mirror of
https://github.com/morluto/rea.git
synced 2026-11-16 11:43:20 +00:00
* fix(native): keep unreported value-flow truncation and archive decode coverage unknown * test(verify): count multiline optional-fact defaults and pin the guard's exclusions * fix(domain): keep unknown counts and costs unknown in comparison and coverage summaries (#11325) * fix(domain): keep unknown counts and costs unknown in comparison and coverage summaries * refactor(verify): exempt legitimate optional-fact defaults with a reasoned marker (#11355) * refactor(verify): exempt legitimate optional-fact defaults with a reasoned marker The optional-fact-default guard counted request defaults and absence that is modelled elsewhere as hidden unknowns. A comment `// optional-fact-default: <reason>` on the default's line or the line before now exempts it; the guard reads unstripped source for the marker and ignores matches inside comment lines. Marked: the two evidence paging offsets and four Swift literal reads whose absence swiftLiteralCoverage reports. Removed instead of marked: jsdiff 9 types Change.count as required, so its two `?? 0` defaults were dead; the evidence link traversal now carries retained records, so it no longer re-looks them up behind `?? []`. Baseline 31 -> 22. * fix(verify): blank comment spans instead of skipping comment-led lines in the optional-fact guard Skipping a match whose line starts like a comment hid live code after a leading block comment and counted defaults inside a block comment without leading stars. The counter now blanks comment text (keeping offsets and line breaks, skipping string literals) and matches the blanked code, while exemption markers are still read from the original lines. * refactor(domain): mark optional-fact defaults whose absence is not an observation (#11389) * refactor(domain): mark optional-fact defaults whose absence is not an observation Per-site review of the remaining optional-fact-default entries. Each marked default is a caller option (trace ignore_fields, environment rules), a list whose producer omits the empty case (unknown_size_frames, the schema's min-1 representation_node_ids), caveat text whose completeness another field carries (call diagnostics, unresolved calls, semantic resource limits), or absence that already turns the result unknown (Wasm interface inventory, WebMCP discovery, managed ExportedType rows, legacy integrity mismatch IDs). Left counted: webCaptureDiff redirects (2; the CDP producer always records them, so absence is a legacy record that should compare as unknown, not as no redirects) and HopperCallerReconciliation (2; index-miss semantics not yet verified). Baseline 22 -> 4. * fix(browser): compare redirects as unknown when a capture predates redirect recording (#11402) * fix(browser): compare redirects as unknown when a capture predates redirect recording The CDP producer records every request's redirect chain, empty when there was none; a record without one predates redirect recording (#986). Comparing it against a current capture reported the network dimension as unchanged, treating the missing chain as "no redirects". Redirects are now compared only when both captures recorded them, and the network dimension is otherwise unknown with a reason; proven membership changes are still reported. * refactor(hopper): interpret a caller-index miss once at the lookup (#11403) The forward index holds observations from the examined procedures only, and relationship coverage stays unknown with the scan facts as its reason. A miss therefore adds no callers without showing there are none; the lookup now says so once instead of four optional reads defaulting at each site. The optional-fact-default baseline is now empty. * docs(javascript): state why an omitted representation list adds no nodes * fix(verify): count optional-fact defaults from parsed TypeScript, not comment-blanked text The guard blanked comments with a hand-written scanner and matched a regular expression on the rest, so a default quoted in a string counted and a regular expression holding /* hid the code after it. It now walks the TypeScript syntax tree for property reads defaulted with ?? 0, ?? [] or ?? false, and reads exemption markers only from comment trivia. * fix(verify): tokenize optional-fact defaults without a compiler and count chained defaults * fix(verify): match whole optional-fact default expressions across calls and operators * fix(verify): count optional-fact defaults in template substitutions and after postfix updates, and tie markers to their own expression * refactor(verify): exempt generation and sample caps from the constant-maximum-slice guard by reason marker (#11454) * refactor(verify): exempt generation and sample caps from the constant-maximum-slice guard by reason marker * fix(verify): read reason markers only from real line comments * fix(verify): read reason markers from the shared source tokenizer * fix(verify): read a regular expression after a control head as a statement start The shared tokenizer treated every closing parenthesis as the end of an operand, so in `if (ready) /[/*]/.test(v)` it read `/[` as division and `/*` inside the regex as a block comment, hiding the rest of the file from the marker-aware guards. Parentheses are now tracked: the `)` closing an if/while/for/with head allows a regular expression, while a call's `)` still means division. * fix(verify): keep a keyword-named member call out of control-head detection `policy.if(ready) / (…)` was read as an if-statement head, so the division was taken for a regular expression and the denominator's optional-fact default was not counted. A control keyword after `.` or `?.` is a member call. * refactor(verify): exempt budget-arithmetic cursors and tallies by a not-an-allowance marker (#11493) * refactor(verify): exempt budget-arithmetic cursors and tallies by a not-an-allowance marker The budget-arithmetic guard cannot tell a scan cursor, an observed total, a string being assembled or a tail buffer's retained length from an allowance when each is compared with a limit. A field declaration marked `// not-an-allowance: <reason>` (read through the shared reason-marker scan) is no longer counted. The four remaining sites carry a marker, so the baseline is empty and removed. * refactor(verify): exempt detail levels, lifecycle states and domain vocabularies from the completeness-enum guard by reason marker (#11510) A zod enum containing "complete" is counted as a local coverage vocabulary. Eleven are not coverage: a caller-chosen detail level, how far a command or session ran, and format states (GPU decode depth, WASM name section, SQLite column details). They now carry `// detail-level:`, `// lifecycle-state:` or `// domain-vocabulary:` with a reason, read through the shared reason-marker scan (which gains an optional predicate for the matches a guard counts). local-completeness-enum: 22 -> 11 on this branch. * fix(verify): bind a not-an-allowance marker to its own class field Exemptions were a file-wide set of names, so a marked tally in one class hid an unmarked allowance with the same name in another, and a marked object property could exempt a class field. Markers now exempt only a member declaration (at class member depth) of the top-level class that contains it, and the named- allowance count is taken per class so the same scoping applies to it. * fix(artifacts): keep main's resources import in the Deno standalone reader after merging main * fix(verify): start an expression after a template substitution and a non-member keyword, and accept any marker reason The shared tokenizer read the `/` after `${` as division, so a regular expression holding `/*` in a substitution hid the rest of the file from every tokenizer-based guard, and it treated property names such as `row.return` as keywords. Tokens now record whether template text opens a substitution and whether a name is an expression keyword (not after `.` or `?.`), and the regular-expression rule and the optional-fact-default operand rule read those facts. The optional-fact-default guard reads markers through reason-marker's isReasonMarker, so any non-empty reason exempts, as for the other reason markers. * fix(verify): blank comments with the shared tokenizer before matching architecture guards (#11573) * fix(verify): blank comments with the shared tokenizer before matching architecture guards The guards stripped comments with a regular expression that took a `/*` inside a string, template or regular expression for a comment opener and deleted code up to the next `*/`. src/flutter/FlutterBuildProvider.ts's "lib/*/libapp.so" text hid four --pool-policy remediation strings from cli-option-text-outside-cli. blankComments in source-tokens.mjs now owns comment removal for the architecture guards and the reason-marker scan: it blanks comment tokens in place (newlines kept) and keeps strings, templates and regular expressions. The scan behind it retains only the last tokens it reads, and whitespace runs are skipped at once. The cli-option-text-outside-cli baseline records FlutterBuildProvider.ts at its true count of 4; no other baseline changed. Fixes #11477 * fix(verify): read a regular expression after a statement block and keep comment offsets for wide characters The tokenizer now records whether each `{` opens a statement block or an object literal (from the token before it) and marks the `}` that closes a block, after which a statement, and so a regular expression, can start; an object literal's `}` still ends an operand. blankComments replaces each run of comment text with one space per UTF-16 unit, so a surrogate pair no longer shifts later offsets. * chore(verify): drop the Flutter CLI-text baseline entry now that main names no option there
977 lines
41 KiB
JavaScript
977 lines
41 KiB
JavaScript
import { readFile, readdir, rm, writeFile } from "node:fs/promises";
|
|
import { join, relative, sep } from "node:path";
|
|
import { fileURLToPath } from "node:url";
|
|
import { recordPredicateCounts } from "./record-predicate-guard.mjs";
|
|
import {
|
|
compareCountedBaseline,
|
|
recordLoweredBaseline,
|
|
reconcileBaselineFile,
|
|
} from "./counted-baseline.mjs";
|
|
import { localCompletenessEnumCount } from "./completeness-enum-guard.mjs";
|
|
import { countKeyedFirstMatches } from "./keyed-first-match.mjs";
|
|
import { budgetArithmeticCount } from "./budget-arithmetic-guard.mjs";
|
|
import { localLimitErrorClassCount } from "./limit-class-guard.mjs";
|
|
import {
|
|
isComparisonOrCountFile,
|
|
optionalFactDefaultCount,
|
|
} from "./optional-fact-default-guard.mjs";
|
|
import { unmarkedMatchCount } from "./reason-marker.mjs";
|
|
import { blankComments } from "./source-tokens.mjs";
|
|
import {
|
|
cliTextGuardSelects,
|
|
cliTextOutsideCliCount,
|
|
} from "./cli-text-guard.mjs";
|
|
import {
|
|
TOOL_REGISTRATION_OWNERS,
|
|
toolRegistrationCount,
|
|
} from "./tool-registration-guard.mjs";
|
|
|
|
// Architecture guards for the strict-identity / best-effort-inference split.
|
|
// - Composite dedup keys must not NUL-join variable-length or free-form
|
|
// MIDDLE fields: ["a\0b"] collapses with ["a", "b"] and drops a finding.
|
|
// The text check below only flags one spelling; it is not a complete audit.
|
|
// - Transparent wrappers unwrap in exactly one place
|
|
// (javascriptAstValues.unwrapJavaScriptExpression); direct isTS* unwrapping
|
|
// elsewhere drifts (the `satisfies` class of bug).
|
|
|
|
const root = fileURLToPath(new URL("../../", import.meta.url));
|
|
|
|
const listFiles = async (dir) => {
|
|
const output = [];
|
|
for (const entry of await readdir(dir, { withFileTypes: true })) {
|
|
const path = join(dir, entry.name);
|
|
if (entry.isDirectory()) output.push(...(await listFiles(path)));
|
|
else if (
|
|
entry.name.endsWith(".ts") &&
|
|
!entry.name.endsWith(".test.ts") &&
|
|
!entry.name.endsWith(".spec.ts") &&
|
|
!entry.name.endsWith(".fixture.ts")
|
|
)
|
|
output.push(path);
|
|
}
|
|
return output;
|
|
};
|
|
|
|
const sources = await listFiles(join(root, "src"));
|
|
const files = await Promise.all(
|
|
sources.map(async (path) => {
|
|
const source = await readFile(path, "utf8");
|
|
return {
|
|
path: relative(root, path).split(sep).join("/"),
|
|
content: blankComments(source),
|
|
source,
|
|
};
|
|
}),
|
|
);
|
|
const failures = [];
|
|
const check = (label, selected, pattern, allow = []) => {
|
|
const offenders = selected.filter(
|
|
(file) => pattern.test(file.content) && !allow.includes(file.path),
|
|
);
|
|
if (offenders.length > 0)
|
|
failures.push(
|
|
`${label}:\n${offenders.map((file) => ` ${file.path}`).join("\n")}`,
|
|
);
|
|
};
|
|
|
|
// Completeness is a claim its producer establishes from examined input. A
|
|
// fallback or default of `true` turns an unstated claim into a complete one.
|
|
check(
|
|
"completeness defaulted to true (leave it unknown or require the producer to state it)",
|
|
files,
|
|
/\b(?:complete|exhaustive|\w+Complete|\w+Exhaustive|\w+_complete|\w+_exhaustive)\w*\s*(?:\?\?\s*true\b|(?::\s*boolean\s*)?=\s*true\s*[,)])/u,
|
|
);
|
|
|
|
// Prerequisite inspection lifetime is an application contract shared by adapters.
|
|
check(
|
|
"provider-owned inspection lifetime (use src/ports/ProviderInspectionScope)",
|
|
files,
|
|
/providers\/ProviderInspectionScope\.js/,
|
|
);
|
|
|
|
check("NUL-joined list key (use compositeKey)", files, /\.join\(["']\\0["']\)/);
|
|
|
|
// Pushed effect invalidation (docs/adr/javascript/pushed-effect-invalidation.md).
|
|
// Opaque-effect checks are private to javascriptSemanticEffectLog.ts, so the
|
|
// compiler keeps readers on its questions, and only the log and its record
|
|
// module read the recorded sites. A method read on an intrinsic
|
|
// instance asks one owner, which checks the method, the instance's prototype
|
|
// link and the intrinsic prototypes.
|
|
check(
|
|
"reader-side read of recorded opaque effects (ask javascriptSemanticEffectLog.ts or the javascriptSemanticEffectRecord.ts accessors)",
|
|
files,
|
|
/\.opaqueEffects\b/u,
|
|
[
|
|
"src/domain/javascript/javascriptSemanticEffectLog.ts",
|
|
"src/domain/javascript/javascriptSemanticEffectRecord.ts",
|
|
],
|
|
);
|
|
|
|
check(
|
|
"reader-side prototype-link member check (use isIntrinsicInstanceMethodCall)",
|
|
files,
|
|
/mayObserve\([^;]*\["__proto__"\]/u,
|
|
["src/domain/javascript/javascriptSemanticIntrinsicInstances.ts"],
|
|
);
|
|
|
|
check(
|
|
"direct transparent-wrapper unwrap outside javascriptAstValues (use unwrapJavaScriptExpression)",
|
|
files,
|
|
/isTSAsExpression|isTSSatisfiesExpression|isTSNonNullExpression|isTSTypeAssertion|isTSInstantiationExpression|isParenthesizedExpression/,
|
|
["src/domain/javascript/javascriptAstValues.ts"],
|
|
);
|
|
|
|
// A producer that pins both the soft and hard ceiling of every resource takes
|
|
// the vector from one owner. prlimitCeilingArguments is the command-less prefix
|
|
// of prlimitArgumentVector; both live in src/process/LinuxResourceLimits.ts.
|
|
// Re-audited 2026-10-22 with util-linux 2.39.3: a bare value sets both soft and
|
|
// hard, the same as soft:hard when the numbers are equal. `soft:` leaves the
|
|
// inherited hard ceiling unchanged.
|
|
// FirmwareCommand and WakaruCommand call prlimitArgumentVector and keep their
|
|
// own file-size and CPU budgets (120 CPU seconds; firmware file size is the
|
|
// call output budget or the 8 MiB report allowance; Wakaru file size is the
|
|
// 128 MiB output allowance). Their 1 GiB address space is
|
|
// BOUNDED_PRODUCER_LIMITS.address_space_bytes. PcapNativeBound calls
|
|
// prlimitCeilingArguments because PcapCommand appends the producer command.
|
|
// Its CPU ceiling stays 300 seconds and it has no resident-input budget;
|
|
// address space and file size use those shared fields.
|
|
// selectEvmWorkerLimits stays a separate selector: Result failures, unit
|
|
// checks, and a refusal when soft exceeds hard. It must not call
|
|
// clampLinuxResourceLimits.
|
|
check(
|
|
"inline paired prlimit ceilings (use prlimitArgumentVector from src/process/LinuxResourceLimits)",
|
|
files,
|
|
/--core=0:0/,
|
|
["src/process/LinuxResourceLimits.ts"],
|
|
);
|
|
|
|
check(
|
|
"bare prlimit value (a bare value sets both ceilings; use prlimitArgumentVector)",
|
|
files,
|
|
/--core=0(?!:)/,
|
|
);
|
|
|
|
// 2026-10-22: soft-only `resource:` options retain the inherited hard ceiling.
|
|
// EVMole budgets differ (3 GiB address space, 16 MiB file, 30 CPU seconds), so
|
|
// this lane must not call prlimitArgumentVector.
|
|
check(
|
|
"soft-only prlimit ceilings (paired producers use prlimitArgumentVector)",
|
|
files,
|
|
/--core=0:(?!0)/,
|
|
["src/evm/EvmoleInterfaceProvider.ts"],
|
|
);
|
|
|
|
// Resident input is the input_bytes field of BOUNDED_PRODUCER_LIMITS. CamelCase
|
|
// inputBytes values with a different policy (an overridable pwntools admission,
|
|
// historical capture intake) are not this budget.
|
|
check(
|
|
"inline resident-input budget (use BOUNDED_PRODUCER_LIMITS from src/domain/producerResourceLimits)",
|
|
files,
|
|
/input_bytes:\s*32 \* 1024 \* 1024/,
|
|
["src/domain/producerResourceLimits.ts"],
|
|
);
|
|
|
|
// Fixed-arity backtick keys audited 2026-10-10: every prefix field is NUL-free
|
|
// (hex digests, file paths, enums, numbers, or compositeKey JSON which escapes
|
|
// NUL), with free-form source strings only in final position. New backtick-\0
|
|
// keys must either meet that shape (and extend this list with justification)
|
|
// or use compositeKey.
|
|
check(
|
|
"new backtick-NUL composite key (audit against the rule above)",
|
|
files,
|
|
/`[^`\n]*\\0/,
|
|
[
|
|
// Bundle fingerprint: fixed packaged paths (no NUL/newline) and 64 hex
|
|
// digest characters per record, never guide contents. Preserve v1 locators.
|
|
"src/application/GuidanceAccess.ts",
|
|
"src/application/javascript/ElectronBoundaryAnalysis.ts",
|
|
"src/application/javascript/ElectronBoundaryGraphContext.ts",
|
|
"src/browser/CdpCaptureCompleteness.ts",
|
|
"src/browser/CdpCaptureEventHandlers.ts",
|
|
"src/browser/CdpPageCapture.ts",
|
|
"src/browser/WebSourceMapFetcher.ts",
|
|
"src/domain/javascript/electronStaticAnalysisBrowser.ts",
|
|
"src/domain/javascript/electronStaticAnalysisIpc.ts",
|
|
"src/domain/javascript/javascriptApplicationChangeGraph.ts",
|
|
"src/domain/javascript/javascriptApplicationVersionItems.ts",
|
|
"src/domain/javascript/javascriptApplicationVersionKeys.ts",
|
|
"src/domain/javascript/javascriptExportShapeSelection.ts",
|
|
"src/domain/javascript/javascriptFeatureSeed.ts",
|
|
"src/domain/javascript/javascriptFeatureTraversal.ts",
|
|
"src/domain/javascript/javascriptNativeHandoff.ts",
|
|
"src/domain/javascript/javascriptRuntimeReconciliationMatching.ts",
|
|
"src/domain/javascript/javascriptRuntimeReconciliationSchemas.ts",
|
|
"src/domain/javascript/javascriptStaticAnalysisBundler.ts",
|
|
"src/domain/javascript/javascriptStaticAnalysisCalls.ts",
|
|
"src/domain/javascript/javascriptStaticAnalysisFindings.ts",
|
|
"src/domain/localPath.ts",
|
|
"src/domain/process/processTraceEvaluation.ts",
|
|
"src/domain/process/processTraceSpecification.ts",
|
|
"src/domain/unicodeCodePointOrder.ts",
|
|
"src/domain/webBundleAnalyzerInspection.ts",
|
|
],
|
|
);
|
|
|
|
// "Non-null object other than an array" has one owner: src/domain/isRecord.ts.
|
|
// A local predicate with that body, under any name, is a copy. The listed files
|
|
// still hold one and may only be removed from the list; a listed file that no
|
|
// longer matches fails so the list shrinks with the code.
|
|
const RECORD_PREDICATE_COPIES = [
|
|
// Exported under a module-specific name; importers still use that name.
|
|
"src/browser/history/CaptureRedaction.ts",
|
|
"src/contracts/compactJsonSchema.ts",
|
|
"src/domain/analysisView/environmentStateRules.ts",
|
|
// Owned by the advertised-schema workstream.
|
|
"src/contracts/inputSchemaPresentation.ts",
|
|
// Narrows a protocol-specific parameter type, not unknown.
|
|
"src/browser/execution/CdpRuntimeInitiator.ts",
|
|
// Named `object`, which also spells the typeof result in these files.
|
|
"src/pcap/PcapDnsWire.ts",
|
|
"src/pcap/PcapOutput.ts",
|
|
];
|
|
const mismatchedRecordCopies = files.filter(({ path, content }) => {
|
|
if (path === "src/domain/isRecord.ts") return false;
|
|
const { named, bodies } = recordPredicateCounts(content);
|
|
return (
|
|
named !== 0 || bodies !== (RECORD_PREDICATE_COPIES.includes(path) ? 1 : 0)
|
|
);
|
|
});
|
|
const pathsWithRecordCopies = new Set(files.map(({ path }) => path));
|
|
const missingRecordCopies = RECORD_PREDICATE_COPIES.filter(
|
|
(path) => !pathsWithRecordCopies.has(path),
|
|
);
|
|
if (mismatchedRecordCopies.length > 0)
|
|
failures.push(
|
|
`local non-array record predicate count changed (import isRecord and update the baseline):\n${mismatchedRecordCopies.map(({ path }) => ` ${path}`).join("\n")}`,
|
|
);
|
|
if (missingRecordCopies.length > 0)
|
|
failures.push(
|
|
`record predicate baseline references missing files; remove those entries:\n${missingRecordCopies.map((path) => ` ${path}`).join("\n")}`,
|
|
);
|
|
|
|
// Node clamps a timer delay above the signed 32-bit bound to 1 ms. Durations
|
|
// use MAX_NODE_TIMER_DELAY_MS or timerDurationMsSchema from their one owner.
|
|
// The allowed files bound signed 32-bit integers that are not timer delays.
|
|
check(
|
|
"signed 32-bit bound literal (use MAX_NODE_TIMER_DELAY_MS or timerDurationMsSchema from src/domain/analysisResourceLimits for durations)",
|
|
files,
|
|
/\b2_?147_?483_?647\b/u,
|
|
[
|
|
"src/domain/analysisResourceLimits.ts",
|
|
// JADX MCP `limit` argument: a Java int, not a duration.
|
|
"src/android/JadxAnalysis.ts",
|
|
// JVM `-XX:ActiveProcessorCount`: a Java int processor count.
|
|
"src/android/JadxConfiguration.ts",
|
|
// TShark `-c` packet count: a C int, not a duration.
|
|
"src/domain/captures/packetCapture.ts",
|
|
// DEX static `int` value range.
|
|
"src/domain/android/androidAnalysis.ts",
|
|
// Signed 32-bit bitmask operand range.
|
|
"src/domain/javascript/javascriptSemanticResources.ts",
|
|
// Recorded crash pid, signal number and code fields.
|
|
"src/domain/native/recordedCrash.ts",
|
|
// Java int parsing for Ghidra launch properties.
|
|
"src/ghidra/GhidraInstallationPolicy.ts",
|
|
],
|
|
);
|
|
|
|
// Native inventory results parse once against their named fact contract
|
|
// (docs/adr/native/typed-operation-facts.md). Hopper's reply table and the
|
|
// advertised output schemas own their own protocol and catalog contracts.
|
|
check(
|
|
"native inventory rows parsed outside their fact contract (use executeFact or a contract from src/domain/native/nativeOperationFacts)",
|
|
files,
|
|
/\bnativeInventoryRowsSchemas\b|z\.array\(\s*addressedValueSchema\b|\baddressedValueSchema\s*\.\s*(?:safeParse|parse|pick)\b/,
|
|
[
|
|
"src/domain/native/nativeInventory.ts",
|
|
"src/domain/native/nativeOperationFacts.ts",
|
|
"src/hopper/HopperReplyContracts.ts",
|
|
"src/contracts/toolOutputSchemaGroups.ts",
|
|
],
|
|
);
|
|
|
|
// Mach-O load-command codes have one table; a local copy drifts from it.
|
|
check(
|
|
"local Mach-O load-command code (import LC from src/domain/apple/machoLoadCommandCodes)",
|
|
files,
|
|
/\bLC_[A-Z0-9_]+\s*=\s*(?:0x[0-9a-f]+|\d+)\b/iu,
|
|
["src/domain/apple/machoLoadCommandCodes.ts"],
|
|
);
|
|
|
|
// Counted baselines record the remaining copies of a pattern that has one
|
|
// owner. A file may not gain copies and a baseline may not grow; when copies
|
|
// are removed, `--shrink-baselines` lowers the recorded counts (and never
|
|
// raises them) so the next removal cannot be undone silently. Each guard's
|
|
// baseline is its own file, scripts/verify/baselines/<key>.json, so changes
|
|
// to different guards never edit the same file.
|
|
const baselineDirectory = join(root, "scripts/verify/baselines");
|
|
const baselineTexts = new Map();
|
|
for (const name of await readdir(baselineDirectory)) {
|
|
if (!name.endsWith(".json")) {
|
|
failures.push(
|
|
`${relative(root, join(baselineDirectory, name))} is not a guard baseline; only <guard-key>.json files belong in ${relative(root, baselineDirectory)}.`,
|
|
);
|
|
continue;
|
|
}
|
|
baselineTexts.set(
|
|
name.slice(0, -".json".length),
|
|
await readFile(join(baselineDirectory, name), "utf8"),
|
|
);
|
|
}
|
|
const baselines = Object.fromEntries(
|
|
[...baselineTexts].map(([key, text]) => [key, JSON.parse(text)]),
|
|
);
|
|
const checkedBaselineKeys = new Set();
|
|
const shrinkBaselines = process.argv.includes("--shrink-baselines");
|
|
let baselinesShrunk = false;
|
|
const checkCounted = (label, key, selected, pattern, owners) => {
|
|
checkedBaselineKeys.add(key);
|
|
const counts = new Map();
|
|
for (const file of selected) {
|
|
if (owners.includes(file.path)) continue;
|
|
const count =
|
|
typeof pattern === "function"
|
|
? pattern(file.content, file)
|
|
: (file.content.match(pattern)?.length ?? 0);
|
|
if (count > 0) counts.set(file.path, count);
|
|
}
|
|
const compared = compareCountedBaseline(
|
|
counts,
|
|
baselines[key] ?? {},
|
|
shrinkBaselines,
|
|
);
|
|
const lines = (differences) =>
|
|
differences
|
|
.map(
|
|
({ path, count, allowed }) =>
|
|
` ${path}: ${count} (baseline ${allowed})`,
|
|
)
|
|
.join("\n");
|
|
if (compared.grown.length > 0)
|
|
failures.push(
|
|
`${label}; new copies are not allowed:\n${lines(compared.grown)}`,
|
|
);
|
|
if (recordLoweredBaseline(baselines, key, compared)) baselinesShrunk = true;
|
|
if (compared.stale.length > 0)
|
|
failures.push(
|
|
`${label}; copies were removed, so lower the baseline with \`node scripts/verify/architecture-guards.mjs --shrink-baselines\`:\n${lines(compared.stale)}`,
|
|
);
|
|
};
|
|
|
|
// Comparison, coverage and count code that reads an optional parsed fact with
|
|
// `?? 0`, `?? []` or `?? false` turns "not reported" into an observed zero,
|
|
// empty list or no. Keep the fact unknown (null with a limitation) instead.
|
|
checkCounted(
|
|
"optional fact defaulted to 0, [] or false in comparison or count code (keep it unknown)",
|
|
"optional-fact-default",
|
|
files.filter(({ path }) => isComparisonOrCountFile(path)),
|
|
// Exemption markers are comments, so this guard reads the unstripped source.
|
|
(_content, { source }) => optionalFactDefaultCount(source),
|
|
[],
|
|
);
|
|
|
|
// A provider failure's captured output comes from one owner,
|
|
// capturedProviderOutput in src/process; local copies drift on byte counts
|
|
// (retained or written) and on how output loss marks truncation.
|
|
checkCounted(
|
|
"captured output copied from a process snapshot (use capturedProviderOutput from src/process/capturedProviderOutput)",
|
|
"captured-output-copy",
|
|
files,
|
|
/\bstdout: [\w.]+\.stdout\.text\b/gu,
|
|
["src/process/capturedProviderOutput.ts"],
|
|
);
|
|
|
|
// Pushed effect invalidation (docs/adr/javascript/pushed-effect-invalidation.md).
|
|
// Recorded effects belong to the effect log. A reader that reads an effect
|
|
// index itself decides on its own which effect kinds and proofs apply; readers
|
|
// ask the log instead. Every reader has moved; the remaining counts are the
|
|
// producers that build these indexes (javascriptSemanticAmbientWrites.ts,
|
|
// javascriptSemanticMemberMutations.ts, and javascriptSemanticAnalysis.ts,
|
|
// which creates and settles them) and the timing-proof owner
|
|
// (javascriptSemanticMutationInitializers.ts), whose position proofs over
|
|
// recorded sites the log delegates to.
|
|
checkCounted(
|
|
"read of a JavaScript effect index outside the effect log (ask javascriptSemanticEffectLog.ts)",
|
|
"javascript-effect-index-read",
|
|
files,
|
|
/\.(?:bindingWrites|memberWrites|escapedPaths|ambientWriteEffects|ambientEscapeEffects|ambientBindingWrites|ambientWrites|ambientEscapes)\b/gu,
|
|
["src/domain/javascript/javascriptSemanticEffectLog.ts"],
|
|
);
|
|
|
|
// Effect timing has one owner (javascriptSemanticMutationInitializers.ts) and
|
|
// one consumer, the effect log. Everything else asks the log whether an effect
|
|
// may precede a read, naming the ordering it relies on.
|
|
check(
|
|
"semanticEffectTiming outside the effect log (ask semanticEffectMayPrecede or a query in javascriptSemanticEffectLog.ts)",
|
|
files,
|
|
/\bsemanticEffectTiming\b/u,
|
|
[
|
|
"src/domain/javascript/javascriptSemanticEffectLog.ts",
|
|
"src/domain/javascript/javascriptSemanticMutationInitializers.ts",
|
|
],
|
|
);
|
|
|
|
// Coverage status has one vocabulary, coverageStatusSchema in
|
|
// src/domain/coverage.ts. A local enum that contains "complete" either copies
|
|
// it or drops `unknown`, which forces an unknown producer to pick a side.
|
|
checkCounted(
|
|
"local completeness enum (use coverageStatusSchema from src/domain/coverage, or mark a detail level, lifecycle state or domain vocabulary with `// detail-level:`, `// lifecycle-state:` or `// domain-vocabulary:` and its reason)",
|
|
"local-completeness-enum",
|
|
files,
|
|
localCompletenessEnumCount,
|
|
["src/domain/coverage.ts"],
|
|
);
|
|
|
|
// Cancellation reasons are required by `cancelledBy` in analysisErrorCore; a
|
|
// direct construction can omit or replace the caller's reason.
|
|
checkCounted(
|
|
"direct AnalysisCancelledError construction (use cancelledBy from src/domain/analysisErrorCore)",
|
|
"analysis-cancelled-constructor",
|
|
files,
|
|
/new AnalysisCancelledError\(/gu,
|
|
["src/domain/analysisErrorCore.ts"],
|
|
);
|
|
|
|
// A limit a parser stopped at is a ParseFailureError<"limit"> (decodeLimit in
|
|
// src/domain/parseFailure); a refusal is an AnalysisRefusal. A local limit
|
|
// class on plain Error hides the limit's name and maximum from both; a limit
|
|
// subclass of BinaryResourceFailure keeps a second limit field beside the
|
|
// allowance its LimitRef already names. The environment-state extraction's private retention stop is control flow that
|
|
// its own entry point catches and turns into `null`; it is never reported.
|
|
checkCounted(
|
|
"local limit class (use decodeLimit from src/domain/parseFailure, a named BinaryResourceFailure allowance, or an AnalysisRefusal)",
|
|
"local-limit-error-class",
|
|
files,
|
|
localLimitErrorClassCount,
|
|
["src/domain/analysisView/EnvironmentStateExtraction.ts"],
|
|
);
|
|
|
|
// A resource allowance is charged and refunded through `Budget` in
|
|
// src/domain/budget, which refuses malformed or oversized charges without
|
|
// changing the remainder. Hand-written `remaining -=` / `retained +=`
|
|
// arithmetic repeats that admission and drifts (charge-then-check, no
|
|
// safe-integer check). A cursor, tally or retained length that is compared
|
|
// with a limit without being charged against it carries
|
|
// `// not-an-allowance: <reason>` on its field declaration (#11408).
|
|
checkCounted(
|
|
"hand-written budget arithmetic (use Budget from src/domain/budget, or mark a cursor or tally field with `// not-an-allowance: <reason>`)",
|
|
"budget-arithmetic",
|
|
files,
|
|
budgetArithmeticCount,
|
|
["src/domain/budget.ts"],
|
|
);
|
|
|
|
// A slice to a named maximum drops items; boundedList in src/domain/boundedList
|
|
// keeps the count of what it dropped beside the retained items, so a consumer
|
|
// can say that more exist and how many. Two caps drop no evidence and carry a
|
|
// reason marker instead (#11411): `// generation-cap: <reason>` limits what REA
|
|
// generates (suggested checks, guards) or the work spent finding it, and
|
|
// `// sample-cap: <reason>` limits illustrative samples whose full totals the
|
|
// same result already reports.
|
|
checkCounted(
|
|
"slice to a named maximum (use boundedList or BoundedListCollector from src/domain/boundedList so the omitted count travels with the items, or mark a generation or sample cap with its reason)",
|
|
"constant-maximum-slice",
|
|
files,
|
|
(_content, file) =>
|
|
unmarkedMatchCount(
|
|
file.source,
|
|
/\.slice\(\s*0\s*,\s*[A-Z][A-Z0-9_]*\s*\)/gu,
|
|
["generation-cap", "sample-cap"],
|
|
),
|
|
["src/domain/boundedList.ts"],
|
|
);
|
|
|
|
// Display text cut with an appended ellipsis comes from ellipsisExcerpt in
|
|
// src/domain/excerpt, which never splits a surrogate pair and marks the result
|
|
// as excerpt text. A hand-written `${text.slice(0, n)}…` does neither.
|
|
checkCounted(
|
|
"hand-written ellipsis excerpt (use ellipsisExcerpt from src/domain/excerpt)",
|
|
"ad-hoc-ellipsis",
|
|
files,
|
|
/\.slice\(\s*0\s*,[^)]*\)\)?\}(?:…|\.\.\.)/gu,
|
|
["src/domain/excerpt.ts"],
|
|
);
|
|
|
|
// A deadline is a monotonic budget owned by `Deadline` in src/domain/deadline;
|
|
// a wall-clock sum moves with clock changes and has no finalize phase. The
|
|
// remaining copy converts a monotonic budget for a browser page, which can
|
|
// only read its own wall clock.
|
|
checkCounted(
|
|
"wall-clock deadline computation (use Deadline from src/domain/deadline)",
|
|
"wall-clock-deadline",
|
|
files,
|
|
/Date\.now\(\)\s*\+/gu,
|
|
[],
|
|
);
|
|
|
|
// `tshark -T json` has one envelope reader, src/domain/pcap/tsharkJsonPackets.ts,
|
|
// which owns malformed JSON, the packet array and each packet's layers.
|
|
checkCounted(
|
|
"local tshark JSON envelope read (use parseTsharkJsonPackets or readTsharkJsonPackets from src/domain/pcap/tsharkJsonPackets)",
|
|
"tshark-json-envelope",
|
|
files,
|
|
/\b_source\s*\.\s*layers\b/gu,
|
|
["src/domain/pcap/tsharkJsonPackets.ts"],
|
|
);
|
|
|
|
// Text from an unknown caught value has one owner, src/kernel/errorMessage.ts.
|
|
// The pattern matches the inline copy with the same expression in all three
|
|
// places; a named local helper with this body is the same copy.
|
|
// Copies that run in another realm cannot import the owner and are excluded by
|
|
// name, so no slice rewrites them: the regex worker's source string and a
|
|
// fetch run inside the browser page through page.evaluate.
|
|
const ISOLATED_ERROR_MESSAGE_COPIES = new Map([
|
|
["src/browser/PlaywrightScenarioRequest.ts", 1],
|
|
["src/process/RegexMatcher.ts", 1],
|
|
]);
|
|
const inlineErrorMessage =
|
|
/(?<![\w$.])([A-Za-z_$][\w$]*(?:\.[A-Za-z_$][\w$]*)*)\s+instanceof\s+Error\s*\?\s*\1\.message\s*:\s*String\(\s*\1\s*\)/gu;
|
|
checkCounted(
|
|
"inline error-to-message conversion (use errorMessage from src/kernel/errorMessage)",
|
|
"inline-error-message",
|
|
files,
|
|
(content, { path }) =>
|
|
Math.max(
|
|
0,
|
|
(content.match(inlineErrorMessage)?.length ?? 0) -
|
|
(ISOLATED_ERROR_MESSAGE_COPIES.get(path) ?? 0),
|
|
),
|
|
["src/kernel/errorMessage.ts"],
|
|
);
|
|
|
|
// Two hexadecimal digits for one byte have one owner, src/kernel/byteToHex.ts.
|
|
checkCounted(
|
|
"inline byte-to-hex formatting (use byteToHex from src/kernel/byteToHex)",
|
|
"inline-byte-hex",
|
|
files,
|
|
/\.toString\(16\)\s*\.padStart\(\s*2\s*,\s*"0"\s*\)/gu,
|
|
["src/kernel/byteToHex.ts"],
|
|
);
|
|
|
|
// A one-shot SHA-256 hex digest of an in-memory value has one owner,
|
|
// src/kernel/sha256Hex.ts. Incremental and streamed hashing (a Hash updated
|
|
// more than once) is not counted.
|
|
checkCounted(
|
|
"one-shot SHA-256 hex digest (use sha256Hex from src/kernel/sha256Hex)",
|
|
"inline-sha256-hex",
|
|
files,
|
|
/createHash\(\s*"sha256"\s*\)\s*\.update\((?:[^()]|\((?:[^()]|\((?:[^()]|\([^()]*\))*\))*\))*\)\s*\.digest\(\s*"hex"\s*\)/gu,
|
|
["src/kernel/sha256Hex.ts"],
|
|
);
|
|
|
|
// Strict UTF-8 decoding has one owner, src/domain/strictUtf8.ts, which makes
|
|
// each site state its byte-order-mark policy. A local fatal decoder silently
|
|
// strips a leading BOM unless it remembers ignoreBOM. The pattern matches the
|
|
// options in any order and an omitted label. The remaining copy runs inside a
|
|
// browser page through page.evaluate, which cannot import the owner.
|
|
checkCounted(
|
|
"local strict UTF-8 TextDecoder (use strictUtf8Decoder, decodeStrictUtf8 or isStrictUtf8 from src/domain/strictUtf8)",
|
|
"strict-utf8-decoder",
|
|
files,
|
|
/new TextDecoder\(\s*(?:(?:["'`]utf-?8["'`]|undefined)\s*,\s*)?\{[^}]*\bfatal:\s*true/giu,
|
|
["src/domain/strictUtf8.ts"],
|
|
);
|
|
|
|
// Turning a completed result into a cancellation after cleanup is the
|
|
// finalization precedence that runOwned owns (src/providers/runOwned.ts).
|
|
checkCounted(
|
|
"completed result reclassified by a hand-read abort signal (finalize through runOwned in src/providers/runOwned)",
|
|
"completed-abort-reclassification",
|
|
files,
|
|
/\.ok\s*&&\s*(?:isAborted\(|[\w.?]*signal\??\.aborted)/gu,
|
|
["src/providers/runOwned.ts"],
|
|
);
|
|
|
|
// An abort reason names who ended the work: a typed `CallerAbort` at the CLI,
|
|
// MCP and HTTP adapters, or an `OwnerCancellation` when REA closes an owner.
|
|
// A bare `.abort()` leaves an anonymous platform AbortError as the cause.
|
|
// Playwright's `route.abort(errorCode)` takes a network error code instead;
|
|
// its call sites pass that code explicitly so the scan stays receiver-independent.
|
|
checkCounted(
|
|
"AbortController.abort() without a reason (pass a CallerAbort or OwnerCancellation from src/domain/analysisErrorCore)",
|
|
"abort-without-reason",
|
|
files,
|
|
/\.abort\(\s*\)/gu,
|
|
[],
|
|
);
|
|
|
|
// GDB MI result records have one parser, src/gdb/GdbMiRecord.ts, which
|
|
// returns typed records and typed failures. A regex over `\^done` accepts a
|
|
// malformed record by its prefix.
|
|
checkCounted(
|
|
"GDB MI result record regex (use parseGdbMiResultRecord or parseGdbMiDoneRecord from src/gdb/GdbMiRecord)",
|
|
"gdb-mi-record-regex",
|
|
files,
|
|
/\\\^\(?(?:\?:)?(?:done|running|connected|error|exit)\b/gu,
|
|
["src/gdb/GdbMiRecord.ts"],
|
|
);
|
|
|
|
// src/cli owns option and command spellings; cli-text-guard.mjs states the
|
|
// rule and its written exemptions (host-only commands, CLI-only files).
|
|
checkCounted(
|
|
"REA CLI option or command named outside the CLI layer (name the contract field or operation; src/cli owns option and command spellings)",
|
|
"cli-option-text-outside-cli",
|
|
files.filter(({ path }) => cliTextGuardSelects(path)),
|
|
cliTextOutsideCliCount,
|
|
[],
|
|
);
|
|
|
|
// tool-registration-guard.mjs states the rule and its written owners.
|
|
checkCounted(
|
|
"MCP tool registered outside the shared binding table (bind it in src/server/serviceToolBindings.ts)",
|
|
"bespoke-mcp-tool-registration",
|
|
files.filter(({ path }) => path.startsWith("src/")),
|
|
toolRegistrationCount,
|
|
[...TOOL_REGISTRATION_OWNERS.keys()],
|
|
);
|
|
|
|
// A runtime setting view throws the ConfigurationError of a malformed launch
|
|
// seed; readRuntimeSetting in src/config/runtimeAnalysisSettings.ts returns it
|
|
// as a Result. Catching it by hand mixes it into each site's own failures.
|
|
checkCounted(
|
|
"hand-caught ConfigurationError (read runtime settings with readRuntimeSetting from src/config/runtimeAnalysisSettings)",
|
|
"hand-caught-configuration-error",
|
|
files,
|
|
/\bcause instanceof ConfigurationError\b/gu,
|
|
["src/config/runtimeAnalysisSettings.ts"],
|
|
);
|
|
|
|
// Byte-format parsers read untrusted input. Bounds, truncation and failure
|
|
// kind belong to BinaryReader (src/domain/binaryContainer.ts) and
|
|
// the ParseFailure algebra (src/domain/parseFailure.ts), not to each parser.
|
|
const PARSER_FOLDERS = [
|
|
"src/android/",
|
|
"src/apkSignature/",
|
|
"src/artifacts/",
|
|
"src/dotnet/",
|
|
"src/electron/",
|
|
"src/erlang/",
|
|
"src/evm/",
|
|
"src/firmware/",
|
|
"src/flutter/",
|
|
"src/go/",
|
|
"src/gpu/",
|
|
"src/il2cpp/",
|
|
"src/model/",
|
|
"src/native/",
|
|
"src/python/",
|
|
// Shared artifact readers (ZIP, directory, ASAR header) live in resources.
|
|
"src/resources/",
|
|
"src/rust/",
|
|
"src/v8/",
|
|
"src/wasm/",
|
|
...[
|
|
"android",
|
|
"apkSignature",
|
|
"apple",
|
|
"electron",
|
|
"erlang",
|
|
"evm",
|
|
"firmware",
|
|
"flutter",
|
|
"go",
|
|
"gpu",
|
|
"il2cpp",
|
|
"managed",
|
|
"model",
|
|
"native",
|
|
"pcap",
|
|
"python",
|
|
"rust",
|
|
"v8",
|
|
"wasm",
|
|
"windows",
|
|
"zip",
|
|
].map((folder) => `src/domain/${folder}/`),
|
|
];
|
|
const parserFiles = files.filter((file) =>
|
|
PARSER_FOLDERS.some((folder) => file.path.startsWith(folder)),
|
|
);
|
|
|
|
// An unchecked Buffer or DataView read throws a bare RangeError past the end,
|
|
// which a catch site cannot tell from a host stack or string limit.
|
|
checkCounted(
|
|
"raw Buffer or DataView numeric read (use BinaryReader from src/domain/binaryContainer)",
|
|
"raw-buffer-read",
|
|
files,
|
|
/\.(?:read(?:Big)?U?Int(?:8|16|32|64)?(?:LE|BE)?|readBigUint64(?:LE|BE)|readUint(?:8|16|32)(?:LE|BE)?|read(?:Float|Double)(?:LE|BE)|get(?:Big)?(?:Uint|Int)(?:8|16|32|64)|getFloat(?:32|64))\(/gu,
|
|
["src/domain/binaryContainer.ts"],
|
|
);
|
|
|
|
// An input fault is a ParseFailureError; a broken internal assumption is a
|
|
// defect, not an input fault, and throws invariantViolation from
|
|
// src/domain/internalInvariant so callers never read it as bad bytes.
|
|
// Process-lifecycle owners in parser folders read their private helper's
|
|
// replies through schemas; their plain throws report the helper's startup,
|
|
// handshake and cleanup outcomes, which are process failures, not parse faults.
|
|
const PROCESS_LIFECYCLE_OWNERS = [
|
|
"src/android/JadxWorkspaceRoot.ts",
|
|
"src/native/pwntools/PwntoolsDecoder.ts",
|
|
];
|
|
const PLAIN_THROW = /throw new (?:Error|TypeError|RangeError)\(/gu;
|
|
checkCounted(
|
|
"plain Error, TypeError or RangeError thrown by a parser (throw a ParseFailureError from src/domain/parseFailure for an input fault, or invariantViolation from src/domain/internalInvariant for a broken internal assumption)",
|
|
"parser-plain-throw",
|
|
parserFiles.filter((file) => !PROCESS_LIFECYCLE_OWNERS.includes(file.path)),
|
|
PLAIN_THROW,
|
|
[],
|
|
);
|
|
// The lifecycle owners' existing process-failure throws are counted, so a
|
|
// parse fault or defect added there later cannot hide behind the exemption.
|
|
checkCounted(
|
|
"plain Error, TypeError or RangeError thrown by a parser-folder process-lifecycle owner (keep its existing process-failure throws; throw a ParseFailureError for an input fault or invariantViolation for a broken internal assumption)",
|
|
"lifecycle-plain-throw",
|
|
parserFiles.filter((file) => PROCESS_LIFECYCLE_OWNERS.includes(file.path)),
|
|
PLAIN_THROW,
|
|
[],
|
|
);
|
|
|
|
// A per-format failure class restates the ParseFailure algebra under its own
|
|
// spelling and needs its own catch-site translation; parsers throw a
|
|
// ParseFailureError instead. A subclass of ParseFailureError (such as
|
|
// CaptureFormatError) keeps the shared kind and is not counted. The name
|
|
// pattern covers format, parse, document, source, decode and syntax failures,
|
|
// and any property-list or archive failure (such as a nesting limit or an
|
|
// unsupported UID), built on any of the host's general error classes.
|
|
checkCounted(
|
|
"per-format failure class extending a host error (throw a ParseFailureError from src/domain/parseFailure)",
|
|
"parser-format-failure-class",
|
|
files,
|
|
/class (?:\w*(?:Format|Parse|Document|Source|Decode|Decoding|Encoding|Syntax)|\w*(?:Plist|PropertyList|Archive)\w*)(?:Failure|Error|Issue) extends (?:Error|TypeError|RangeError|SyntaxError)\b/gu,
|
|
["src/domain/parseFailure.ts"],
|
|
);
|
|
|
|
// Mach-O load-command codes have one table; a literal code in a load-command
|
|
// reader restates it and drifts from the named command it means.
|
|
checkCounted(
|
|
"literal Mach-O load-command code (use LC from src/domain/apple/machoLoadCommandCodes)",
|
|
"macho-load-command-literal",
|
|
files.filter((file) =>
|
|
/MachoLoadCommandCursor|readMachoLoadCommand|machoLoadCommandCodes/u.test(
|
|
file.content,
|
|
),
|
|
),
|
|
// An equality test, or membership in a literal list of codes.
|
|
/\b(?:frame\.command|command\.kind|command|cmd)\s*[!=]==\s*(?:0x[0-9a-fA-F]+|\d+)\b|\[\s*(?:0x[0-9a-fA-F]+|\d+)(?:\s*,\s*(?:0x[0-9a-fA-F]+|\d+))*\s*,?\s*\]\s*\.(?:includes|indexOf)\(\s*(?:\w+\.)*(?:command|cmd|kind)\s*\)/gu,
|
|
["src/domain/apple/machoLoadCommandCodes.ts"],
|
|
);
|
|
|
|
// JSON.parse keeps the last of repeated member names without saying so; a
|
|
// parser of artifact content must state its duplicate-member policy
|
|
// (src/domain/jsonDuplicateMember.ts finds a repeated name).
|
|
checkCounted(
|
|
"plain JSON.parse of parser input (state the duplicate-member policy; see src/domain/jsonDuplicateMember)",
|
|
"parser-json-parse",
|
|
parserFiles,
|
|
/JSON\.parse\(/gu,
|
|
[],
|
|
);
|
|
|
|
// Raw JSON Schema keywords and `x-rea-*` markers in `.meta()` bypass the
|
|
// canonical Zod checks and reach every JSON Schema projection, including the
|
|
// advertised one. Presentation facts are typed annotations
|
|
// (src/domain/schemaAnnotations); constraints belong in Zod checks, or in a
|
|
// src/domain/schemaConstraints helper that adds the check and its keyword
|
|
// together.
|
|
checkCounted(
|
|
"raw JSON Schema keyword or x-rea marker in Zod .meta() (use a typed annotation from src/domain/schemaAnnotations, a Zod check, or a src/domain/schemaConstraints helper)",
|
|
"raw-json-schema-meta",
|
|
files,
|
|
/\.meta\(\s*\{\s*"?(?:x-rea-[\w-]+|\$ref|not|allOf|anyOf|oneOf|if|then|else|dependentRequired|dependentSchemas|dependencies|uniqueItems|(?:min|max)(?:Length|Items|Properties|imum)|exclusive(?:Min|Max)imum|multipleOf|type|const|enum|format|pattern|patternProperties|properties|required|additionalProperties|items|prefixItems|contains|content(?:MediaType|Encoding|Schema)|propertyNames)"?\s*:/gu,
|
|
["src/domain/schemaConstraints.ts"],
|
|
);
|
|
|
|
// A RangeError can mean a stack overflow, a host string or buffer limit, a
|
|
// raw read or write past a buffer, or whatever its thrower intended. Only
|
|
// known host limits are classified; parsers throw a ParseFailureError.
|
|
checkCounted(
|
|
"direct RangeError interpretation (use hostRangeErrorKind or parseFailureFromCause from src/domain/parseFailure)",
|
|
"range-error-interpretation",
|
|
files,
|
|
/instanceof RangeError/gu,
|
|
["src/domain/parseFailure.ts"],
|
|
);
|
|
|
|
// Cleanup completion is recorded by the owner and read by errors when asked;
|
|
// AnalysisError has no push API. Composing dependency sets with
|
|
// bindCleanupDependencies() is being replaced, so it may not spread.
|
|
checkCounted(
|
|
"bindCleanupDependencies() composition outside the error base (compose through analysisErrorWithCleanupDependencies, analysisErrorWithCloseRejection(s) for close rejections whose completion may be unknown, or runOwned, or copy lease claims)",
|
|
"cleanup-dependency-binding",
|
|
files,
|
|
/\.bindCleanupDependencies\(/gu,
|
|
["src/domain/analysisErrorBase.ts", "src/domain/analysisErrorCleanup.ts"],
|
|
);
|
|
|
|
// A launcher's PATH search order (POSIX empty and relative entries and `..`;
|
|
// Windows quoted entries and PATHEXT) belongs to src/process/executableSearch;
|
|
// callers choose only how they probe each candidate.
|
|
checkCounted(
|
|
"PATH or PATHEXT split outside the executable search owner (use posixPathCandidates or windowsPathCandidates from src/process/executableSearch)",
|
|
"path-search-split",
|
|
files,
|
|
/\.split\(\s*(?:posix\.|win32\.)?(?:delimiter|pathDelimiter)\s*\)|\bPATH\b[^\n]{0,12}\.split\(":"\)|\bPATHEXT\b[^\n]{0,60}\.split\(/gu,
|
|
["src/process/executableSearch.ts"],
|
|
);
|
|
|
|
// The running host is read once by src/platform/HostPlatform; composition
|
|
// takes it from there and passes the platform or architecture on as a
|
|
// required parameter. An ambient read elsewhere mixes the host REA runs on
|
|
// with the host a caller asked about, and cannot be selected by a test.
|
|
// Named imports from node:process or node:os and os.platform()/os.arch() are
|
|
// the same read under another spelling.
|
|
const ambientHostRead = (property, osFunction) =>
|
|
new RegExp(
|
|
String.raw`\bprocess\.${property}\b|\bos\.${osFunction}\(\)|\bimport\s*(?:type\s*)?\{[^}]*\b(?:${property}|${osFunction})\b[^}]*\}\s*from\s*["'](?:node:)?(?:process|os)["']`,
|
|
"gu",
|
|
);
|
|
checkCounted(
|
|
"ambient process.platform read (take HOST_PLATFORM from src/platform/HostPlatform at composition and pass the platform as a parameter)",
|
|
"ambient-host-platform",
|
|
files,
|
|
ambientHostRead("platform", "platform"),
|
|
["src/platform/HostPlatform.ts"],
|
|
);
|
|
checkCounted(
|
|
"ambient process.arch read (take HOST_PLATFORM from src/platform/HostPlatform at composition and pass the architecture as a parameter)",
|
|
"ambient-host-architecture",
|
|
files,
|
|
ambientHostRead("arch", "arch"),
|
|
["src/platform/HostPlatform.ts"],
|
|
);
|
|
|
|
// Tests select a platform or architecture by injecting it (a parameter, a
|
|
// `host` option, or a mock of src/platform/HostPlatform). Redefining
|
|
// process.platform or process.arch stops exercising its branch once the module
|
|
// under test takes the host from HostPlatform, and still passes. Each module
|
|
// migration switches its tests to injection and lowers this baseline.
|
|
const listTestFiles = async (dir) => {
|
|
const output = [];
|
|
for (const entry of await readdir(dir, { withFileTypes: true })) {
|
|
const path = join(dir, entry.name);
|
|
if (entry.isDirectory()) output.push(...(await listTestFiles(path)));
|
|
else if (/\.(?:test|spec|fixture)\.ts$/u.test(entry.name))
|
|
output.push(path);
|
|
}
|
|
return output;
|
|
};
|
|
const testFiles = await Promise.all(
|
|
[
|
|
...(await listTestFiles(join(root, "src"))),
|
|
...(await listFiles(join(root, "tests"))),
|
|
...(await listTestFiles(join(root, "tests"))),
|
|
].map(async (path) => ({
|
|
path: relative(root, path).split(sep).join("/"),
|
|
content: await readFile(path, "utf8"),
|
|
})),
|
|
);
|
|
checkCounted(
|
|
"test redefines process.platform or process.arch (inject the platform: pass it as a parameter or host option, or mock src/platform/HostPlatform)",
|
|
"test-host-platform-stub",
|
|
testFiles,
|
|
/\b(?:defineProperty|spyOn)\(\s*(?:globalThis\.)?process\s*,\s*["'`](?:platform|arch)["'`]/gu,
|
|
[],
|
|
);
|
|
|
|
// `find` on a parsed section, program-header or load-command table silently
|
|
// keeps the first of repeated declarations. Select keyed entries through
|
|
// src/domain/keyedCollection (`uniqueKeyed`, `allKeyed`, or
|
|
// `firstKeyed`/`lastKeyed` with the format's precedence rule); an
|
|
// address-containment test that selects nothing uses `some`. GuidanceAccess
|
|
// looks up a guidance document's heading section by its locator identity,
|
|
// not a parsed binary section table.
|
|
checkCounted(
|
|
"first-match section, program or command lookup (select through src/domain/keyedCollection)",
|
|
"section-first-match",
|
|
files,
|
|
/\b(?:sections|programs|commands)\.find(?:Last)?\(/gu,
|
|
["src/application/GuidanceAccess.ts"],
|
|
);
|
|
|
|
// The same first-match choice, recognized by its shape rather than the
|
|
// table's variable name (see keyed-first-match.mjs). Select through
|
|
// src/domain/keyedCollection instead.
|
|
checkCounted(
|
|
"first-match lookup by key field in a parser (select through src/domain/keyedCollection)",
|
|
"keyed-first-match",
|
|
parserFiles,
|
|
countKeyedFirstMatches,
|
|
[],
|
|
);
|
|
|
|
// A verifier fixture root is removed only after the MCP clients, sessions and
|
|
// processes that may still use it confirm cleanup; otherwise it is retained
|
|
// and named. scripts/lib/verifier-fixture.mjs owns that ordering on a
|
|
// src/process PrivateRuntimeRoot, so verifiers allocate roots through it.
|
|
const verifierScripts = await Promise.all(
|
|
(await readdir(join(root, "scripts"), { recursive: true }))
|
|
.map((path) => path.split(sep).join("/"))
|
|
.filter(
|
|
(path) =>
|
|
/^(?:verify-[^/]+|verify\/.+)\.mjs$/u.test(path) &&
|
|
path !== "verify/architecture-guards.mjs",
|
|
)
|
|
.map(async (path) => ({
|
|
path: `scripts/${path}`,
|
|
content: blankComments(
|
|
await readFile(join(root, "scripts", path), "utf8"),
|
|
),
|
|
})),
|
|
);
|
|
// The Windows native lane tests PrivateRuntimeRoot itself: it calls
|
|
// PrivateRuntimeRoot.create under a parent whose DELETE lease forces a sharing
|
|
// violation and asserts the typed refusal. That call is the subject under
|
|
// test, not a verifier root, so exactly one occurrence of that expression is
|
|
// exempt; any other allocation in the file, including a second copy of it,
|
|
// still counts.
|
|
const verifierRootSubjectsUnderTest = new Map([
|
|
[
|
|
"scripts/verify-windows-native.mjs",
|
|
"PrivateRuntimeRoot.create({ parent: sharingParent.observation.path })",
|
|
],
|
|
]);
|
|
checkCounted(
|
|
"verifier fixture root allocated outside VerifierFixture (use scripts/lib/verifier-fixture.mjs)",
|
|
"verifier-fixture-root",
|
|
verifierScripts.map((file) => {
|
|
const subject = verifierRootSubjectsUnderTest.get(file.path);
|
|
return subject === undefined
|
|
? file
|
|
: { ...file, content: file.content.replace(subject, "") };
|
|
}),
|
|
/\b(?:mkdtemp(?:Sync)?|createCanonicalTemporaryDirectory)\(|\bPrivateRuntimeRoot\.create\(/gu,
|
|
[],
|
|
);
|
|
|
|
for (const key of baselineTexts.keys())
|
|
if (!checkedBaselineKeys.has(key))
|
|
failures.push(
|
|
`${relative(root, join(baselineDirectory, `${key}.json`))} names no counted guard; delete it or restore its guard.`,
|
|
);
|
|
for (const key of checkedBaselineKeys) {
|
|
const path = join(baselineDirectory, `${key}.json`);
|
|
const shown = relative(root, path);
|
|
const baselineFile = reconcileBaselineFile(
|
|
baselineTexts.get(key),
|
|
baselines[key],
|
|
shrinkBaselines,
|
|
);
|
|
if (baselineFile.action === "write") {
|
|
await writeFile(path, baselineFile.text);
|
|
console.log(`Rewrote ${shown}.`);
|
|
} else if (baselineFile.action === "delete") {
|
|
await rm(path);
|
|
console.log(`Removed ${shown}; its guard has no remaining copies.`);
|
|
} else if (baselineFile.action === "reject")
|
|
failures.push(
|
|
`${shown} is not in its written form (paths sorted by code unit); rewrite it with \`node scripts/verify/architecture-guards.mjs --shrink-baselines\`.`,
|
|
);
|
|
}
|
|
if (baselinesShrunk) console.log("Lowered counted baselines.");
|
|
|
|
if (failures.length > 0) {
|
|
console.error(`Architecture guards failed:\n\n${failures.join("\n\n")}\n`);
|
|
process.exit(1);
|
|
}
|
|
console.log(`Architecture guards passed across ${files.length} source files.`);
|