flakelab 0.1.0 → 0.1.1

package/README.md

@@ -18,34 +18,93 @@ failure is rare:
 ```bash
 npx flakelab@latest tests/checkout.spec.ts --runs 20
 npx flakelab@latest tests/checkout.spec.ts --runs 20 --verbose
+npx flakelab@latest scan tests/checkout.spec.ts --runs 20 --json
 ```
 
-The default command repeatedly executes the selected target in isolated Playwright processes,
-distinguishes intermittent failures from consistent test failures, prints a concise result, and
-writes machine-readable evidence to `.flakelab/runs/scan.json`. It does not call Groq or Solari.
+The default command starts one Playwright process and uses Playwright's native `--repeat-each`
+and JSON reporter to classify every selected test as `no-failure-observed`, `mixed-outcomes`,
+`failed-every-run`, `skipped`, or `errored`. These labels describe only the bounded sample; they
+do not claim that a test is permanently stable. The human summary includes the 80% Wilson
+confidence interval—or the upper bound when no failure was observed—and writes machine-readable evidence to
+`.flakelab/runs/scan.json`. The artifact contains direct run totals, per-test identities,
+confidence bounds, and bounded failure-signature clusters. Each cluster records its observed
+rate, first and last attempt, a redacted representative reason, and up to three representative
+trace, screenshot, or video paths. Multiple independent failure modes are called out explicitly.
+The scan does not call
+Groq or Solari. Use `--json` when another tool needs the scan artifact on stdout; progress remains
+on stderr so the JSON stream stays parseable. A scan exits `0` when it observes no failure, `1`
+when it observes mixed outcomes or failure in every run, and `2` when infrastructure makes the
+result inconclusive.
+
+For a large target, the terminal summary lists problematic tests first and collapses clean
+observations into a count. The saved artifact, `--json`, and `--verbose` retain every test record.
+
+Paths stored in scan evidence are relative to the scanned project. Outside-project paths are
+reduced to a non-identifying placeholder and basename, and the JSON does not record its own output
+location. Playwright directories containing only `.last-run.json` are removed after the scan.
+Evidence-bearing directories receive a FlakeLab ownership marker and only the five newest are
+retained; unmarked directories are always treated as user-owned and are never removed.
+
+Every scan target is deliberate: `flakelab .` is the shortest whole-project form, while the
+spelled-out `flakelab scan` command requires its target and never silently substitutes `.`.
+
+## Analyze an existing CI failure
+
+Point FlakeLab at a Playwright blob-report archive or a directory of shard archives:
 
-For the complete causal workflow, use FlakeLab's Playwright fixture in tests that should accept
-controlled browser faults:
-
-```ts
-import { expect, test } from "flakelab/playwright"
+```bash
+npx flakelab@latest analyze ./blob-report
+npx flakelab@latest analyze ./blob-report --baseline .flakelab/runs/analyze.json --json
 ```
 
+Analysis does not rerun tests and needs no provider credentials. FlakeLab copies and validates
+the archives inside an owned working directory, asks Playwright's supported report merger for
+the canonical JSON report, and leaves the source reports untouched. Unsafe paths, links,
+malformed ZIP structures, excessive expansion, and unbounded archive sets are rejected before
+the merger runs.
+
+The terminal summary ranks mixed outcomes, recurring failures, diagnostic attachments, and
+failure modes not present in the optional baseline. The complete result is written to
+`.flakelab/runs/analyze.json`; `--json` emits the same validated artifact on stdout. Referenced
+traces, screenshots, and videos are retained with the owned analysis evidence under bounded
+retention, while temporary archive copies and conversion files are removed.
+
+The scan explicitly disables Playwright retries so a retry is never mistaken for an independent
+statistical repetition. It uses one Playwright worker by default because arbitrary test suites may
+share ports, databases, caches, or workspace files. Use `--concurrency` greater than `1` only when
+those resources are safe for normal Playwright worker parallelism. Traces, screenshots, videos,
+and ordinary test output stay in a unique directory beneath the selected evidence directory;
+FlakeLab removes only its temporary JSON reporter file.
+
+The native JSON reporter contract is tested against `@playwright/test` 1.62.1, which is the current
+peer requirement. The integration fixture covers repetitions, multiple projects, expected
+failures, skips, timeouts, file-backed attachments, and paths containing spaces.
+
+The causal workflow works with ordinary Playwright tests. Keep importing `test` and `expect` from
+`@playwright/test`; no per-test FlakeLab fixture or source edit is required. For each experiment,
+FlakeLab creates a temporary config beside the existing Playwright config, retains its projects,
+dependencies, web server, authentication settings, application environment, and reporters, and
+adds a local fault proxy through Playwright's public configuration. The generated config and proxy
+are removed after the trial, including after failure or interruption. The current project-level
+bridge injects matching HTTP requests; it reports an explicit unsupported-path error instead of
+claiming success when a selected route is HTTPS or no request matches.
+
 Then explicitly opt into AI investigation and isolated proof:
 
 ```bash
-npx flakelab@latest tests/checkout.spec.ts \
-  --prove \
+npx flakelab@latest prove tests/checkout.spec.ts \
+  --fault network-delay \
   --pattern "**/api/checkout" \
   --open
 ```
 
-`--prove` requires `GROQ_API_KEY` and `SOLARI_API_KEY`. It chains trigger discovery, reproducer
+The equivalent target shortcut is `npx flakelab@latest tests/checkout.spec.ts --prove`.
+Proof mode requires `GROQ_API_KEY` and `SOLARI_API_KEY`. It chains trigger discovery, reproducer
 verification, bounded AI investigation, candidate generation, proof in a disposable Solari
 microVM, and the offline evidence report. The candidate remains a reviewable diff and is never
 applied automatically.
 
-See every command and default without running tests:
+See every command and its accepted options without running tests:
 
 ```bash
 npx flakelab@latest --help
@@ -61,18 +120,56 @@ The doctor checks Node.js, Playwright configuration, Chromium, ignored evidence
 credential availability, and test-process credential isolation. It reports only whether a
 credential is configured and where it came from; it never prints the value or a fingerprint.
 
-The local diagnostic core can already run controlled baseline and fault-injected trials:
+Use `diagnose` as the adaptive product journey. A target begins with the bounded native scan;
+an existing CI report begins with read-only analysis:
 
 ```bash
-pnpm flakelab diagnose tests/fixtures/flaky-checkout.spec.ts \
-  --runs 4 \
-  --seed 42 \
-  --delay-ms 250
+npx flakelab@latest diagnose tests/checkout.spec.ts
+npx flakelab@latest diagnose --report ./blob-report
+```
+
+The command explains what was observed and writes `.flakelab/runs/diagnose.json` with the cheapest
+useful next experiment, its trial bound, an expected duration based on the local scan when
+available, required credentials, and the Solari estimate—or an explicit statement that no
+reliable estimate is available. Local diagnosis never requests provider credentials.
+
+Continue only as far as the evidence justifies:
+
+```bash
+npx flakelab@latest diagnose tests/checkout.spec.ts --discover
+npx flakelab@latest diagnose tests/checkout.spec.ts --investigate
+npx flakelab@latest diagnose tests/checkout.spec.ts --repair
+```
+
+`--discover` remains local and creates a minimized reproducer. `--investigate` explicitly enables
+bounded Groq usage and retains both the reproducer and investigation evidence. `--repair`
+explicitly enables candidate generation and a disposable Solari proof; a rejected candidate still
+produces the portable evidence report. When starting from `--report`, supply a test target before
+requesting new experiments so FlakeLab never guesses which test to execute.
+
+Every diagnosis writes `diagnose.json` atomically after each completed phase and when work fails
+or is interrupted. Continue the exact saved workflow with:
+
+```bash
+npx flakelab@latest resume .flakelab/runs/diagnose.json
 ```
 
-The command runs alternating control and network-delay trials, prints their failure rates,
-and writes a validated NDJSON event stream under `.flakelab/runs/`. The included checkout
-fixture passes normally and fails reliably when its checkout request is delayed.
+Resume validates the checkpoint and its original input hash, confines saved paths to the current
+project, and starts at the next safe phase. It does not repeat a confirmed scan, discovery, or
+investigation. Completed checkpoints are a no-op apart from printing their summary. Partial
+checkpoints remain readable and are labelled `running`, `interrupted`, or `failed`; they never
+masquerade as a completed result.
+
+The checkpoint shows planned and actual trial counts, wall time, AI token usage and estimated
+cost, Solari sandbox creation and release counts, snapshot-cache decisions, and cleanup status.
+It stores normalized options and project-relative artifact paths, not credentials. `Ctrl+C` stops
+new trials; local Playwright process trees and disposable remote resources are cleaned up by the
+same bounded execution paths used outside diagnosis.
+
+Prepared Solari demo snapshots are reused only when a key covering the commit, lockfile, fixture,
+template, runtime command, application path, port, and timeout matches. The metrics artifact gives
+the key and a reason for every hit or miss. Candidate repair workspaces are intentionally not
+cached because each contains a unique patch; their checkpoint explains that decision.
 
 ## Discover and replay a minimal reproducer
 
@@ -81,25 +178,95 @@ failure-rate and confidence thresholds:
 
 ```bash
 pnpm flakelab discover tests/fixtures/flaky-checkout.spec.ts \
+  --fault network-delay \
   --trials 4 \
-  --concurrency 4 \
+  --concurrency 1 \
   --seed 42 \
   --max-delay 125 \
+  --max-seconds 300 \
   --min-rate 0.7 \
   --output flakelab.repro.yaml
 ```
 
 The command writes a strict portable YAML reproducer and a JSON discovery sidecar containing
-the evidence for every candidate. It normalizes Playwright failures into stable signatures,
-uses an 80% Wilson lower confidence bound, and accepts a trigger only when both the observed
-rate and lower bound meet the requested minimum. Replay the result with:
+the evidence for every candidate. Every candidate alternates matched-seed control and intervention
+trials. A pre-existing failure is allowed, but the intervention is causal only when the same
+normalized failure signature reaches the requested rate and its 80% lower confidence bound exceeds
+the control's upper bound. Discovery prints each completed trial to stderr
+and stops at the five-minute default elapsed-time ceiling; set `--max-seconds` explicitly when a
+suite needs a different local budget. Replay the result with:
 
 ```bash
-pnpm flakelab replay flakelab.repro.yaml --concurrency 4
+pnpm flakelab replay flakelab.repro.yaml --concurrency 1
 ```
 
-The current browser-fault matrix supports deterministic network delay and injected HTTP
-failures. Fault routes are removed after every trial, including failed tests.
+The current project-level fault matrix supports deterministic network delay, type-aware resource
+loading delay, injected HTTP failures, response truncation, response-tail duplication, and response
+reordering. Minimize
+malformed payload triggers with `--fault response-truncation --max-remove-bytes 1024` or
+`--fault response-duplication --max-duplicate-bytes 1024`. Find request races with
+`--fault response-reordering --max-hold-ms 250`; the first response in each adjacent matching pair
+is held while the second proceeds. Response mutations run in the order stored in the bounded fault
+set, so a reproducer can express and replay an exact composition without introducing another schema.
+Isolate startup dependencies with
+`--fault resource-loading-delay --resource-type script --pattern "**/assets/*" --max-delay 250`.
+Resource loading can target `document`, `script`, `stylesheet`, `image`, or `font` without slowing
+unrelated API traffic. Discovery reports the minimum observed timing boundary and saves a
+stability-margin trigger between that boundary and the configured maximum, so the portable
+reproducer is less sensitive to ordinary machine jitter.
+Delay application lifecycle listeners independently with
+`--fault startup-event-delay --startup-event dom-content-loaded --pattern "**/checkout"`.
+The fault supports `dom-content-loaded` and `load`, injects a temporary same-origin bootstrap into
+matching HTML, preserves an existing script nonce, and verifies that the bootstrap actually ran.
+It delays application listeners rather than the browser lifecycle event itself, network traffic,
+or the event loop.
+Probe missed main-thread deadlines with
+`--fault event-loop-stall --max-stall-ms 500 --stall-after-ms 0 --pattern "**/checkout"`.
+The offset is measured from the native `DOMContentLoaded` event and the stall duration is bounded
+at two seconds. Discovery finds the minimum observed duration, confirms it with repeated trials,
+and saves a stability-margin reproducer. The fault blocks only the matching page's main thread;
+it does not simulate system-wide CPU saturation.
+Test expired session handling without recording credentials with
+`--fault auth-cookie-expiry --cookie-name session-id --pattern "**/api/session"`.
+Only the named cookie is withheld from matched requests, and an absent cookie is reported as an
+unapplied fault rather than false causal evidence. Reproducers contain the cookie name but never
+its value. The direct browser adapter restores the original cookie in `finally`.
+Test partially initialized browser state with
+`--fault storage-state-delay --storage local-storage --storage-key auth-token --max-delay 500`
+and a document URL pattern. The bootstrap temporarily makes `getItem` return `null` only for that
+key; it does not delete or record the stored value. Both local and session storage are supported,
+and discovery minimizes the visibility delay before saving a stability-margin trigger.
+Expose wall-clock assumptions with
+`--fault clock-jump --clock-offset-ms 3600000 --jump-after-ms 25 --pattern "**/app"`.
+The fault changes `Date` wall time after a deterministic offset while leaving `performance.now`,
+timeouts, and interval scheduling monotonic. Use `--clock-offset-ms=-3600000` for a backward jump.
+Exercise internationalization assumptions with
+`--fault locale --locale fr-FR --pattern "**/app"` or
+`--fault timezone --timezone America/New_York --pattern "**/app"`. Locale tags and IANA timezone
+names are validated before trials. These two settings are applied through Playwright's browser
+context configuration; the document pattern verifies that the intended target was reached. The
+direct page adapter rejects them explicitly because an existing Playwright context cannot be
+reconfigured faithfully after creation.
+Probe responsive and accessibility-specific behavior with
+`--fault viewport --viewport-width 375 --viewport-height 667` and
+`--fault reduced-motion`. Viewports are bounded from 200 to 7680 pixels in each dimension, and
+reduced motion uses the browser's native media emulation. Test animation races with
+`--fault animation-speed --animation-rate 5`; rates are bounded from 0.1x to 10x and alter Web
+Animations and CSS animation/transition playback without accelerating application timers. Each
+visual fault is reversible, uses a paired control, and requires a matched document before FlakeLab
+accepts its causal evidence.
+Expose suite-level isolation bugs with
+`--fault worker-pressure --max-workers 4`. FlakeLab raises Playwright's supported worker limit,
+enables full parallel scheduling for the selected target, and saves the smallest worker count that
+reproduces the failure. Probe shared accounts, records, ports, and other cross-test state with
+`--fault shared-state-interference --max-copies 4`; this overlaps repeated copies of the selected
+test through Playwright's supported `repeatEach` and worker controls. Both faults use matched-seed
+single-worker controls, are bounded to 16 workers or copies, require an independent 12-trial
+confirmation, and store the selected test path—not a URL glob—as their artifact target.
+The local proxy and generated Playwright config are removed after every trial, including failed and
+interrupted tests. Solari repair proof consumes the same validated fault set as local replay.
+Revision bisect still rejects unsupported fault combinations explicitly rather than approximating
+remote behavior.
 
 ## AI investigator
 
@@ -109,7 +276,7 @@ terminal requests it through hidden input and keeps it in memory for this run on
 ```bash
 pnpm flakelab investigate tests/fixtures/flaky-checkout.spec.ts \
   --trials 4 \
-  --concurrency 4 \
+  --concurrency 1 \
   --max-delay 125
 ```
 
@@ -168,17 +335,24 @@ tree:
 ```bash
 pnpm flakelab repair flakelab.investigation.json \
   --reproducer flakelab.repro.yaml \
+  --source src/checkout.ts \
+  --max-cost 0.25 \
   --patch candidate.diff \
   --proof flakelab.proof.json
 ```
 
-This command requires both `GROQ_API_KEY` and `SOLARI_API_KEY`. The model can propose only exact,
-bounded edits to application source it previously received. FlakeLab rejects test changes,
+This command requires both `GROQ_API_KEY` and `SOLARI_API_KEY`. `--max-cost` bounds candidate
+generation, and the result reports observed model tokens and estimated cost. Repeat `--source` to
+approve up to seven application source files that a black-box Playwright test does not import.
+Every approved file remains project-confined, JavaScript/TypeScript-only, credential-blocked, and
+inside the shared 64 KiB context limit. The model can propose only exact, bounded edits to source it
+received. FlakeLab rejects test changes,
 assertion weakening, lint suppressions, credential-like additions, path escapes, and numeric-only
 timeout increases before execution.
 
 The candidate is copied into a disposable Solari microVM. Typecheck, lint, hostile trials, clean
-controls, and nearby regression tests all run there. The machine is destroyed afterward, and the
+controls, and up to 50 nested or co-located `.spec.*` and `.test.*` regression files all run there.
+The machine is destroyed afterward, and the
 candidate is returned as a reviewable diff; FlakeLab never applies it to the local checkout.
 Cold validation installs the pinned Node/pnpm toolchain and Chromium, so it is intentionally
 slower than the future snapshot-backed warm path.
@@ -199,12 +373,19 @@ pnpm flakelab report flakelab.investigation.json \
 The React/Recharts interface is bundled by Vite into a single HTML file. It explains the root
 cause, deterministic ownership classification, experiment timeline, competing hypotheses,
 minimal trigger, before/after proof matrix, static checks, model usage, and reviewable artifacts.
+Each causal conclusion links to both its intervention and clean control. Investigation mode retains
+project-relative Playwright traces, compares one representative passing control with one failing
+intervention, and links each recording from its experiment. The report also gives an exact replay
+command, the bounded source files shown to the investigator, and the line of every candidate edit.
 The evidence is schema-validated and credential-redacted before rendering. A restrictive content
-security policy blocks runtime network access, so the report remains usable offline.
+security policy blocks runtime network access, unsafe evidence links are rejected, and the report
+remains usable offline.
 
-`--open` launches only the local file. `--publish` is optional and always asks for interactive
-confirmation immediately before creating a public Solari preview. Published reports expire and
-their hosting sandbox is automatically killed after at most 60 minutes.
+Interactive report generation offers to open the local file in the default browser. `--open`
+accepts that choice without prompting, while CI and redirected sessions never prompt or open a
+browser implicitly. `--publish` is optional and always asks for interactive confirmation
+immediately before creating a public Solari preview. Published reports expire and their hosting
+sandbox is automatically killed after at most 60 minutes.
 
 ## Statistical Git bisect
 
@@ -264,8 +445,14 @@ Playwright configuration, manifest, or lockfile behavior changes, it falls back
 to keep time and cost predictable; `.flakelab/changed-tests.json` preserves the complete selection.
 
 The uploaded artifact is retained for seven days and includes the offline HTML report,
-investigation, reproducer, discovery evidence, proof, candidate diff, and selection manifest. A
-rejected repair still uploads its evidence before the job reports failure.
+investigation traces, reproducer, discovery evidence, proof, candidate diff, and selection manifest.
+A rejected repair still uploads its evidence before the job reports failure.
+
+The same composite action can analyze an existing Playwright blob report without Groq or Solari
+credentials. Pass `blob-report` and, optionally, an artifact downloaded from an earlier run as
+`baseline`. The action does not rerun tests in this mode; it merges shards, ranks failures, marks
+signatures absent from the baseline, uploads the validated analysis, and writes a concise Markdown
+job summary. No hosted database is required.
 
 ## Parallel Solari runner
 

package/dist/analysis/blob-report.d.ts

@@ -0,0 +1,15 @@
+import type { ScanTestResult } from "../scan/schema.js";
+import type { AnalysisSourceKind } from "./schema.js";
+export interface BlobReportAnalysisOptions {
+    artifactDirectory: string;
+    playwrightCliPath?: string;
+    signal?: AbortSignal;
+}
+export interface BlobReportAnalysisResult {
+    archiveCount: number;
+    artifactDirectory: string | null;
+    runnerErrors: string[];
+    sourceKind: AnalysisSourceKind;
+    tests: ScanTestResult[];
+}
+export declare function analyzeBlobReports(projectRoot: string, source: string, options: BlobReportAnalysisOptions): Promise<BlobReportAnalysisResult>;

package/dist/analysis/blob-report.js

@@ -0,0 +1,137 @@
+import { spawn } from "node:child_process";
+import { copyFile, lstat, mkdir, mkdtemp, readFile, readdir, rm } from "node:fs/promises";
+import { basename, extname, isAbsolute, join, relative, resolve } from "node:path";
+import { retainOwnedArtifacts } from "../artifacts/retention.js";
+import { redactText } from "../report/redaction.js";
+import { parseNativeScanReport } from "../runner/native-scan.js";
+import { createPlaywrightEnvironment, resolvePlaywrightCliPath, } from "../runner/playwright-executor.js";
+import { waitForProcessTree } from "../runner/process-tree.js";
+import { inspectZipArchive } from "./zip-archive.js";
+const MAX_ARCHIVES = 100;
+const MAX_TOTAL_EXPANDED_BYTES = 1_024 * 1_024 * 1_024;
+async function discoverInput(sourcePath) {
+    const information = await lstat(sourcePath);
+    if (information.isSymbolicLink()) {
+        throw new Error("Blob report input cannot be a symbolic link");
+    }
+    if (information.isFile()) {
+        if (extname(sourcePath).toLowerCase() !== ".zip") {
+            throw new Error("Blob report file must use the .zip extension");
+        }
+        return { archives: [sourcePath], kind: "blob-archive" };
+    }
+    if (!information.isDirectory()) {
+        throw new Error("Blob report input must be a ZIP archive or directory");
+    }
+    const entries = await readdir(sourcePath, { withFileTypes: true });
+    const archiveEntries = entries.filter((entry) => extname(entry.name).toLowerCase() === ".zip");
+    if (archiveEntries.some((entry) => entry.isSymbolicLink() || !entry.isFile())) {
+        throw new Error("Blob report directory contains a non-regular ZIP entry");
+    }
+    const archives = archiveEntries
+        .map((entry) => resolve(sourcePath, entry.name))
+        .sort((left, right) => left.localeCompare(right));
+    if (archives.length === 0) {
+        throw new Error("Blob report directory does not contain any ZIP archives");
+    }
+    return { archives, kind: "blob-directory" };
+}
+async function validateArchives(archives) {
+    if (archives.length > MAX_ARCHIVES) {
+        throw new Error(`Blob report input exceeds the ${MAX_ARCHIVES}-archive safety limit`);
+    }
+    let expandedBytes = 0;
+    for (const archive of archives) {
+        const summary = await inspectZipArchive(archive);
+        expandedBytes += summary.expandedBytes;
+        if (expandedBytes > MAX_TOTAL_EXPANDED_BYTES) {
+            throw new Error("Blob reports expand beyond the 1 GiB combined safety limit");
+        }
+    }
+}
+function mergeArguments(playwrightCliPath, inputDirectory) {
+    return [playwrightCliPath, "merge-reports", "--reporter=json", inputDirectory];
+}
+async function copyArchives(archives, directory) {
+    for (const [index, archive] of archives.entries()) {
+        const destination = join(directory, `${String(index + 1).padStart(3, "0")}-${basename(archive)}`);
+        await copyFile(archive, destination);
+    }
+}
+function assertDirectChild(parent, directory) {
+    const child = relative(resolve(parent), resolve(directory));
+    if (!child || child.startsWith("..") || isAbsolute(child) || child.includes("/") || child.includes("\\")) {
+        throw new Error("Refusing to clean an analysis path outside its owned parent");
+    }
+}
+async function removeWorkDirectory(parent, directory) {
+    assertDirectChild(parent, directory);
+    await rm(directory, { force: true, recursive: true });
+}
+function mergeFailure(diagnostic) {
+    const safeDiagnostic = redactText(diagnostic).trim().slice(-2_000);
+    return new Error(safeDiagnostic || "Playwright could not merge the blob reports");
+}
+async function removeConversionFiles(directory) {
+    const entries = await readdir(directory, { withFileTypes: true });
+    const files = entries.filter((entry) => entry.isFile() || entry.isSymbolicLink());
+    await Promise.all(files.map(async (entry) => rm(join(directory, entry.name), { force: true })));
+}
+export async function analyzeBlobReports(projectRoot, source, options) {
+    const sourcePath = resolve(projectRoot, source);
+    const input = await discoverInput(sourcePath);
+    const workParent = resolve(projectRoot, options.artifactDirectory, "playwright-analysis");
+    await mkdir(workParent, { recursive: true });
+    const workDirectory = await mkdtemp(join(workParent, "run-"));
+    let cleanupComplete = false;
+    try {
+        await copyArchives(input.archives, workDirectory);
+        const archiveCopies = (await readdir(workDirectory, { withFileTypes: true }))
+            .filter((entry) => entry.isFile() && extname(entry.name).toLowerCase() === ".zip")
+            .map((entry) => resolve(workDirectory, entry.name))
+            .sort((left, right) => left.localeCompare(right));
+        await validateArchives(archiveCopies);
+        const reportPath = join(workDirectory, "flakelab-merged.json");
+        const playwrightCliPath = options.playwrightCliPath ?? resolvePlaywrightCliPath(projectRoot);
+        const child = spawn(process.execPath, mergeArguments(playwrightCliPath, workDirectory), {
+            cwd: projectRoot,
+            detached: process.platform !== "win32",
+            env: {
+                ...createPlaywrightEnvironment(),
+                PLAYWRIGHT_JSON_OUTPUT_FILE: reportPath,
+            },
+            shell: false,
+            windowsHide: true,
+        });
+        const execution = await waitForProcessTree(child, options.signal);
+        if (options.signal?.aborted) {
+            throw new Error("Playwright blob report analysis was interrupted");
+        }
+        if (execution.spawnError || execution.exitCode !== 0) {
+            throw mergeFailure(execution.spawnError ?? execution.diagnostic);
+        }
+        let parsed;
+        try {
+            parsed = parseNativeScanReport(await readFile(reportPath, "utf8"), projectRoot);
+        }
+        catch {
+            throw new Error("Playwright produced a missing or malformed merged JSON report");
+        }
+        await removeConversionFiles(workDirectory);
+        const artifactDirectory = await retainOwnedArtifacts(projectRoot, workDirectory);
+        cleanupComplete = true;
+        return {
+            archiveCount: input.archives.length,
+            artifactDirectory,
+            runnerErrors: parsed.runnerErrors,
+            sourceKind: input.kind,
+            tests: parsed.tests,
+        };
+    }
+    finally {
+        if (!cleanupComplete) {
+            await removeWorkDirectory(workParent, workDirectory);
+        }
+    }
+}
+//# sourceMappingURL=blob-report.js.map

package/dist/analysis/blob-report.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"blob-report.js","sourceRoot":"","sources":["../../src/analysis/blob-report.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAE,MAAM,oBAAoB,CAAA;AAC1C,OAAO,EAAE,QAAQ,EAAE,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,EAAE,EAAE,MAAM,kBAAkB,CAAA;AACzF,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,UAAU,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,WAAW,CAAA;AAElF,OAAO,EAAE,oBAAoB,EAAE,MAAM,2BAA2B,CAAA;AAChE,OAAO,EAAE,UAAU,EAAE,MAAM,wBAAwB,CAAA;AACnD,OAAO,EAAE,qBAAqB,EAAE,MAAM,0BAA0B,CAAA;AAChE,OAAO,EACL,2BAA2B,EAC3B,wBAAwB,GACzB,MAAM,kCAAkC,CAAA;AACzC,OAAO,EAAE,kBAAkB,EAAE,MAAM,2BAA2B,CAAA;AAG9D,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAA;AAEpD,MAAM,YAAY,GAAG,GAAG,CAAA;AACxB,MAAM,wBAAwB,GAAG,KAAK,GAAG,KAAK,GAAG,KAAK,CAAA;AAqBtD,KAAK,UAAU,aAAa,CAAC,UAAkB;IAC7C,MAAM,WAAW,GAAG,MAAM,KAAK,CAAC,UAAU,CAAC,CAAA;IAC3C,IAAI,WAAW,CAAC,cAAc,EAAE,EAAE,CAAC;QACjC,MAAM,IAAI,KAAK,CAAC,6CAA6C,CAAC,CAAA;IAChE,CAAC;IACD,IAAI,WAAW,CAAC,MAAM,EAAE,EAAE,CAAC;QACzB,IAAI,OAAO,CAAC,UAAU,CAAC,CAAC,WAAW,EAAE,KAAK,MAAM,EAAE,CAAC;YACjD,MAAM,IAAI,KAAK,CAAC,8CAA8C,CAAC,CAAA;QACjE,CAAC;QACD,OAAO,EAAE,QAAQ,EAAE,CAAC,UAAU,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,CAAA;IACzD,CAAC;IACD,IAAI,CAAC,WAAW,CAAC,WAAW,EAAE,EAAE,CAAC;QAC/B,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAA;IACzE,CAAC;IACD,MAAM,OAAO,GAAG,MAAM,OAAO,CAAC,UAAU,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,CAAC,CAAA;IAClE,MAAM,cAAc,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,WAAW,EAAE,KAAK,MAAM,CAAC,CAAA;IAC9F,IAAI,cAAc,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,cAAc,EAAE,IAAI,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,CAAC;QAC9E,MAAM,IAAI,KAAK,CAAC,wDAAwD,CAAC,CAAA;IAC3E,CAAC;IACD,MAAM,QAAQ,GAAG,cAAc;SAC5B,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,CAAC,UAAU,EAAE,KAAK,CAAC,IAAI,CAAC,CAAC;SAC/C,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,KAAK,CAAC,CAAC,CAAA;IACnD,IAAI,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,yDAAyD,CAAC,CAAA;IAC5E,CAAC;IACD,OAAO,EAAE,QAAQ,EAAE,IAAI,EAAE,gBAAgB,EAAE,CAAA;AAC7C,CAAC;AAED,KAAK,UAAU,gBAAgB,CAAC,QAAkB;IAChD,IAAI,QAAQ,CAAC,MAAM,GAAG,YAAY,EAAE,CAAC;QACnC,MAAM,IAAI,KAAK,CAAC,iCAAiC,YAAY,uBAAuB,CAAC,CAAA;IACvF,CAAC;IACD,IAAI,aAAa,GAAG,CAAC,CAAA;IACrB,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;QAC/B,MAAM,OAAO,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,CAAA;QAChD,aAAa,IAAI,OAAO,CAAC,aAAa,CAAA;QACtC,IAAI,aAAa,GAAG,wBAAwB,EAAE,CAAC;YAC7C,MAAM,IAAI,KAAK,CAAC,4DAA4D,CAAC,CAAA;QAC/E,CAAC;IACH,CAAC;AACH,CAAC;AAED,SAAS,cAAc,CAAC,iBAAyB,EAAE,cAAsB;IACvE,OAAO,CAAC,iBAAiB,EAAE,eAAe,EAAE,iBAAiB,EAAE,cAAc,CAAC,CAAA;AAChF,CAAC;AAED,KAAK,UAAU,YAAY,CAAC,QAAkB,EAAE,SAAiB;IAC/D,KAAK,MAAM,CAAC,KAAK,EAAE,OAAO,CAAC,IAAI,QAAQ,CAAC,OAAO,EAAE,EAAE,CAAC;QAClD,MAAM,WAAW,GAAG,IAAI,CAAC,SAAS,EAAE,GAAG,MAAM,CAAC,KAAK,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,GAAG,CAAC,IAAI,QAAQ,CAAC,OAAO,CAAC,EAAE,CAAC,CAAA;QACjG,MAAM,QAAQ,CAAC,OAAO,EAAE,WAAW,CAAC,CAAA;IACtC,CAAC;AACH,CAAC;AAED,SAAS,iBAAiB,CAAC,MAAc,EAAE,SAAiB;IAC1D,MAAM,KAAK,GAAG,QAAQ,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC,CAAA;IAC3D,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC;QACzG,MAAM,IAAI,KAAK,CAAC,6DAA6D,CAAC,CAAA;IAChF,CAAC;AACH,CAAC;AAED,KAAK,UAAU,mBAAmB,CAAC,MAAc,EAAE,SAAiB;IAClE,iBAAiB,CAAC,MAAM,EAAE,SAAS,CAAC,CAAA;IACpC,MAAM,EAAE,CAAC,SAAS,EAAE,EAAE,KAAK,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAA;AACvD,CAAC;AAED,SAAS,YAAY,CAAC,UAAkB;IACtC,MAAM,cAAc,GAAG,UAAU,CAAC,UAAU,CAAC,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,KAAK,CAAC,CAAA;IAClE,OAAO,IAAI,KAAK,CAAC,cAAc,IAAI,6CAA6C,CAAC,CAAA;AACnF,CAAC;AAED,KAAK,UAAU,qBAAqB,CAAC,SAAiB;IACpD,MAAM,OAAO,GAAG,MAAM,OAAO,CAAC,SAAS,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,CAAC,CAAA;IACjE,MAAM,KAAK,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,MAAM,EAAE,IAAI,KAAK,CAAC,cAAc,EAAE,CAAC,CAAA;IACjF,MAAM,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE,CAAC,EAAE,CAAC,IAAI,CAAC,SAAS,EAAE,KAAK,CAAC,IAAI,CAAC,EAAE,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAA;AACjG,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,kBAAkB,CACtC,WAAmB,EACnB,MAAc,EACd,OAAkC;IAElC,MAAM,UAAU,GAAG,OAAO,CAAC,WAAW,EAAE,MAAM,CAAC,CAAA;IAC/C,MAAM,KAAK,GAAG,MAAM,aAAa,CAAC,UAAU,CAAC,CAAA;IAC7C,MAAM,UAAU,GAAG,OAAO,CAAC,WAAW,EAAE,OAAO,CAAC,iBAAiB,EAAE,qBAAqB,CAAC,CAAA;IACzF,MAAM,KAAK,CAAC,UAAU,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAA;IAC5C,MAAM,aAAa,GAAG,MAAM,OAAO,CAAC,IAAI,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC,CAAA;IAC7D,IAAI,eAAe,GAAG,KAAK,CAAA;IAC3B,IAAI,CAAC;QACH,MAAM,YAAY,CAAC,KAAK,CAAC,QAAQ,EAAE,aAAa,CAAC,CAAA;QACjD,MAAM,aAAa,GAAG,CAAC,MAAM,OAAO,CAAC,aAAa,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,CAAC,CAAC;aAC1E,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,MAAM,EAAE,IAAI,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,WAAW,EAAE,KAAK,MAAM,CAAC;aACjF,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,CAAC,aAAa,EAAE,KAAK,CAAC,IAAI,CAAC,CAAC;aAClD,IAAI,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,KAAK,CAAC,CAAC,CAAA;QACnD,MAAM,gBAAgB,CAAC,aAAa,CAAC,CAAA;QACrC,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,EAAE,sBAAsB,CAAC,CAAA;QAC9D,MAAM,iBAAiB,GAAG,OAAO,CAAC,iBAAiB,IAAI,wBAAwB,CAAC,WAAW,CAAC,CAAA;QAC5F,MAAM,KAAK,GAAG,KAAK,CACjB,OAAO,CAAC,QAAQ,EAChB,cAAc,CAAC,iBAAiB,EAAE,aAAa,CAAC,EAChD;YACE,GAAG,EAAE,WAAW;YAChB,QAAQ,EAAE,OAAO,CAAC,QAAQ,KAAK,OAAO;YACtC,GAAG,EAAE;gBACH,GAAG,2BAA2B,EAAE;gBAChC,2BAA2B,EAAE,UAAU;aACxC;YACD,KAAK,EAAE,KAAK;YACZ,WAAW,EAAE,IAAI;SAClB,CACF,CAAA;QACD,MAAM,SAAS,GAAG,MAAM,kBAAkB,CAAC,KAAK,EAAE,OAAO,CAAC,MAAM,CAAC,CAAA;QACjE,IAAI,OAAO,CAAC,MAAM,EAAE,OAAO,EAAE,CAAC;YAC5B,MAAM,IAAI,KAAK,CAAC,iDAAiD,CAAC,CAAA;QACpE,CAAC;QACD,IAAI,SAAS,CAAC,UAAU,IAAI,SAAS,CAAC,QAAQ,KAAK,CAAC,EAAE,CAAC;YACrD,MAAM,YAAY,CAAC,SAAS,CAAC,UAAU,IAAI,SAAS,CAAC,UAAU,CAAC,CAAA;QAClE,CAAC;QACD,IAAI,MAAgD,CAAA;QACpD,IAAI,CAAC;YACH,MAAM,GAAG,qBAAqB,CAAC,MAAM,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,CAAA;QACjF,CAAC;QAAC,MAAM,CAAC;YACP,MAAM,IAAI,KAAK,CAAC,+DAA+D,CAAC,CAAA;QAClF,CAAC;QACD,MAAM,qBAAqB,CAAC,aAAa,CAAC,CAAA;QAC1C,MAAM,iBAAiB,GAAG,MAAM,oBAAoB,CAAC,WAAW,EAAE,aAAa,CAAC,CAAA;QAChF,eAAe,GAAG,IAAI,CAAA;QACtB,OAAO;YACL,YAAY,EAAE,KAAK,CAAC,QAAQ,CAAC,MAAM;YACnC,iBAAiB;YACjB,YAAY,EAAE,MAAM,CAAC,YAAY;YACjC,UAAU,EAAE,KAAK,CAAC,IAAI;YACtB,KAAK,EAAE,MAAM,CAAC,KAAK;SACpB,CAAA;IACH,CAAC;YAAS,CAAC;QACT,IAAI,CAAC,eAAe,EAAE,CAAC;YACrB,MAAM,mBAAmB,CAAC,UAAU,EAAE,aAAa,CAAC,CAAA;QACtD,CAAC;IACH,CAAC;AACH,CAAC"}

package/dist/analysis/cli.d.ts

@@ -0,0 +1,7 @@
+import type { AnalyzeOptions } from "../commands/options.js";
+export interface AnalyzeInvocation {
+    command: "analyze";
+    options: AnalyzeOptions;
+    target: string;
+}
+export declare function parseAnalyzeArguments(args: string[]): AnalyzeInvocation;

package/dist/analysis/cli.js

@@ -0,0 +1,28 @@
+import { parseArgs } from "node:util";
+export function parseAnalyzeArguments(args) {
+    const parsed = parseArgs({
+        args,
+        allowPositionals: true,
+        strict: true,
+        options: {
+            artifacts: { type: "string", default: ".flakelab/runs" },
+            baseline: { type: "string" },
+            json: { type: "boolean", default: false },
+            verbose: { type: "boolean", default: false },
+        },
+    });
+    if (parsed.positionals.length !== 1) {
+        throw new Error("analyze requires exactly one blob report path");
+    }
+    return {
+        command: "analyze",
+        options: {
+            artifacts: parsed.values.artifacts ?? ".flakelab/runs",
+            baseline: parsed.values.baseline,
+            json: parsed.values.json ?? false,
+            verbose: parsed.values.verbose ?? false,
+        },
+        target: parsed.positionals[0],
+    };
+}
+//# sourceMappingURL=cli.js.map

package/dist/analysis/cli.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.js","sourceRoot":"","sources":["../../src/analysis/cli.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,MAAM,WAAW,CAAA;AAUrC,MAAM,UAAU,qBAAqB,CAAC,IAAc;IAClD,MAAM,MAAM,GAAG,SAAS,CAAC;QACvB,IAAI;QACJ,gBAAgB,EAAE,IAAI;QACtB,MAAM,EAAE,IAAI;QACZ,OAAO,EAAE;YACP,SAAS,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,gBAAgB,EAAE;YACxD,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;YAC5B,IAAI,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,OAAO,EAAE,KAAK,EAAE;YACzC,OAAO,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,OAAO,EAAE,KAAK,EAAE;SAC7C;KACF,CAAC,CAAA;IACF,IAAI,MAAM,CAAC,WAAW,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACpC,MAAM,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAA;IAClE,CAAC;IACD,OAAO;QACL,OAAO,EAAE,SAAS;QAClB,OAAO,EAAE;YACP,SAAS,EAAE,MAAM,CAAC,MAAM,CAAC,SAAS,IAAI,gBAAgB;YACtD,QAAQ,EAAE,MAAM,CAAC,MAAM,CAAC,QAAQ;YAChC,IAAI,EAAE,MAAM,CAAC,MAAM,CAAC,IAAI,IAAI,KAAK;YACjC,OAAO,EAAE,MAAM,CAAC,MAAM,CAAC,OAAO,IAAI,KAAK;SACxC;QACD,MAAM,EAAE,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC;KAC9B,CAAA;AACH,CAAC"}