fix(spec): re-spell step17.rationale as a joined fragment array, collapsing registry.ts AST depth 977 to 76

[qq9340100]

Fixes#10122

step17.rationale in packages/spec/src/migrations/registry.ts was a single
left-nested + chain of 971 string literals. @typescript-eslint/parser builds
its ESTree AST by recursion, so what it costs is depth, not size — that one
expression gave the whole file a max AST depth of 977, roughly 13x the repo's
runner-up, and put it over V8's default main-thread stack.

Re-spelled as an array of the same literals joined with .join('').

Before / after, measured on this branch

	before (origin/main @ 6439f8b42)	after (192d3c58e)
registry.ts max AST depth	977	76
step17.rationale initializer subtree depth	971	4
step17.rationale initializer kind	BinaryExpression	CallExpression
longest + chain in the file	971 operands, L397–1367	70 operands, L4968–5037 (step18.rationale, untouched)
min --stack-size for eslint to parse the file	1099 KB	104 KB
parses at V8's default stack (984 KB, no flag)?	no — cliffs	yes

Depth convention: SourceFile = depth 1, walked with ts.forEachChild. The card
measured 976 / 970 operands at 1800ffac2; the file has since grown by one
fragment, and the numbers above are re-measured on current main.

--stack-size method: binary search over
node --stack-size=N node_modules/eslint/bin/eslint.js PATH --no-inline-config,
1 KB resolution, cliff detected by the literal Maximum call stack size exceeded
string. Node v22.22.2. The 1099 KB "before" figure sits close to the 1085 KB
recorded in eslint.config.mjs — the difference is the one fragment appended
since, at the ~1.10 KB/fragment cost that comment measured.

After the change the file needs 9.5x less stack than V8 gives it by default,
where before it was ~12% over budget and could only parse on borrowed headroom.

The acceptance criterion: the string value is unchanged

The 971 fragments are carried over byte-for-byte — same quoting (941 single,
30 double), same escapes, one fragment per line. Only the + separators became
,, plus the [ / ].join('') wrapper.

Proven three ways, all mechanical:

1. Source-level, pre-commit. Both spellings evaluated and compared with strict
   === plus sha256 — 79328 chars, 2099de977992696945f00d772042b27c3be79135d2a4d44109433973f979de1f, identical.
2. Independent re-check on the committed artifacts. The initializer was
   re-extracted from origin/main's file and from this branch's file via the
   TypeScript AST and re-evaluated: same length, same sha256, strict === : true.
3. At runtime, out of the built package.import('packages/spec/dist/index.mjs')
   then MIGRATIONS_BY_MAJOR[17].rationale → 79328 chars, sha256
   2099de97…de1f. The value a consumer actually receives is the same string.

A gap check also ran before the rewrite: every one of the 970 gaps between
consecutive operands was asserted to be whitespace plus a single +, so no comment
or non-literal operand could be silently dropped. All 971 operands are plain
StringLiteral (0 other kinds).

Nothing else moved

The rewrite replaced exactly the byte range [19762, 108025) — the initializer —
and the surrounding bytes were asserted unchanged on both sides. git diff is a
single hunk, @@ -397,971 +397,973 @@.

• The os-generated marker regions are untouched. step17's markers are at L1432 and
  L4926; the chain ended at L1367, entirely outside them.
• Other steps' rationales, conversionIds, RETIRED_KEYS_BY_MAJOR and
  RETIRED_DEFS_BY_MAJOR are byte-identical.
• The custom merge driver's marker structure is preserved — pnpm check:merge-driver
  is green.

A template literal was considered and rejected: the content contains backticks
(`execute`, `target` …) and ${-hostile text, so it would have required
escaping inside every fragment — which is exactly the byte-level churn this change
is trying not to introduce. The .join('') spelling keeps each literal verbatim.

Gates

Union re-derived from the actual diff with node scripts/pm/dispatch-gates.mjs
(no paths — it reads the change set from the merge base itself) and run at
192d3c58e, the final commit.

The two that this card is about, both green:

=== pnpm lint :: EXIT=0
=== check:slot-lookup :: EXIT=0

pnpm lint is the whole-repo run — the one that was returning
registry.ts 0:0 error Parsing error: Maximum call stack size exceeded.

Also green at 192d3c58e: check:merge-driver, check:changeset-gate-self-tests,
check:cross-package-test-inputs, check:spec-parsed-alias,
check:type-source-resolution, check-adr-0087-registration,
check-cross-package-test-inputs, check-affected-docs, and the spec-liveness
family (check:empty-state, check:liveness, check:strictness-ledger,
check:variant-docs), plus pnpm --filter @objectstack/spec typecheck and
pnpm --filter @objectstack/spec test.

check:doc-formula-expressions red on first pass with
ERR_MODULE_NOT_FOUND … @objectstack/formula/dist/index.mjs — a missing build
artifact in a fresh worktree, not this diff. After
pnpm --filter '@objectstack/formula...' build it is green:

✓ check:doc-formula-expressions: 22 record-scoped formula example(s) across 407 files / 1416 TS blocks judged clean

One declared narrowing:node scripts/check-dev-prereqs.mjs was not brought
green locally. Its own verdict names the cause and it is not this diff —
✗ The workspace is not built — 1 unmet precondition, not a list of problems. / 66 of 67 workspace packages declare an entry point under dist/ that is not on disk. Clearing it needs a full pnpm build of the workspace, which would hold
the shared verify lock for a long time on a container running several agents; CI
builds the workspace and runs this gate anyway.

No changeset — deliberate, and here is what dist actually contains

registry.ts does ship in the published @objectstack/spec, and its emitted
bytes do change: dist/index.js / dist/index.mjs are not minified, so they
carry the source spelling verbatim — rationale: [ … ].join('') where they
previously carried the + chain.

What does not change is anything a consumer can observe: the exported string is
the same 79328 characters with the same sha256 (verified out of the built artifact,
above), no export was added or removed, and no type changed. Per AGENTS.md — "Add a
changeset for feature work … Pure bug fixes do not require a changeset" — this is a
repo-tooling fix with no functional delta, so it carries the skip-changeset label
instead of declaring a release of its own.

For the closers of the duplicate reports

#10071 and #10121 report the same signature and were already triaged as
duplicates of the card this PR addresses; they remain open or closed at their
own triagers' discretion — nothing here changes their state. The evidence
recorded on the card at 18:45Z today — PR #10396
dequeued from the merge queue with reason: CI_FAILURE while its own head was
green (24 checks, 0 failing, mergeable_state: clean), the queue build failing on
packages/spec/src/migrations/registry.ts:0:0 — Parsing error: Maximum call stack size exceeded for a file not in that PR's diff — is the collateral-eviction mode
this removes. #10030 raised the parser's headroom to --stack-size=4000; this
lowers the floor instead, from 1099 KB to 104 KB.

Two things deliberately left alone

1. eslint.config.mjs still says --stack-size=4000, and its comment block
   (L31–91) still describes the pre-fix state. Whether that headroom should now be
   lowered or kept as margin is a real decision with a measurement attached, not a
   mechanical follow-on, and this PR's file surface is one file by claim. Filed
   separately as [finding] Once the registry.ts depth fix lands, eslint.config.mjs's --stack-size=4000 and its 60-line rationale describe a state that no longer exists #10448 (unassigned), which lays out both dispositions.
2. step18.rationale is a 70-operand chain (file depth 76). That is the same
   authoring habit at harmless scale — the repo's runner-up across all 4659 linted
   files was already 71 — and the card explicitly scopes the smaller same-shape
   chains out.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/spec, touching 3 documentable anchor(s).

5 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/automation/flows.mdx(via /api/v1/apps (route))
• content/docs/automation/hook-bodies.mdx(via crypto.hash (literal))
• content/docs/getting-started/quick-reference.mdx(via /api/v1/apps (route))
• content/docs/protocol/kernel/http-protocol.mdx(via /api/v1/apps (route))
• content/docs/protocol/objectui/widget-contract.mdx(via my_picker (literal))

What this run could not see

• 1 name(s) were too generic to anchor anything (single lowercase words)

Coarse fallback — 116 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 91f303cb4bfa47bc3526633f5275d0aafc5aad24 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 95d8adee5003c8b3faaf7588aeacc0a9f474214e — the merge of head 192d3c58e71f5fcfdb90a22cecb2dafba4893d3e into base 91f303cb4bfa47bc3526633f5275d0aafc5aad24, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 95d8adee5003c8b3faaf7588aeacc0a9f474214e && git checkout 95d8adee5003c8b3faaf7588aeacc0a9f474214e
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 91f303cb4bfa47bc3526633f5275d0aafc5aad24 192d3c58e71f5fcfdb90a22cecb2dafba4893d3e && git checkout -B drift-repro 91f303cb4bfa47bc3526633f5275d0aafc5aad24 && git merge --no-ff 192d3c58e71f5fcfdb90a22cecb2dafba4893d3e
node scripts/docs-audit/affected-docs.mjs --json 91f303cb4bfa47bc3526633f5275d0aafc5aad24

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 91f303cb4bfa47bc3526633f5275d0aafc5aad24 → pass the list as
args.docs, on the commit named under Which tree this was computed on.