bringup: pin PTO ASLRef functional-model chain

[zhoubot]

Summary

Consolidate the existing PTO 0.58.5 review projection and pin the new ARM-style
ASLRef functional-model chain:

• tools/pto-spec → feat: unify ASLRef functional-model architecture boundary PTO-ISA/pto-spec#188
• tools/asl-model → feat: add ASLRef-backed PTO reference model PTO-ISA/asl-model#2
• tools/SuperScalarModel → LinxISA/SuperScalarModel#428
• exact ASLRef commit 0b6e09066d4186c8a26e02e3bb884bd664d5eb34

The v0.58 model lock binds commits, trees, PTO release manifest, model lock,
model ABI, worker protocol, consumer modeling spec, and the validated
scalar/block/Tile result hashes. Any gitlink or identity mismatch fails closed.

This PR also carries the previously prepared PTO 0.58.5 ISA/docs/AVS projection.
Legacy Shared source variants are retired from the C codec; B.SUBVIEW and
B.ASSEMBLE remain explicitly decode-only in Sail until their carrier-state
semantics are implemented, rather than being silently modeled as no-ops.

Closes #190

Validation

• bash tools/ci/check_repo_layout.sh
• python3 tools/isa/build_golden.py --profile v0.58 --check
• python3 tools/isa/validate_spec.py --profile v0.58
• python3 tools/isa/check_canonical_v058.py --root .
• python3 tools/bringup/check_avs_contract.py --matrix avs/linx_avs_v1_test_matrix.yaml
• python3 tools/bringup/check_sail_model.py --require-parser --require-c-backend
• mkdocs build --strict
• (Optional) python3 tools/bringup/check_avs_profile_closure.py --matrix avs/linx_avs_v1_test_matrix.yaml --status avs/linx_avs_v1_test_matrix_status.json --tier pr
• (Optional) bash tools/regression/run.sh

Additional validation:

• python3 -m unittest discover -s tools/isa -p 'test_*.py': 38/38 PASS
• compiler fixture tests: 16/16 PASS
• python3 tools/isa/check_pto_functional_model_lock.py: PASS
• documentation integrity/projection check: PASS
• Sail parser, directed tests, coverage and C backend: PASS
• SuperScalarModel dependency: 21/21 CTest plus three gfrun ELF cases PASS

Submodules (if touched)

• Submodule SHAs are intentional and minimal
• .gitmodules URLs remain in the LinxISA org
  • PTO architecture and reference-model owners intentionally use the PTO-ISA
    organization; this is validated by the exact component/model lock.
• Any cross-repo change was merged upstream first
  • Deliberately false while this PR is Draft. The three dependency PRs remain
    open and must not be merged automatically.

Draft policy

Do not merge this PR until PTO-ISA/pto-spec#188, PTO-ISA/asl-model#2, and
LinxISA/SuperScalarModel#428 are reviewed and their final heads are repinned.

[zhoubot]

Pinned draft dependencies: PTO-ISA/pto-spec#188, PTO-ISA/asl-model#2, and LinxISA/SuperScalarModel#428. Final gitlinks must be refreshed after review; do not merge automatically.

[zhoubot]

Scope decision and downstream update: the functional-model lock intentionally carries only the three stop-PC result hashes. Host-exit is deferred to PTO-ISA/pto-spec#150. The tools/model codec gate is now handled by LinxISA/linx-model#19 (Issue #18) and pinned at e9f14d9228732f0589f309285cbc8aabaa698353.

[zhoubot]

Downstream QEMU closure added: LinxISA/qemu#79 (Issue #78), pinned at b6fbb378a7614c0f8a8805b27e55b687eb088828. Local evidence: canonical TEPL selector gate passes with 85 operations, opcode metadata audit passes, and QEMU source regressions pass 25/25.

[jiale-wangOwO]

Cross-repository decisions and merge order are now tracked in PTO-ISA/asl-model#7. This pinning PR must remain last and be refreshed only after the upstream model/spec/gfrun heads stabilize.

[zhoubot]

Scalar rollback refresh is staged in #193. It repins merged PTO-SPEC 33637171… and merged asl-model 3e775759…, restores Sail SrcRType/CSEL behavior, imports AGU raw-11 reservation and bit-granular HL.BFI, and passes the hosted Sail parser/C-backend/directed/AVS lane. #193 remains Draft because this base PR still has the private SuperScalarModel checkout blocker and the existing tools/model catalog freshness failure; neither was bypassed.

[jiale-wangOwO]

Current CI diagnosis (2026-09-02):

• model fails in linx_model_check_gen_minst_codec: the lock still points tools/model at bf9d73cff60009e9d40e7f3d240411d96af8cf87 (old catalog). The authoritative regenerated projection is isa: regenerate codec for PTO 0.58.5 linx-model#19, e9f14d9228732f0589f309285cbc8aabaa698353, tree 970ca59157a53349d2ccd75d942526f007df7acc. I prepared dependent PR fix: repin linx-model codec projection in #191 #194, which updates the gitlink and component-lock.v0.58.json together. It passes check_component_lock.py locally.
• guards fails before any lock checks: git submodule update --init tools/SuperScalarModel cannot clone https://github.com/LinxISA/SuperScalarModel.git because that repository is private and the GitHub-hosted GITHUB_TOKEN has no access to sibling private repositories. This is an infrastructure permission issue, not a pin mismatch. The same access must be supplied to the trusted-push workflow, or a trusted runner must have a read credential. Do not make the consumer URL public or replace it with a different repository; the lock must continue to identify the actual consumer.
• guard is the aggregate policy doing its intended fail-closed behavior because the trusted push result is not available.

Recommended order: merge/land linx-model#19, merge #194 into this branch, then configure a read-only organization credential (for example an organization GitHub App/PAT secret) for the private SuperScalarModel checkout in both CI workflows. Re-run guards/model before changing any remaining pins. The branch remains Draft until the exact leaf PRs are stable.

[jiale-wangOwO]

I also prepared dependent PR #195 to fix the CI checkout blocker: #195

It adds explicit SUPERSCALARMODEL_READ_TOKEN handling to both the static and trusted-push workflows and documents the required least-privilege credential. The submodule URL and lock identity remain unchanged; no public mirror or fallback is introduced. A maintainer must add the repository secret (GitHub App/fine-grained token with read-only contents access to LinxISA/SuperScalarModel) before this can make the guards job progress past submodule initialization.

[jiale-wangOwO]

To reduce the number of overlapping dependent PRs, consolidated the two fixes into dependent PR #196: #196

#196 supersedes and closes #194/#195. It atomically repins tools/model to the accepted linx-model#19 codec projection, makes the model job use its immutable authority checkout, and adds the least-privilege private SuperScalarModel checkout configuration to both workflows. No lock identity, fallback, or fail-closed policy is weakened.