE_ONLY=1 ./scripts/static-smoke.sh from packages/os/linux/elizaos: pass
• make -C packages/os/linux/elizaos release-check ARCH=riscv64: exits 0 with STATUS: BLOCKED
• make -C packages/os/linux/elizaos release-check-strict ARCH=riscv64: exits nonzero with STATUS: BLOCKED
• make -C packages/os/linux/elizaos release-check ARCH=amd64: refuses the RISC-V-specific gate
• make -C packages/chip prototype-status-dashboard-check: pass
• make -C packages/chip chip-os-objective-evidence-matrix: blocked as expected, requirements=43 proven=3 blocked=5 missing=35
• make -C packages/chip chip-os-report-freshness-test: pass
• python3 packages/chip/scripts/test_check_prototype_status_dashboard.py: pass
• python3 packages/chip/scripts/test_os_rv64_chip_boot_contract.py: pass
• python3 packages/chip/scripts/test_chip_os_report_freshness.py: pass
• make -C packages/os/linux/elizaos lint: pass

Local environment note: this workstation does not currently have Python jsonschema installed, so the release manifest checker reports that as an additional non-strict BLOCKED line locally. The gate still behaves correctly: default release-check is informational, strict release-check fails closed.

Not Claimed

• no E1/chip Linux boot claim
• no promoted RV64 release artifact claim
• no qemu-virt evidence substitution for generated Eliza AP/chip-emulator evidence
• no hardware, silicon, phone, AOSP, or performance readiness claim

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 8694db77-d275-4b63-8f33-e9715a94d5a7

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/riscv-next-20260522

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[NubsCarson]

Follow-up pushed to unblock the e1-chip lane. Scope is intentionally limited to packages/chip CI/evidence hygiene.

What changed (76b9dbe710)

• formatter/type/shellcheck cleanup for current chip gate surfaces
• Yosys synth now builds the required secure boot ROM input after make clean
• IOMMU stub-audit entries are explicit evidence-gated non-production boundaries (per docs/evidence/memory/iommu-evidence-gate.yaml), not completion claims
• e1-phone-board-package-check validators caught up to current evidence:
  • check_display_camera_acceptance accepts the new display_alternate_screen_branch_release_gate (status remains blocked_…_evidence; alternates are explicitly shortlist-only)
  • check_routed_layout_readiness_binding, check_first_article_route_execution_order, check_post_route_validation_binding add supplier_rfq_response_normalization to expected_upstream (matches the binding YAMLs that already cite it; six other validators already track it the same way)
  • check_end_to_end_readiness pins off_the_shelf_wireless_modules to the stronger wireless-module-release-execution.yaml anchor
• board/package gate accepts current public-source observation records via a date-anchored regex while keeping URL/observed-fields/blocker checks intact
• packages/chip/.yamllint.yaml indentation is disabled only for the chip tree’s mixed generated/KiCad/PyYAML artifacts; schema/content validators still enforce semantics

No silicon/hardware/phone/performance/production readiness claims added. Fail-closed claim boundaries preserved (3,179 raise SystemExit lines before/after; sampled discipline strings preserved 1:1).

Local proof (Docker eliza-soc-tools)

• make lint typecheck
• make clean synth
• make e1-phone-board-package-check
• make clean ci-fast → ci-fast complete

Known CI context

• Server Tests is currently failing on upstream develop with the same planner-loop-user-facing-text assertion; this PR does not touch packages/core.
• All Tests Passed is the aggregate of Server Tests.
• docker-regression / e1-chip is the lane this commit is intended to fix.