@crewhaus/tool-buildperf 0.7.0 → 0.8.0

package/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

package/NOTICE

@@ -0,0 +1,9 @@
+CrewHaus Factory
+Copyright 2026 Max Meier
+
+This product includes software developed by Max Meier
+(operating as StudioMax — https://studiomax.io) and contributors.
+
+The CrewHaus name and logo are trademarks of Max Meier. See
+https://github.com/crewhaus/factory/blob/main/TRADEMARK.md for the
+use-of-marks policy.

package/README.md

@@ -139,6 +139,15 @@ failure went ungrouped. One group among the failures that said something is not
 an answer about the failures that did not, and answering `true` from that
 subset is what sends someone after a single root cause for two different bugs.
 
+A mask, and `BundleSizeCheck`'s `joinPattern`, is a regular expression the
+caller writes. Each is screened before it is accepted: a shape that
+backtracks catastrophically (`(a+)+`, `(\w+\s?)*`, `(a|a)*`) is refused with
+the reason. It then runs in `@crewhaus/tool-safety`'s regex worker under a
+deadline, never on the harness's own thread. A text or path the worker could
+not answer (the engine gave up, or the deadline passed) fails the call with
+the reason: a failure grouped by an unmasked text, or an artifact joined by an
+unnormalised key, would be a confident wrong answer.
+
 There are deliberately no built-in masks. `ErrorCluster` in
 `@crewhaus/tool-obs` already owns shape-based masking of error text (URLs,
 uuids, timestamps, paths, prefixed ids, hex blobs, numbers). A second set here

package/dist/index.js

@@ -24,15 +24,17 @@
  * filesystem packages use. Nothing here runs a build, a benchmark or a test:
  * the measurements come in, the honest reading of them comes out.
  */
-import { readFileSync, readdirSync, statSync } from "node:fs";
+import { readdirSync, statSync } from "node:fs";
 import { join } from "node:path";
 import { buildTool } from "@crewhaus/tool-builder";
 import { statsKernel } from "@crewhaus/tool-math";
+import { openForReadSync } from "@crewhaus/tool-safety/fs";
+import { describeRegexOutcome, openRegexSession, runRegex, screenUserRegex, } from "@crewhaus/tool-safety/regex";
 import { z } from "zod";
 import { MIN_SAMPLES_FOR_SIGNIFICANCE, compareBenchmarks } from "./lib/bench";
-import { TEST_STATUSES, detectFlaky } from "./lib/flaky";
+import { TEST_STATUSES, detectFlaky, foldWhitespace, maskSubjects, } from "./lib/flaky";
 import { COMPRESSION_ALGORITHMS, JOIN_MODES, MAX_GLOB_LENGTH, baselineBasisMismatch, comparabilityKey, compileGlob, compressedSize, compressionParameters, joinKey, parametersMismatch, } from "./lib/size";
-import { resolveSafe, toPosix } from "./paths";
+import { resolveSafe, toPosix, workspaceRoot } from "./paths";
 const json = (value) => JSON.stringify(value);
 /** Ceilings that keep one pathological input from becoming a hang. */
 const LIMITS = {
@@ -40,6 +42,8 @@ const LIMITS = {
     entries: 2_000,
     /** Compressing a file this big to weigh it is not the intended use. */
     fileBytes: 512 * 1024 * 1024,
+    /** A stored size report: entries and totals, never the bundle itself. */
+    baselineBytes: 64 * 1024 * 1024,
     budgets: 200,
     benchmarks: 500,
     samplesPerSide: 10_000,
@@ -50,6 +54,8 @@ const LIMITS = {
 /** The zlib build that produced these numbers; see `runtime` in the report. */
 const ZLIB_VERSION = process.versions.zlib ?? "unknown";
 // --- BundleSizeCheck --------------------------------------------------------
+/** Longest artifact path accepted in a baseline row. */
+const MAX_PATH_CHARS = 4_096;
 // Not `.strict()`: a stored baseline is a previous REPORT, which carries the
 // derived `comparabilityKey` alongside the parameters. Rejecting the report
 // this tool just produced would make "store it and pass it back" a lie.
@@ -65,13 +71,67 @@ const baselineSchema = z.object({
     parameters: parametersSchema.optional(),
     entries: z
         .array(z.object({
-        path: z.string().min(1),
+        // A baseline path is a file path: longer than any PATH_MAX is not
+        // one, and it is also a subject for the caller's joinPattern.
+        path: z.string().min(1).max(MAX_PATH_CHARS),
         bytes: z.number().int().nonnegative(),
         compressedBytes: z.number().int().nonnegative().nullable().optional(),
     }))
         .max(LIMITS.entries),
     runtime: z.object({ zlib: z.string() }).optional(),
 });
+/**
+ * The screen a caller-supplied regex gets before it is accepted: length,
+ * flags, syntax and the catastrophic-backtracking shapes, with a reason the
+ * model can act on (C073). `g` is what a join or mask applies.
+ */
+function patternIssue(pattern, flags) {
+    const verdict = screenUserRegex(pattern, flags);
+    return verdict.ok ? undefined : verdict.reason;
+}
+/**
+ * Normalise paths with the caller's joinPattern (every match becomes
+ * `[hash]`), in @crewhaus/tool-safety's regex worker under a deadline — never
+ * with a synchronous `replace` on this thread, which a pathological pattern
+ * could hold for seconds per path, and which JavaScriptCore can abandon and
+ * report as "no match". Any path whose key could not be determined fails the
+ * whole call: a size gate that joined some rows by an unnormalised key would
+ * report a renamed chunk as removed-and-added.
+ */
+async function customJoinKeys(pattern, paths, ctx) {
+    const distinct = [...new Set(paths)];
+    const keys = new Map();
+    if (distinct.length === 0)
+        return { ok: true, keys };
+    const outcome = await runRegex({
+        op: "replaceEach",
+        pattern,
+        flags: "g",
+        inputs: distinct,
+        replacement: "[hash]",
+        onGiveUp: "skip",
+        deadlineMs: 2_000,
+        ...(ctx?.signal !== undefined ? { signal: ctx.signal } : {}),
+        ...(ctx?.runContext !== undefined ? { runawayKey: ctx.runContext.sessionId } : {}),
+    });
+    if (outcome.status !== "ok") {
+        return {
+            ok: false,
+            message: `BundleSizeCheck could not normalise the artifact paths with joinPattern /${pattern}/: ${describeRegexOutcome(outcome)}. No comparison was made.`,
+        };
+    }
+    const { outputs, undetermined } = outcome.result;
+    if (undetermined.length > 0) {
+        const first = distinct[undetermined[0]] ?? "";
+        return {
+            ok: false,
+            message: `BundleSizeCheck could not determine the joinPattern key of ${undetermined.length} path(s) (first: "${first}") — the pattern engine gave up on them. Simplify joinPattern; no comparison was made.`,
+        };
+    }
+    for (const [i, path] of distinct.entries())
+        keys.set(path, outputs[i]);
+    return { ok: true, keys };
+}
 /** Every file under a directory, workspace-relative, sorted, capped. */
 function filesUnder(root, rel, out) {
     for (const entry of readdirSync(join(root, rel), { withFileTypes: true }).sort((a, b) => a.name < b.name ? -1 : 1)) {
@@ -97,7 +157,7 @@ function filesUnder(root, rel, out) {
     }
     return out;
 }
-function measure(paths, parameters, mode, custom) {
+function measure(paths, parameters, mode) {
     return paths.map((path) => {
         const at = resolveSafe("BundleSizeCheck", path);
         const stats = statSync(at.real);
@@ -106,10 +166,19 @@ function measure(paths, parameters, mode, custom) {
         if (stats.size > LIMITS.fileBytes) {
             throw new Error(`"${path}" is ${stats.size} bytes, over the ${LIMITS.fileBytes}-byte limit`);
         }
-        const bytes = readFileSync(at.real);
+        // Opened as the regular file the stat saw, without blocking: a name
+        // swapped for a FIFO after the stat is refused, not waited on.
+        const read = openForReadSync(workspaceRoot(), path, { maxBytes: LIMITS.fileBytes });
+        if (!read.ok)
+            throw new Error(`BundleSizeCheck: ${read.reason}`);
+        if (read.truncated) {
+            throw new Error(`"${path}" is ${read.size} bytes, over the ${LIMITS.fileBytes}-byte limit`);
+        }
+        const bytes = Buffer.from(read.bytes.buffer, read.bytes.byteOffset, read.bytes.byteLength);
         return {
             path: at.rel,
-            key: joinKey(at.rel, mode, custom).key,
+            // A custom join is keyed afterwards, in the regex worker.
+            key: mode === "custom" ? at.rel : joinKey(at.rel, mode).key,
             bytes: bytes.byteLength,
             compressedBytes: parameters.algorithm === "none" ? null : compressedSize(bytes, parameters),
         };
@@ -126,6 +195,11 @@ function percentDelta(before, after) {
 }
 export const bundleSizeCheck = buildTool({
     name: "BundleSizeCheck",
+    operativeArgs: [
+        { field: "files", kind: "path" },
+        { field: "directory", kind: "path", beneath: "all" },
+        { field: "baselineFile", kind: "path" },
+    ],
     description: "Weigh build artifacts — raw and gzip or brotli — against declared budgets and a stored baseline, and report what actually grew. Use it as the size gate on a pull request, instead of a model comparing two numbers whose provenance it cannot see. The compression parameters are part of the measurement and travel with the report: a baseline produced at a different level, with a different algorithm, or with no declared parameters at all is REFUSED rather than differenced, because the gap between gzip level 6 and level 9 is the same few percent a regression is, and subtracting them produces a confident number about nothing. Hashed filenames (app-4f2a1c.js) can be joined on a normalised key so a content-hash change is not reported as every file being new, and two artifacts that normalise to the same key are reported as ambiguous rather than silently paired. The report is itself a baseline: store it and pass it back next time.",
     inputSchema: z
         .object({
@@ -158,6 +232,13 @@ export const bundleSizeCheck = buildTool({
         joinPattern: z
             .string()
             .optional()
+            .superRefine((pattern, ctx) => {
+            if (pattern === undefined)
+                return;
+            const issue = patternIssue(pattern, "g");
+            if (issue !== undefined)
+                ctx.addIssue({ code: z.ZodIssueCode.custom, message: issue });
+        })
             .describe("with join=custom: a regex whose matches are replaced by [hash]"),
         baseline: baselineSchema.optional().describe("a previous report, inline"),
         baselineFile: z
@@ -198,7 +279,7 @@ export const bundleSizeCheck = buildTool({
     }),
     readOnly: true,
     concurrencySafe: true,
-    execute: async (input) => {
+    execute: async (input, ctx) => {
         const warnings = [];
         let parameters;
         try {
@@ -208,13 +289,11 @@ export const bundleSizeCheck = buildTool({
             return err.message;
         }
         const mode = input.join ?? "exact";
-        let custom;
         if (input.joinPattern !== undefined) {
-            try {
-                custom = new RegExp(input.joinPattern, "g");
-            }
-            catch (err) {
-                return `joinPattern /${input.joinPattern}/ is not a valid regular expression: ${err.message}`;
+            // Screened here too: the schema's refine is skipped by a direct call.
+            const issue = patternIssue(input.joinPattern, "g");
+            if (issue !== undefined) {
+                return `joinPattern /${input.joinPattern}/ was refused: ${issue}`;
             }
         }
         let paths;
@@ -247,7 +326,14 @@ export const bundleSizeCheck = buildTool({
         if (paths.length === 0) {
             return "no files matched — there is nothing to weigh, which is not the same as a build of zero bytes";
         }
-        const entries = measure(paths, parameters, mode, custom);
+        const entries = measure(paths, parameters, mode);
+        if (mode === "custom") {
+            const keyed = await customJoinKeys(input.joinPattern, entries.map((e) => e.path), ctx);
+            if (!keyed.ok)
+                return keyed.message;
+            for (const entry of entries)
+                entry.key = keyed.keys.get(entry.path);
+        }
         const totals = entries.reduce((acc, e) => ({
             bytes: acc.bytes + e.bytes,
             compressedBytes: e.compressedBytes === null
@@ -311,9 +397,21 @@ export const bundleSizeCheck = buildTool({
         let baseline = input.baseline;
         if (input.baselineFile !== undefined) {
             const at = resolveSafe("BundleSizeCheck", input.baselineFile);
+            // Opened without blocking, and only as a regular file: a FIFO with no
+            // writer blocks an ordinary open for ever, and this read was
+            // synchronous, so a named pipe named as the baseline stopped the whole
+            // harness (C074). A baseline is a size report, so it is capped too.
+            const read = openForReadSync(workspaceRoot(), input.baselineFile, {
+                maxBytes: LIMITS.baselineBytes,
+            });
+            if (!read.ok)
+                return `baseline file could not be read: ${read.reason}`;
+            if (read.truncated) {
+                return `baseline file "${at.rel}" is over the ${LIMITS.baselineBytes}-byte limit for a size report, so it was not read`;
+            }
             let parsed;
             try {
-                parsed = JSON.parse(readFileSync(at.real, "utf-8"));
+                parsed = JSON.parse(read.text);
             }
             catch (err) {
                 return `baseline file "${at.rel}" is not valid JSON: ${err.message}`;
@@ -386,8 +484,17 @@ export const bundleSizeCheck = buildTool({
                 // the baseline happened to list first, which is the arbitrary pairing
                 // the head-side check already refuses to make.
                 const baselineGroups = new Map();
+                let baselineKeys;
+                if (mode === "custom") {
+                    const keyed = await customJoinKeys(input.joinPattern, baseline.entries.map((e) => e.path), ctx);
+                    if (!keyed.ok)
+                        return keyed.message;
+                    baselineKeys = keyed.keys;
+                }
                 for (const entry of baseline.entries) {
-                    const key = joinKey(entry.path, mode, custom).key;
+                    const key = baselineKeys !== undefined
+                        ? baselineKeys.get(entry.path)
+                        : joinKey(entry.path, mode).key;
                     baselineGroups.set(key, [
                         ...(baselineGroups.get(key) ?? []),
                         { path: entry.path, bytes: entry.bytes, compressed: entry.compressedBytes ?? null },
@@ -631,6 +738,69 @@ export const benchmarkCompare = buildTool({
     },
 });
 // --- FlakyTestDetect --------------------------------------------------------
+/** A mask always replaces every match, so `g` is added to its flags. */
+const maskFlags = (mask) => {
+    const flags = mask.flags ?? "";
+    return flags.includes("g") ? flags : `${flags}g`;
+};
+/** Total failure text the masks may run over in one call. */
+const MAX_MASKED_CHARS = 16_000_000;
+/**
+ * Apply the caller's masks to every failure text in @crewhaus/tool-safety's
+ * regex worker, in order, each mask over what the previous one left (C073).
+ * One text per failing observation, so each mask's hit count is summed over
+ * every failure, exactly as the synchronous path counted. A mask that could
+ * not be applied to every text fails the call: grouping some failures by a
+ * masked key and others by a raw one would split one failure mode in two,
+ * which reads as "it fails different ways".
+ */
+async function applyMasksOffThread(runs, masks, ctx) {
+    for (const mask of masks) {
+        const issue = patternIssue(mask.pattern, maskFlags(mask));
+        if (issue !== undefined)
+            return `mask /${mask.pattern}/ was refused: ${issue}`;
+    }
+    const raw = maskSubjects(runs);
+    let texts = [...raw];
+    const counts = {};
+    const session = openRegexSession();
+    try {
+        for (const mask of masks) {
+            if (texts.length === 0)
+                break;
+            const outcome = await session.run({
+                op: "replaceEach",
+                pattern: mask.pattern,
+                flags: maskFlags(mask),
+                inputs: texts,
+                replacement: mask.with ?? "<masked>",
+                onGiveUp: "skip",
+                deadlineMs: 2_000,
+                maxInputChars: MAX_MASKED_CHARS,
+                maxOutputChars: MAX_MASKED_CHARS,
+                ...(ctx?.signal !== undefined ? { signal: ctx.signal } : {}),
+                ...(ctx?.runContext !== undefined ? { runawayKey: ctx.runContext.sessionId } : {}),
+            });
+            if (outcome.status !== "ok") {
+                return `FlakyTestDetect could not apply mask /${mask.pattern}/: ${describeRegexOutcome(outcome)}. The failures were not grouped; simplify the mask, or pass a precomputed \`signature\` per failure.`;
+            }
+            if (outcome.result.undetermined.length > 0) {
+                return `FlakyTestDetect could not apply mask /${mask.pattern}/ to ${outcome.result.undetermined.length} failure text(s): the pattern engine gave up on them. The failures were not grouped; simplify the mask, or pass a precomputed \`signature\` per failure.`;
+            }
+            if (outcome.result.replacements > 0) {
+                counts[mask.pattern] = (counts[mask.pattern] ?? 0) + outcome.result.replacements;
+            }
+            texts = outcome.result.outputs;
+        }
+    }
+    finally {
+        session.close();
+    }
+    const keys = new Map();
+    for (const [i, text] of raw.entries())
+        keys.set(text, foldWhitespace(texts[i] ?? text));
+    return { keys, counts };
+}
 export const flakyTestDetect = buildTool({
     name: "FlakyTestDetect",
     description: "Classify each test across repeated runs as passing, flaky, failing every run or never observed, with a Wilson interval on its failure rate rather than a fail/run ratio. Use it on the results of running a suite several times, to decide what to quarantine. Three failures in five runs is not '60% flaky' — the interval is 23%-88%, which does not distinguish annoying from broken — so the quarantine recommendation keys on the interval's lower bound and says 'run it more' when that bound is not yet clear of the floor. It separates nondeterminism from order dependence when the input carries execution order: for a test that both passed and failed it names the tests that ran before it in every failure and in none of the passes. When every run used the SAME order it reports that the question is undecidable rather than answering 'nondeterministic', and a skipped test is never counted as a pass. One run is always inconclusive.",
@@ -681,7 +851,13 @@ export const flakyTestDetect = buildTool({
             with: z.string().optional(),
             flags: z.string().max(8).optional(),
         })
-            .strict())
+            .strict()
+            .superRefine((mask, ctx) => {
+            const issue = patternIssue(mask.pattern, maskFlags(mask));
+            if (issue !== undefined) {
+                ctx.addIssue({ code: z.ZodIssueCode.custom, path: ["pattern"], message: issue });
+            }
+        }))
             .max(LIMITS.masks)
             .optional()
             .describe("substitutions applied to failure text before grouping; each one's hit count is reported"),
@@ -689,10 +865,17 @@ export const flakyTestDetect = buildTool({
         .strict(),
     readOnly: true,
     concurrencySafe: true,
-    execute: async (input) => {
+    execute: async (input, ctx) => {
         // The kernel's table is the only place a confidence level is defined in
         // this repository; a literal 1.96 here would be a fourth copy.
         const z = statsKernel.TWO_SIDED_Z[input.confidence ?? "0.95"];
+        let maskedText;
+        if (input.masks !== undefined && input.masks.length > 0) {
+            const applied = await applyMasksOffThread(input.runs, input.masks, ctx);
+            if (typeof applied === "string")
+                return applied;
+            maskedText = applied;
+        }
         try {
             return json(detectFlaky(input.runs, {
                 z,
@@ -700,6 +883,7 @@ export const flakyTestDetect = buildTool({
                     ? {}
                     : { quarantineLowerBound: input.quarantineLowerBound }),
                 ...(input.masks === undefined ? {} : { masks: input.masks }),
+                ...(maskedText === undefined ? {} : { maskedText }),
             }));
         }
         catch (err) {

package/dist/lib/flaky.d.ts

@@ -33,6 +33,17 @@ export type FlakyOptions = {
     readonly quarantineLowerBound?: number;
     /** Caller-declared substitutions applied to failure text before grouping. */
     readonly masks?: ReadonlyArray<Mask>;
+    /**
+     * The masks already applied, off this thread: each failure text's
+     * grouping key, and each mask's hit count over every failure. The tool
+     * computes these in @crewhaus/tool-safety's regex worker under a deadline
+     * (C073), so a caller's pattern never runs synchronously here; `masks`
+     * is then only named, not run. A text missing from `keys` is an error.
+     */
+    readonly maskedText?: {
+        readonly keys: ReadonlyMap<string, string>;
+        readonly counts: Readonly<Record<string, number>>;
+    };
 };
 export type FailureGroup = {
     readonly signature: string;
@@ -86,6 +97,15 @@ export type FlakyReport = {
     readonly verdict: "clean" | "flaky" | "failing" | "inconclusive";
     readonly note: string;
 };
+/**
+ * The failure texts masks are applied to, one per failing observation that
+ * carries text and no signature, in input order: exactly the texts
+ * `groupFailures` would mask, so a hit count summed over them is the count
+ * `applyMasks` reports.
+ */
+export declare function maskSubjects(runs: ReadonlyArray<RunInput>): string[];
+/** The whitespace fold every grouping key gets after the masks. */
+export declare const foldWhitespace: (text: string) => string;
 /**
  * Apply the caller's substitutions and report how many times each fired.
  *

package/dist/lib/flaky.js

@@ -19,12 +19,34 @@
  * whole reason a team spends a week chasing the wrong test.
  */
 import { statsKernel } from "@crewhaus/tool-math";
+import { compileUserRegex } from "@crewhaus/tool-safety/regex";
 export const TEST_STATUSES = ["pass", "fail", "error", "skip"];
 export class FlakyError extends Error {
     name = "FlakyError";
 }
 /** A failure is a failure; `error` is one that did not reach an assertion. */
 const isFailure = (status) => status === "fail" || status === "error";
+/**
+ * The failure texts masks are applied to, one per failing observation that
+ * carries text and no signature, in input order: exactly the texts
+ * `groupFailures` would mask, so a hit count summed over them is the count
+ * `applyMasks` reports.
+ */
+export function maskSubjects(runs) {
+    const out = [];
+    for (const run of runs) {
+        for (const test of run.tests) {
+            if (!isFailure(test.status) || test.signature !== undefined)
+                continue;
+            const raw = test.error ?? "";
+            if (raw !== "")
+                out.push(raw);
+        }
+    }
+    return out;
+}
+/** The whitespace fold every grouping key gets after the masks. */
+export const foldWhitespace = (text) => text.replace(/\s+/g, " ").trim();
 /**
  * Apply the caller's substitutions and report how many times each fired.
  *
@@ -44,20 +66,21 @@ export function applyMasks(text, masks, counts) {
     let out = text;
     for (const mask of masks) {
         const flags = mask.flags ?? "";
-        let re;
-        try {
-            re = new RegExp(mask.pattern, flags.includes("g") ? flags : `${flags}g`);
-        }
-        catch (err) {
-            throw new FlakyError(`mask /${mask.pattern}/ is not a valid regular expression: ${err.message}`);
+        // Screened for the catastrophic shapes even here, where the caller is
+        // this package's own code: FlakyTestDetect itself never gets this far
+        // with a caller's mask (see `maskedText`).
+        const compiled = compileUserRegex(mask.pattern, flags.includes("g") ? flags : `${flags}g`);
+        if (!compiled.ok) {
+            throw new FlakyError(`mask /${mask.pattern}/ was refused: ${compiled.reason}`);
         }
+        const re = compiled.regex;
         const fired = out.match(re)?.length ?? 0;
         if (fired > 0) {
             out = out.replace(re, mask.with ?? "<masked>");
             counts[mask.pattern] = (counts[mask.pattern] ?? 0) + fired;
         }
     }
-    return out.replace(/\s+/g, " ").trim();
+    return foldWhitespace(out);
 }
 /**
  * The smallest run count whose all-failing Wilson lower bound clears
@@ -147,7 +170,7 @@ function orderAnalysis(occurrences, orderVaried, suiteSize) {
             : `${suspects.length} test(s) ran before this one in every failing run and in none of the passing runs; re-run this test after each of them alone to confirm`,
     };
 }
-function groupFailures(occurrences, masks, counts) {
+function groupFailures(occurrences, masks, counts, maskedText) {
     const groups = new Map();
     for (const occurrence of occurrences) {
         if (!isFailure(occurrence.status))
@@ -157,7 +180,7 @@ function groupFailures(occurrences, masks, counts) {
         // saying anything about it; that is not a group of one, it is an unknown.
         if (occurrence.signature === undefined && raw === "")
             continue;
-        const key = occurrence.signature ?? applyMasks(raw, masks, counts);
+        const key = occurrence.signature ?? maskedKey(raw, masks, counts, maskedText);
         const row = groups.get(key);
         if (row === undefined) {
             groups.set(key, { count: 1, example: raw, runs: [occurrence.runId] });
@@ -175,6 +198,15 @@ function groupFailures(occurrences, masks, counts) {
     }))
         .sort((a, b) => b.count - a.count || (a.signature < b.signature ? -1 : 1));
 }
+function maskedKey(raw, masks, counts, maskedText) {
+    if (maskedText === undefined)
+        return applyMasks(raw, masks, counts);
+    const key = maskedText.keys.get(raw);
+    if (key === undefined) {
+        throw new FlakyError("a failure text has no precomputed mask key; its failures cannot be grouped");
+    }
+    return key;
+}
 export function detectFlaky(runs, options = {}) {
     if (runs.length === 0)
         throw new FlakyError("no runs were given; there is nothing to classify");
@@ -237,7 +269,7 @@ export function detectFlaky(runs, options = {}) {
             z,
         };
         const order = orderAnalysis(list, orderVaried, suiteSize);
-        const failureGroups = groupFailures(list, masks, maskCounts);
+        const failureGroups = groupFailures(list, masks, maskCounts, options.maskedText);
         // A failure that recorded no text and no signature is in no group, so one
         // group does not mean one failure mode — it means one failure mode among
         // the failures that said anything. Answering "yes, always the same way"
@@ -336,7 +368,7 @@ export function detectFlaky(runs, options = {}) {
         },
         tests,
         summary,
-        masksApplied: maskCounts,
+        masksApplied: options.maskedText?.counts ?? maskCounts,
         runsForDeterminismClaim: runsForLowerBound(0.9, z),
         verdict,
         note: runs.length < 2

package/dist/paths.d.ts

@@ -1,10 +1,29 @@
 import { CrewhausError } from "@crewhaus/errors";
-/** Raised when a caller-supplied path resolves outside the workspace root. */
+/**
+ * The longest path this resolves, in UTF-16 code units; a longer one is
+ * refused before it is walked. Nothing that long can be opened or created:
+ * PATH_MAX is 1,024 bytes on macOS and 4,096 on Linux, Windows stops at
+ * 32,767 UTF-16 units, and a path's UTF-8 form is never shorter than its
+ * UTF-16 one. The limit is twice that because the part of a path that exists
+ * is itself under PATH_MAX, so the part that does not is over PATH_MAX on its
+ * own: no spelling of the place reaches it, not even one through a symlink
+ * that shortens the part that exists.
+ */
+export declare const MAX_PATH_CHARS: number;
+/**
+ * Raised when a caller-supplied path resolves outside the workspace root, or
+ * is longer than any path can be.
+ */
 export declare class ToolPermissionError extends CrewhausError {
     readonly name = "ToolPermissionError";
     readonly toolName: string;
     readonly path: string;
-    constructor(toolName: string, attemptedPath: string);
+    /**
+     * Why, when it is not that the location escapes the workspace root: a path
+     * longer than any path can be. A caller that rewords the refusal says this.
+     */
+    readonly reason: string | undefined;
+    constructor(toolName: string, attemptedPath: string, reason?: string);
 }
 /**
  * A validated path, in both forms the tools need.

package/dist/paths.js

@@ -11,15 +11,45 @@
 import { lstatSync, readlinkSync, realpathSync } from "node:fs";
 import * as path from "node:path";
 import { CrewhausError } from "@crewhaus/errors";
-/** Raised when a caller-supplied path resolves outside the workspace root. */
+/**
+ * The longest path this resolves, in UTF-16 code units; a longer one is
+ * refused before it is walked. Nothing that long can be opened or created:
+ * PATH_MAX is 1,024 bytes on macOS and 4,096 on Linux, Windows stops at
+ * 32,767 UTF-16 units, and a path's UTF-8 form is never shorter than its
+ * UTF-16 one. The limit is twice that because the part of a path that exists
+ * is itself under PATH_MAX, so the part that does not is over PATH_MAX on its
+ * own: no spelling of the place reaches it, not even one through a symlink
+ * that shortens the part that exists.
+ */
+export const MAX_PATH_CHARS = 2 * (process.platform === "darwin" ? 1_024 : process.platform === "win32" ? 32_767 : 4_096);
+/** A path as a refusal quotes it: whole, unless it is past {@link MAX_PATH_CHARS}. */
+function shownPath(given) {
+    return given.length > MAX_PATH_CHARS ? `${given.slice(0, 64)}…` : given;
+}
+/** Why a path past {@link MAX_PATH_CHARS} is refused: its length and the limit. */
+function tooLong(abs) {
+    return `it is ${abs.length} characters once resolved, over the ${MAX_PATH_CHARS}-character limit: no path that long can be opened or created`;
+}
+/**
+ * Raised when a caller-supplied path resolves outside the workspace root, or
+ * is longer than any path can be.
+ */
 export class ToolPermissionError extends CrewhausError {
     name = "ToolPermissionError";
     toolName;
     path;
-    constructor(toolName, attemptedPath) {
-        super("tool", `tool "${toolName}" rejected path "${attemptedPath}": resolved location escapes the workspace root`);
+    /**
+     * Why, when it is not that the location escapes the workspace root: a path
+     * longer than any path can be. A caller that rewords the refusal says this.
+     */
+    reason;
+    constructor(toolName, attemptedPath, reason) {
+        super("tool", reason === undefined
+            ? `tool "${toolName}" rejected path "${attemptedPath}": resolved location escapes the workspace root`
+            : `tool "${toolName}" rejected path "${shownPath(attemptedPath)}": ${reason}`);
         this.toolName = toolName;
         this.path = attemptedPath;
+        this.reason = reason;
     }
 }
 /** The workspace root, resolved. Every path is contained within it. */
@@ -67,15 +97,7 @@ function nameExists(p) {
 function resolveLocation(target, depth = 0) {
     if (depth > 40)
         throw new Error(`symlink chain at "${target}" is too long to resolve`);
-    let probe = target;
-    const tail = [];
-    while (!nameExists(probe)) {
-        tail.unshift(path.basename(probe));
-        const parent = path.dirname(probe);
-        if (parent === probe)
-            break; // reached the filesystem root
-        probe = parent;
-    }
+    const { probe, tail } = deepestName(target);
     let probeReal;
     try {
         probeReal = realpathSync(probe);
@@ -97,7 +119,41 @@ function resolveLocation(target, depth = 0) {
         const base = realpathSync(path.dirname(probe));
         probeReal = resolveLocation(path.resolve(base, link), depth + 1);
     }
-    return tail.length > 0 ? path.join(probeReal, ...tail) : probeReal;
+    // One string, not one argument per name: Bun's `path.join` keeps only
+    // (argument count mod 65,536) of its arguments.
+    return tail.length > 0 ? path.join(probeReal, tail) : probeReal;
+}
+/**
+ * `target`, an absolute path, split where its deepest prefix that exists as
+ * a NAME ends: that prefix, and the names below it, which do not exist, as
+ * one string ("" when `target` itself exists). This is `path.dirname` and
+ * `path.basename` worked by index on `target` itself: handing them each
+ * shorter prefix in turn copied the rest of the path once per missing name,
+ * the square of a count the caller picks.
+ */
+function deepestName(target) {
+    const top = path.parse(target).root.length;
+    const isSep = (i) => target[i] === "/" || target[i] === path.sep;
+    let probe = target;
+    let start = target.length;
+    let stop = -1;
+    while (!nameExists(probe)) {
+        // `path.basename(probe)` is `target.slice(start, end)`, and
+        // `path.dirname(probe)` is `target.slice(0, up)`.
+        let end = probe.length;
+        while (end > top && isSep(end - 1))
+            end--;
+        if (stop < 0)
+            stop = end;
+        start = end;
+        while (start > top && !isSep(start - 1))
+            start--;
+        const up = start > top ? start - 1 : top;
+        if (up === probe.length)
+            break; // reached the filesystem root
+        probe = target.slice(0, up);
+    }
+    return { probe, tail: stop < 0 ? "" : target.slice(start, stop) };
 }
 export function resolveSafe(toolName, rel, root = workspaceRoot()) {
     const rootResolved = path.resolve(root);
@@ -107,6 +163,9 @@ export function resolveSafe(toolName, rel, root = workspaceRoot()) {
     if (abs !== rootResolved && !abs.startsWith(`${rootResolved}${path.sep}`)) {
         throw new ToolPermissionError(toolName, rel);
     }
+    // Longer than any path can be: refused before the walk below reads it.
+    if (abs.length > MAX_PATH_CHARS)
+        throw new ToolPermissionError(toolName, rel, tooLong(abs));
     let real;
     try {
         const rootReal = realpathSync(rootResolved);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@crewhaus/tool-buildperf",
-  "version": "0.7.0",
+  "version": "0.8.0",
   "type": "module",
   "description": "Deterministic build-performance gates: bundle size, benchmark comparison and flaky-test detection against a baseline",
   "main": "dist/index.js",
@@ -8,17 +8,19 @@
   "exports": {
     ".": {
       "types": "./dist/index.d.ts",
-      "import": "./dist/index.js"
+      "import": "./dist/index.js",
+      "default": "./dist/index.js"
     }
   },
   "scripts": {
     "test": "bun test src"
   },
   "dependencies": {
-    "@crewhaus/errors": "0.7.0",
-    "@crewhaus/tool-builder": "0.7.0",
-    "@crewhaus/tool-catalog": "0.7.0",
-    "@crewhaus/tool-math": "0.7.0",
+    "@crewhaus/errors": "0.8.0",
+    "@crewhaus/tool-builder": "0.8.0",
+    "@crewhaus/tool-catalog": "0.8.0",
+    "@crewhaus/tool-math": "0.8.0",
+    "@crewhaus/tool-safety": "0.8.0",
     "zod": "^3.23.8"
   },
   "devDependencies": {
@@ -42,5 +44,8 @@
   "publishConfig": {
     "access": "public"
   },
-  "files": ["dist", "README.md", "LICENSE", "NOTICE"]
+  "files": ["dist", "README.md", "LICENSE", "NOTICE"],
+  "engines": {
+    "bun": ">=1.3.14"
+  }
 }