spec(port): share tree helpers across eight gen_w ports (Closes #9633)

[dmitrii-f-t27]

Closes #9633.

Draft. BLOCKED, not mergeable even with every check green: gen-verilog drops the helper functions that moved behind use (#8769). See "Blocked" below. This PR depends on the shared base listed under Dependencies.

What

Eleven paths change. The first commit, 3518b72, has exactly ten of them (264 insertions, 1094 deletions; historical figures). A later commit adds the eleventh, a data file (below).

• specs/port/scripts/gen_w_tree.t27 (new): TreeBuild, indent, elem_x, elem_y, build_level, build_tree, flat_index and is_ascii, written once, with 7 tests.
• specs/port/scripts/gen_w{698,709,713,716,721,728,729,732}.t27: the local copies of those items are removed and use port::scripts::gen_w_tree; is added. Header constants, generate() and every test are unchanged.
• tools/duplicate_bodies_baseline.txt: two lines removed (flat_index 2, is_ascii 2). Nothing is raised.
• docs/reports/suite_expectations.json (later commit, 665cc91): the one entry for specs/port/scripts/gen_w698.t27 (phase gen-verilog) is removed and max_entries goes from 83 to 82. No other entry, cap or expiry changes, and no broad bless was run.
  This is a conditional candidate, not a proven improvement (see below).

What was observed about the expectation

The corpus ratchet of the first commit failed (run 38764724462, job 116524835401): ledger 83 of 83, UNEXPECTED FAILURES 0, UNEXPECTED PASSES 1: specs/port/scripts/gen_w698.t27 [gen-verilog], "fixed -- remove from ledger".
gen_w698 had spelled indent as a table, an array of strings (W469, #1959); that table is gone. It was an observed unexpected pass with a critical unresolved dependency, not a demonstrated fix: see Blocked. With the entry removed the suite ratchet
is clean locally (UNEXPECTED FAIL 0, UNEXPECTED PASS 0), but that check does not detect lost helper definitions, so a clean result does not prove the code is preserved.

Blocked: #8769

gen-verilog imports enum and struct types through use but not functions. Same base t27c, two outputs: the gen-verilog of the unchanged gen_w709 defines function indent, elem_x, elem_y, build_level, build_tree and more (/tmp/t27-ci-gen-w709-base.v);
the gen-verilog of gen_w698 after its helpers moved behind use defines only function generate, and its test names are only comments (/tmp/t27-ci-gen-w698-after.v). The Zig and Rust targets and t27c test-report pass, but the Verilog output loses code.
Exit code 0 does not prove preservation. This PR must not merge until #8769 is fixed and the helper definitions are present in the generated Verilog; the expectation removal is justified only after that. The root issue is #8769.

Same pattern as specs/port/scripts/gen_w_blocks.t27 for gen_w310..gen_w322. No Rust, no workflow, no generated file, no seal, no exception entry, no file of #9622 or #9630 is touched.

Checks run on this commit

• Shared helpers, reference: 7 of 7 tests, 33 asserts. Native t27b: 7 of 7, 33 asserts (/tmp/t27-ci-tree-tests.log, /tmp/t27-ci-tree-native.log).
• The eight ports, reference: 13 of 13 tests each, 104 of 104 in total, 0 vacuous (/tmp/t27-ci-tree-ports-reference.log). The master artifact also had 13 reference tests for each of them.
• The eight ports, native: blocked ConstDecl, exactly as in the master artifact. This PR does not claim they are native green.
• The ordinary lefthook pre-commit and pre-push hooks both passed.

What this PR does and does not turn green

• It removes the build_level, indent, flat_index and is_ascii violations from python3 tools/dupe_scan.py.
• The full duplicate scanner is still red on this branch without the STEP1/STEP2 work of ci: remove copied helpers from STEP1/STEP2 t27 ports #9631, so the duplicate-bodies check of this PR is red by itself until ci: remove copied helpers from STEP1/STEP2 t27 ports #9631 lands.
• The local tests above ran on the exact committed file contents of the first commit with the same base t27c.
• CI of the first commit (head 3518b72...) at 16:01 UTC: duplicate-bodies, cli-tri and the corpus ratchet failing, others still running. The duplicate and cli failures are the expected dependencies below; the corpus failure was the gen_w698 observation above.
• CI of head 665cc91 at 16:13 UTC: 25 success, 2 failure, 2 in progress; the corpus ratchet succeeds. Those checks cannot see the Verilog loss, so this table is not evidence of completion. Nothing here is called CI green or mergeable.

Dependencies

• ci: remove copied helpers from STEP1/STEP2 t27 ports #9631: STEP1/STEP2 duplicate groups, owned by another lane; needed for the full duplicate-bodies check.
• ci: restore rust_convert native pass (reference CSE hoist regression) #9632: rust_convert CSE hoist, owned by another lane; needed for the t27b ratchet.
• t27b-native ratchet is red on every master run since it landed: ledger 099ac2224 predates 16 improvements and 40+ new specs #9600: the stale t27b ledger; the ratchet cannot be green before it is refreshed.
• cli-tri: retire deleted-workflow test and refresh measured census pins #9634: the CLI build and census failure; the cli-tri check stays red until it lands (owner decision pending there).
• mesh-node port: stop random keypair test from reading undefined primitive results #9635: the mesh-node crypto mismatch, a separate PR.

PRs #9622 and #9630 stay red until these are integrated; this PR alone does not make them green.

Dmitrii Fedorov (@dmitrii-f-t27)

[dmitrii-f-t27]

Critical review finding on this PR: do not merge, even with green checks. gen-verilog loses the moved helper functions (#8769).

Evidence, same base t27c, no code changed:

• gen-verilog of the unchanged specs/port/scripts/gen_w709.t27 defines function indent, elem_x, elem_y, build_level, build_tree and the other helpers.
• gen-verilog of specs/port/scripts/gen_w698.t27 after its helpers moved behind use port::scripts::gen_w_tree; defines only function generate. Every helper definition is gone and the test names survive only as comments.
• gen-verilog imports enum and struct types through use, not functions (root issue gen-verilog: a function brought in by use is called but never defined, and gen-verilog exits 0 #8769).

What this changes:

• The corpus ratchet's UNEXPECTED PASS for specs/port/scripts/gen_w698.t27 [gen-verilog] is a symptom of the loss, not proof that W469 is fixed. Removing that expectation (83 to 82 in docs/reports/suite_expectations.json) is a conditional candidate only, and is justified only after gen-verilog: a function brought in by use is called but never defined, and gen-verilog exits 0 #8769 is fixed and the helper definitions appear in the generated Verilog.
• The clean local suite ratchet (UNEXPECTED FAIL 0, UNEXPECTED PASS 0) and a return code of 0 do not detect lost definitions. A green check table cannot prove preservation.
• Correction of what was claimed before about other targets: t27c test-report (the Zig reference) passes 13 of 13 for each of the eight ports (104 of 104), and the shared helpers pass 7 of 7 natively. The Rust target was NOT tested when the earlier text was written, and "Rust passes" is withdrawn.
  Checked since then: the Rust generated for gen_w728 does contain the helper definitions (/tmp/t27-ci-tree-port-gen.rs), but rustc --test on it exits 1 (/tmp/t27-ci-rust-output-compile.log): is_ascii(s: &'static str) with for c in s is not supported. The Rust generated for the unchanged base gen_w709
  (/tmp/t27-ci-tree-port-gen-base.rs) fails rustc the same way, exit 1 (/tmp/t27-ci-rust-output-base-compile.log). So that is a limitation of the base compiler's Rust output, not caused by this change, and the generated Rust of these specs does not compile either before or after. A working generated Rust is also a goal of the original issues, so it is stated here as not passing.
  The Verilog loss above is specific to this change: the base Verilog has the helpers and the changed one does not.

Status: not mergeable and not CI green. The PR stays draft. Dependencies as listed in the body: #8769 (blocker), #9631, #9632, #9634, #9600. No new request to the owner is made by this comment.

Dmitrii Fedorov (@dmitrii-f-t27)

[gHashTag]

Superseded by #9970. That PR carries this PR's two commits unchanged, with authorship preserved, and adds the STEP1/STEP2 ledger entries with a documented reason (#8769), so duplicate-bodies turns green. Leaving this draft open for its author to close.