pi-lens 4.3.0 → 4.4.0

package/dist/clients/console-guard-install.js

@@ -1,26 +0,0 @@
-// Capture this before the guard's other module initialization so host boot is
-// not charged to extension evaluation (#1374). Lives in its own
-// side-effect-free module (#1434 S3d) — see eval-timestamp.ts's docstring for
-// why: `startup-timing.ts` needs this constant but must never trigger a
-// console-guard install merely by importing it.
-import { PI_LENS_EVAL_STARTED_MS } from "./eval-timestamp.js";
-// #1333: install the console guard as an import side-effect so it runs before
-// any other module's initialization code. Installing inside the extension
-// factory is too late — by the time the factory runs, every static/transitive
-// import has already evaluated, so a dependency writing to console during
-// module init would still hit the TUI raw (#1338 review finding). This module
-// MUST remain the first import of index.ts; the enforcement test pins that.
-// installConsoleGuard() itself is idempotent and no-ops under test mode and
-// PI_LENS_CONSOLE_GUARD=0.
-// #1434: the guard captures only while pi-lens owns execution, so the module
-// window must be open for the rest of the import graph. index.ts closes it on
-// its last line; an unref'd backstop closes it if that never runs.
-import { installConsoleGuard, openModuleLoadConsoleWindow, } from "./extension-log.js";
-installConsoleGuard();
-openModuleLoadConsoleWindow();
-// Not re-exported: nothing should import the timestamp from this module.
-// `clients/startup-timing.ts` imports it straight from `eval-timestamp.ts`,
-// which is what keeps that import side-effect-free (S3d). The `void` keeps
-// the import from looking unused to lint while documenting that this module
-// only needs the ordering (evaluate eval-timestamp.ts first), not the value.
-void PI_LENS_EVAL_STARTED_MS;

package/dist/clients/cooperative-budget.js

@@ -1,42 +0,0 @@
-/** Monotonic cooperative-work budgeting for event-loop-sensitive bulk work. */
-export function createDeadline(budgetMs) {
-    const boundedBudget = Math.max(0, budgetMs);
-    let startedAt = performance.now();
-    return {
-        expired: () => performance.now() - startedAt >= boundedBudget,
-        reset: () => {
-            startedAt = performance.now();
-        },
-    };
-}
-export async function yieldIfOverBudget(deadline) {
-    if (!deadline.expired())
-        return false;
-    await new Promise((resolve) => setImmediate(resolve));
-    deadline.reset();
-    return true;
-}
-export async function forEachCooperatively(items, fn, options) {
-    const deadline = createDeadline(options.budgetMs);
-    const assertContinuing = () => {
-        if (options.shouldContinue?.() === false) {
-            throw new Error(options.abortMessage ?? "cooperative work superseded");
-        }
-    };
-    let index = 0;
-    for (const item of items) {
-        // Check before each unit: this per-unit check is what bounds abort
-        // latency (one work unit, not an iteration checkpoint). The post-yield
-        // check below only matters after the final unit's yield, so superseded
-        // work is reported aborted rather than completed.
-        assertContinuing();
-        const result = fn(item, index++);
-        if (result !== undefined &&
-            typeof result.then === "function") {
-            await result;
-        }
-        if (deadline.expired() && (await yieldIfOverBudget(deadline))) {
-            assertContinuing();
-        }
-    }
-}

package/dist/clients/dead-code-client.js

@@ -1,510 +0,0 @@
-/**
- * Cross-file dead-code detection for non-JS/TS ecosystems (#127).
- *
- * Knip (clients/knip-client.ts) gives JS/TS projects project-wide unused
- * exports/files/deps at session_start. Per-file dispatch linters can't do that
- * for other languages — "this exported function is unused anywhere" needs a
- * whole-project scan. This module is the per-language harness that closes the
- * gap, paralleling KnipClient's lifecycle (detect → ensureAvailable → analyze,
- * cached at session_start, surfaced as a turn_end advisory).
- *
- * Phase 1 ships Python via `vulture`. Future phases add Go/Rust/etc. by
- * implementing DeadCodeClient and adding to getDeadCodeClients().
- */
-import { createSubsystemLogger } from "./extension-log.js";
-import { incrementDegradationCount } from "./degradation-ledger.js";
-import * as fs from "node:fs";
-import * as path from "node:path";
-import { findLocalBinsAt, VENV_BIN_DIRS } from "./package-manager.js";
-import { findNearestMarkerRoot } from "./path-utils.js";
-import { getScratchTreeFnmatchPatterns } from "./scratch-tree-policy.js";
-import { safeSpawnAsync } from "./safe-spawn.js";
-import { probeToolAsync } from "./tool-probe.js";
-import { classifyProbeFailure, createAvailabilityLatch, logAvailabilityDecision, startHostStallSampler, } from "./dispatch/runners/utils/availability-policy.js";
-import { firstOutputLine, spawnFailedWithNoOutput, } from "./dispatch/runners/utils/spawn-outcome.js";
-import { formatToolFailure } from "./dispatch/runners/utils/tool-failure.js";
-function emptyResult(language) {
-    return {
-        success: false,
-        language,
-        unusedExports: [],
-        unusedFiles: [],
-        unusedDeps: [],
-        unlistedDeps: [],
-    };
-}
-const ANALYSIS_TIMEOUT_MS = 30_000;
-// Directories never worth scanning for the user's own dead code.
-//
-// Single-sourced from `scratch-tree-policy.ts`'s `EXCLUDED_DIRS`-derived list
-// (#1562 sweep finding: this array had hand-drifted from `EXCLUDED_DIRS` —
-// missing `.pi`/`.claude`/`.next`/`.turbo`/etc, the same single-source-of-
-// truth defect #883 named for per-language lists). `site-packages`/`.eggs`
-// are Python-packaging-specific and have no `EXCLUDED_DIRS` counterpart (no
-// other scanner needs them), so they stay as a small local addition rather
-// than polluting the shared cross-tool list.
-const VULTURE_PYTHON_ONLY_EXCLUDES = ["*/site-packages/*", "*/.eggs/*"];
-const VULTURE_EXCLUDES = [
-    ...getScratchTreeFnmatchPatterns(),
-    ...VULTURE_PYTHON_ONLY_EXCLUDES,
-];
-// Decorators whose target is called by a framework, never by name in the tree.
-// vulture cannot see those call sites, so every such symbol is a permanent
-// false positive that reappears in every scan — the fastest way to teach a
-// reader to stop reading the advisory. Glob-matched by vulture itself.
-const VULTURE_IGNORE_DECORATORS = [
-    "@pytest.fixture",
-    "@pytest.fixture(*",
-    "@fixture",
-    "@fixture(*",
-    "@app.*",
-    "@router.*",
-    "@celery.task*",
-    "@task",
-    "@shared_task*",
-];
-// vulture line: `path/to/file.py:12: unused function 'foo' (60% confidence)`
-const VULTURE_LINE = /^(.*?):(\d+): unused (\w[\w ]*?) '([^']+)' \((\d+)% confidence\)\s*$/;
-/**
- * Parse vulture's text output into normalized issues. Pure (no spawn/fs) so the
- * parser is unit-testable against captured output. Unrecognized lines (banner,
- * `unreachable code` without a quoted name) are ignored. `root` makes file
- * paths project-relative when possible.
- */
-export function parseVultureOutput(output, root) {
-    const issues = [];
-    for (const raw of output.split(/\r?\n/)) {
-        const m = raw.match(VULTURE_LINE);
-        if (!m)
-            continue;
-        const [, file, line, kind, name, confidence] = m;
-        // All map to the "export" bucket (a defined symbol used nowhere); the
-        // `kind` preserves function/class/import/etc. for display.
-        let rel = file;
-        try {
-            rel = path.relative(root, file) || file;
-        }
-        catch {
-            rel = file;
-        }
-        issues.push({
-            category: "export",
-            kind: kind.trim(),
-            name,
-            file: rel,
-            line: Number.parseInt(line, 10),
-            confidence: Number.parseInt(confidence, 10),
-        });
-    }
-    return issues;
-}
-/**
- * Python dead-code via vulture (https://github.com/jendrikseipp/vulture).
- *
- * vulture finds unused functions, classes, methods, imports, variables and
- * attributes by static analysis of the whole tree. It has no JSON reporter, so
- * we parse its stable one-line-per-finding text output. A clean run exits 0;
- * found dead code exits 3 with findings on stdout. A non-zero exit with
- * parseable output is success, not failure (see runAnalyze for the table).
- */
-export class PythonDeadCodeClient {
-    id = "python";
-    language = "Python";
-    /**
-     * Transient-aware memo (#1467): only a durable "vulture is not installed"
-     * verdict is remembered for the session. A probe that timed out expires and
-     * is re-probed, so vulture cannot be disabled for the process by one stall.
-     */
-    availabilityLatch = createAvailabilityLatch();
-    resolved = null;
-    ensureInFlight = null;
-    inFlight = new Map();
-    log;
-    constructor(verbose = false) {
-        this.log = verbose ? createSubsystemLogger("dead-code:python") : () => { };
-    }
-    get minConfidence() {
-        const raw = Number.parseInt(process.env.PI_LENS_VULTURE_MIN_CONFIDENCE ?? "60", 10);
-        return Number.isFinite(raw) ? Math.min(100, Math.max(0, raw)) : 60;
-    }
-    detect(cwd) {
-        return this.resolveProjectRoot(cwd) !== null;
-    }
-    owns(filePath) {
-        const ext = path.extname(filePath).toLowerCase();
-        return ext === ".py" || ext === ".pyi";
-    }
-    /**
-     * Nearest dir with a Python project marker, never at/above $HOME and never
-     * escaping a VCS boundary — same containment rules as KnipClient so a scan
-     * launched from a bare cwd can't recurse the whole home tree (#250/#296).
-     * Delegates to the shared path-utils helper (refs #625) rather than
-     * hand-rolling the climb; only the marker list differs from KnipClient's.
-     */
-    resolveProjectRoot(startDir, homeDirOverride) {
-        return findNearestMarkerRoot(startDir, [
-            "pyproject.toml",
-            "setup.py",
-            "setup.cfg",
-            "requirements.txt",
-            "Pipfile",
-            "tox.ini",
-        ], { boundaries: [".git", ".hg", ".svn"], homeDir: homeDirOverride });
-    }
-    async ensureAvailable(root) {
-        const memo = this.availabilityLatch.read();
-        if (memo !== null)
-            return memo;
-        if (this.ensureInFlight)
-            return this.ensureInFlight;
-        this.ensureInFlight = this.doEnsureAvailable(root);
-        try {
-            return await this.ensureInFlight;
-        }
-        finally {
-            this.ensureInFlight = null;
-        }
-    }
-    /**
-     * `<root>/.venv/bin/vulture` (and `venv`/Windows `Scripts` equivalents),
-     * checked by existence only — no spawn. A project's own venv is the vulture
-     * that project's own config and dependency set were written against; the
-     * bare `vulture`/`python -m vulture` PATH probes below only ever see
-     * whatever happens to be active in the CALLING shell, not the project's
-     * (#1731, discipline B — the same venv-first shape sqlfluff's binary
-     * resolution already uses, `runner-helpers.ts` `createVenvFinder`).
-     *
-     * The directory list is the shared `VENV_BIN_DIRS` — this was a third
-     * private spelling of the same four venv directories, with a
-     * `process.platform` branch that meant the ubuntu lane never saw the
-     * `Scripts` half at all (defect shape 30).
-     *
-     * The search is a fixed lookup at `root` ONLY (`findLocalBinsAt`, no
-     * ancestor walk) — #2544 review round 3 folded this onto the shared
-     * ancestor walker, reasoning a monorepo package should find its
-     * repo-root venv, but that was an unrequested behaviour change with no
-     * #2514 defect behind it. Reverted in round 4 F2. `findLocalBinsAt`
-     * returns EVERY match at `root` rather than the first, because
-     * `doEnsureAvailable` probes candidates in order and must keep the
-     * `venv/` fallback when a `.venv/` vulture exists but cannot run.
-     */
-    venvCandidates(root) {
-        return findLocalBinsAt(["vulture"], root, {
-            windowsExt: ".exe",
-            binDirs: VENV_BIN_DIRS,
-        }).map((full) => ({ cmd: full, prefix: [] }));
-    }
-    async doEnsureAvailable(root) {
-        // Presence-gated, NOT auto-installed. vulture is a pure-Python package
-        // with no standalone binary, so "auto-install" would mean `pip install`
-        // into whatever Python environment happens to be active — wrong and
-        // intrusive for uv / poetry / conda / pipx users. So we only use vulture
-        // when the user already has it, probing the project's own venv first
-        // (#1731), then the `vulture` console script and `python -m vulture` on
-        // PATH (the script dir is frequently not on PATH even when the package
-        // is installed). Mirrors govulncheck's no-install gating.
-        const candidates = [
-            ...(root ? this.venvCandidates(root) : []),
-            { cmd: "vulture", prefix: [] },
-            { cmd: "python", prefix: ["-m", "vulture"] },
-            { cmd: "python3", prefix: ["-m", "vulture"] },
-        ];
-        // A timeout on ANY candidate means the machine, not the tool, answered —
-        // the run gets a bounded retry instead of a permanent skip (#1467).
-        let sawTransient = false;
-        let transientCause = "probe-timeout";
-        // Accumulated across ALL candidates, because the failure verdicts below
-        // are about the whole sweep rather than any one probe. Reporting zero
-        // here would erase the evidence that cracked #1467: four 5s probes and
-        // a stalled host look identical to an absent tool without these two
-        // numbers.
-        const sweepStartedAt = Date.now();
-        let sweepHostStallMs = 0;
-        /** What the last classified candidate returned (#1500 review): this was
-         * computed and thrown away, so every vulture row said "missing" with no
-         * trace of which candidate reported what. */
-        let sweepEvidence;
-        for (const c of candidates) {
-            const sampler = startHostStallSampler();
-            const startedAt = Date.now();
-            let probe;
-            let hostStallMs;
-            try {
-                probe = await probeToolAsync(c.cmd, [...c.prefix, "--version"], {
-                    timeout: 5000,
-                });
-            }
-            finally {
-                hostStallMs = sampler.stop();
-                sweepHostStallMs += hostStallMs;
-            }
-            if (!probe.error && probe.status === 0) {
-                this.resolved = c;
-                this.availabilityLatch.noteAvailable();
-                this.log(`vulture found: ${[c.cmd, ...c.prefix].join(" ")}`);
-                logAvailabilityDecision({
-                    tool: "vulture",
-                    verdict: "available",
-                    outcome: "success",
-                    cause: "ok",
-                    elapsedMs: Date.now() - startedAt,
-                    latched: true,
-                    hostStallMs,
-                    budgetMs: 5000,
-                    classifiedBy: "probe",
-                });
-                return true;
-            }
-            const classified = classifyProbeFailure(probe, {
-                hostStallMs,
-                command: c.cmd,
-            });
-            sweepEvidence = classified.evidence;
-            if (classified.outcome === "transient") {
-                sawTransient = true;
-                transientCause = classified.cause;
-            }
-        }
-        if (sawTransient) {
-            const retryAfterMs = this.availabilityLatch.noteUnavailable("transient", transientCause);
-            this.log("vulture probe timed out; will retry (not treated as missing)");
-            logAvailabilityDecision({
-                tool: "vulture",
-                verdict: "unavailable",
-                outcome: "transient",
-                cause: transientCause,
-                elapsedMs: Date.now() - sweepStartedAt,
-                hostStallMs: sweepHostStallMs,
-                latched: false,
-                retryAfterMs,
-                budgetMs: 5000,
-                classifiedBy: "probe",
-                ...(sweepEvidence !== undefined && { evidence: sweepEvidence }),
-            });
-            return false;
-        }
-        this.availabilityLatch.noteUnavailable("missing", "not-found");
-        this.log("vulture not installed; skipping (no auto-install)");
-        logAvailabilityDecision({
-            tool: "vulture",
-            verdict: "unavailable",
-            outcome: "missing",
-            cause: "not-found",
-            elapsedMs: Date.now() - sweepStartedAt,
-            hostStallMs: sweepHostStallMs,
-            latched: true,
-            budgetMs: 5000,
-            // Every candidate probe was classified, and none was transient (#1500).
-            classifiedBy: "probe",
-            ...(sweepEvidence !== undefined && { evidence: sweepEvidence }),
-        });
-        return false;
-    }
-    async analyze(cwd) {
-        const root = this.resolveProjectRoot(cwd || process.cwd());
-        if (!root) {
-            return {
-                ...emptyResult(this.language),
-                success: true,
-                summary: "No Python project root found; vulture skipped",
-            };
-        }
-        if (!(await this.ensureAvailable(root))) {
-            return {
-                ...emptyResult(this.language),
-                success: true,
-                summary: "vulture not installed; skipped. Install vulture (pip/uv/pipx) to enable Python dead-code detection.",
-            };
-        }
-        const key = path.resolve(root);
-        const existing = this.inFlight.get(key);
-        if (existing)
-            return existing;
-        const promise = this.runAnalyze(key);
-        const wrapped = promise.finally(() => {
-            // Identity-guarded release (#1968, #1967's pattern): delete only if
-            // THIS build is still the registered one. A bare delete-by-key lets
-            // a late-settling build A evict a live build B that replaced the
-            // entry mid-flight, and the next caller starts a duplicate.
-            if (this.inFlight.get(key) === wrapped)
-                this.inFlight.delete(key);
-        });
-        this.inFlight.set(key, wrapped);
-        return wrapped;
-    }
-    /**
-     * True when `<root>/pyproject.toml` carries a `[tool.vulture]` table.
-     * vulture's own config discovery reads that table, but CLI flags override
-     * it — passing `--min-confidence`/`--exclude` unconditionally silently
-     * overrode a project's own thresholds and ignore list (#1731, discipline A).
-     * Same shape as `hasSqlfluffConfig`/`hasMypyConfig` in `tool-policy.ts`: a
-     * plain string search, no TOML parser, matching the section header only.
-     */
-    hasProjectVultureConfig(root) {
-        try {
-            const content = fs.readFileSync(path.join(root, "pyproject.toml"), "utf-8");
-            return content.includes("[tool.vulture]");
-        }
-        catch {
-            return false;
-        }
-    }
-    async runAnalyze(root) {
-        const startMs = Date.now();
-        const invocation = this.resolved ?? { cmd: "vulture", prefix: [] };
-        // Let the project's own [tool.vulture] config win (#1731, discipline A):
-        // vulture discovers it unaided, but these two flags override it when
-        // passed, so they are omitted whenever the project ships that table.
-        const hasConfig = this.hasProjectVultureConfig(root);
-        const args = [
-            ...invocation.prefix,
-            ".",
-            ...(hasConfig ? [] : [`--min-confidence=${this.minConfidence}`]),
-            ...(hasConfig ? [] : [`--exclude=${VULTURE_EXCLUDES.join(",")}`]),
-            `--ignore-decorators=${VULTURE_IGNORE_DECORATORS.join(",")}`,
-        ];
-        const result = await safeSpawnAsync(invocation.cmd, args, {
-            timeout: ANALYSIS_TIMEOUT_MS,
-            cwd: root,
-        });
-        const durationMs = Date.now() - startMs;
-        // Spawn-level failure (ENOENT, timeout) — not a "found issues" exit.
-        if (result.error) {
-            this.log(`scan error: ${result.error.message}`);
-            return {
-                ...emptyResult(this.language),
-                summary: `Error: ${result.error.message}`,
-                durationMs,
-            };
-        }
-        // Verified exit-code table (vulture 2.16, probed live during the #1758
-        // review): a clean run exits 0 with empty stdout; dead code found exits
-        // 3 with findings on stdout; invalid input or a parse error exits 1
-        // with empty stdout and the error on stderr. #1736 sweep: the ORIGINAL
-        // guard here required non-empty stderr to call it an error, so a
-        // nonzero exit with BOTH empty stdout and empty stderr (a silent crash)
-        // still fell through to "No dead code found" -- the same
-        // empty-distinguishes-clean-from-errored gap the knip fix closes. A
-        // nonzero exit with no findings on stdout is never clean now,
-        // regardless of whether stderr said anything.
-        const output = result.stdout || "";
-        if (!output.trim()) {
-            const stderr = (result.stderr || "").trim();
-            // Same discriminator every dispatch/runners linter uses
-            // (`spawnFailedWithNoOutput`) rather than a parallel hand-rolled
-            // check (result.error is already handled above, so this reduces to
-            // the `status !== 0` half).
-            if (spawnFailedWithNoOutput(result, output)) {
-                // #1816: one shared wording, one truncation, signal named.
-                const reason = formatToolFailure({
-                    tool: "vulture",
-                    status: result.status,
-                    signal: result.signal,
-                    stderr: result.stderr,
-                });
-                this.log(reason);
-                incrementDegradationCount({
-                    kind: "runner-empty-result",
-                    subject: "vulture",
-                    reason,
-                });
-                return {
-                    ...emptyResult(this.language),
-                    summary: stderr
-                        ? `vulture error: ${firstOutputLine(stderr)}`
-                        : reason,
-                    durationMs,
-                };
-            }
-            // vulture's verified exit-code table (above): exit 0 with empty
-            // stdout IS the clean-run signal, so this ran over the root
-            // (#2154) — unlike knip, whose clean run always prints JSON.
-            return {
-                ...emptyResult(this.language),
-                success: true,
-                analyzed: true,
-                summary: "No dead code found",
-                durationMs,
-            };
-        }
-        return { ...this.parseOutput(output, root), durationMs };
-    }
-    parseOutput(output, root) {
-        const unusedExports = parseVultureOutput(output, root);
-        const total = unusedExports.length;
-        return {
-            ...emptyResult(this.language),
-            success: true,
-            // #2154: parsed vulture output for this root.
-            analyzed: true,
-            unusedExports,
-            summary: total === 0
-                ? "No dead code found"
-                : `Found ${total} unused Python symbol(s)`,
-        };
-    }
-}
-/**
- * The dead-code clients to run at session_start. Each is offered every project;
- * the orchestrator calls detect() to decide which actually apply (polyglot
- * repos may run several). Phase 1: Python only.
- */
-export function getDeadCodeClients(verbose = false) {
-    return [new PythonDeadCodeClient(verbose)];
-}
-/** Total issue count across all buckets — convenience for advisories/telemetry. */
-export function deadCodeIssueCount(result) {
-    return (result.unusedExports.length +
-        result.unusedFiles.length +
-        result.unusedDeps.length +
-        result.unlistedDeps.length);
-}
-/** Every bucket flattened — delta diffing and diagnostic mapping want one list. */
-export function deadCodeIssues(result) {
-    return [
-        ...result.unusedExports,
-        ...result.unusedFiles,
-        ...result.unusedDeps,
-        ...result.unlistedDeps,
-    ];
-}
-/**
- * Joins finding-identity parts into a stable key for diffing one scan
- * against the previous one. The string-only signature is load-bearing: a
- * line/column number cannot be passed without a compile error. An edit
- * shifts every line below it, and a scan delta is then filtered to exactly
- * the files the edit touched — so a line in the key turns each shifted
- * pre-existing finding into a "newly unused" report under a heading that
- * blames the agent's own edit for orphaning it. Inserting four lines above
- * one real finding produced four false ones (#1477, #1483). Shared by
- * knip's per-turn delta in runtime-turn.ts so the rule lives in one place.
- */
-export function stableFindingKey(...parts) {
-    return parts.map((part) => part ?? "").join(":");
-}
-/** Stable identity for diffing one scan against the previous one. */
-export function deadCodeIssueKey(issue) {
-    return stableFindingKey(issue.category, issue.file, issue.name);
-}
-/**
- * Format this turn's ATTRIBUTABLE delta — symbols that became unused in files
- * the agent just edited — or "" when there is nothing to report. Mirrors the
- * Knip advisory's contract deliberately: the project-wide list is not injected
- * per turn (hundreds of pre-existing findings would drown the blockers and burn
- * context every turn), it stays available on demand via lens_diagnostics.
- */
-export function formatDeadCodeDelta(issues, language, max = 5) {
-    if (issues.length === 0)
-        return "";
-    let report = `💀 Newly unused ${language} symbols in files you edited — check if callers need updating (dead-code):\n`;
-    for (const issue of issues.slice(0, max)) {
-        const loc = issue.file
-            ? `${issue.file}${issue.line ? `:${issue.line}` : ""}`
-            : "(unknown)";
-        report += `  ${loc} — unused ${issue.kind} ${issue.name}\n`;
-    }
-    if (issues.length > max) {
-        report += `  … and ${issues.length - max} more\n`;
-    }
-    return report;
-}

package/dist/clients/dead-code-logger.js

@@ -1,30 +0,0 @@
-/**
- * NDJSON telemetry for cross-file dead-code scans (#127). One event per
- * session_start scan per language, so we can answer: which languages get
- * scanned, how many findings, and how long the whole-project scan takes (the
- * input to phasing decisions in the issue). Mirrors `ast-grep-tool-logger.ts`
- * for shape + size-based rotation.
- */
-import * as path from "node:path";
-import { isTestMode } from "./env-utils.js";
-import { getGlobalPiLensLogDir } from "./probe-home-state.js";
-import { createNdjsonLogger } from "./ndjson-logger.js";
-const LOG_DIR = getGlobalPiLensLogDir();
-const LOG_FILE = path.join(LOG_DIR, "dead-code.log");
-const LOG_BACKUP_FILE = path.join(LOG_DIR, "dead-code.log.1");
-const MAX_LOG_BYTES = Math.max(128 * 1024, Number.parseInt(process.env.PI_LENS_DEAD_CODE_LOG_MAX_BYTES ?? "1048576", 10) || 1048576);
-const writer = createNdjsonLogger({
-    filePath: LOG_FILE,
-    maxBytes: MAX_LOG_BYTES,
-    backupPath: LOG_BACKUP_FILE,
-});
-/**
- * Append one scan event. Fire-and-forget: telemetry must never break a scan, so
- * every fs error is swallowed. Skipped under test mode to keep the suite from
- * writing to the user's real ~/.pi-lens.
- */
-export function logDeadCodeScan(event) {
-    if (isTestMode())
-        return;
-    writer.log({ ts: new Date().toISOString(), ...event });
-}