@bli-cockpit/cli 0.2.237 → 0.2.239

package/dist/upload-sync.js

@@ -22,157 +22,181 @@ export async function syncLocalAmbientEnvelope(options = {}) {
     if (!provenance) {
         throw new Error("Ambient upload requires collector provenance.");
     }
-    // Assigned progressively so the spool path below can still report what did
-    // land if the upload pass throws part-way through.
-    let uploadOutcomes = [];
-    let uploadedChunkCount = 0;
-    const attemptedDate = options.now ?? new Date(attemptedAt);
-    const staging = await readRawEvidenceStagingState(paths.state_dir);
+    const run = {
+        options, paths, cursor, built, fetchImpl, provenance, attemptedAt,
+        attemptedDate: options.now ?? new Date(attemptedAt),
+        staging: await readRawEvidenceStagingState(paths.state_dir),
+        uploadOutcomes: [], uploadedChunkCount: 0,
+    };
     try {
-        if (built.raw_evidence_upload_files.length > 0) {
-            // Objects still inside their backoff window are not offered at all. They
-            // become held outcomes below, which keeps them out of the envelope, keeps
-            // the sync in retry_pending, and names the reason — the 559th identical
-            // attempt is what BLI-3066 made impossible. An operator-invoked retry
-            // (`cockpit backfill`, doctor) offers them anyway and says so: the window
-            // is the scheduler's cadence, not an answer to a person asking now
-            // (BLI-3118).
-            const { deliverable, held, bypassed } = partitionHeldEvidenceFiles(built.raw_evidence_upload_files, staging, attemptedDate, options.evidenceDeliveryMode);
-            logEvidenceHeldByBackoff(held, attemptedAt);
-            logEvidenceBackoffBypassed(bypassed, attemptedAt);
-            uploadOutcomes = held;
-            if (deliverable.length > 0) {
-                const upload = await uploadRawEvidenceFilesChunked({
-                    fetchImpl,
-                    dashboardUrl: built.dashboard_url,
-                    deviceToken: built.device_token,
-                    provenance,
-                    generatedAt: built.envelope.generated_at,
-                    files: deliverable,
-                });
-                uploadOutcomes = [...held, ...upload.outcomes];
-                uploadedChunkCount = upload.uploaded_chunk_count;
-            }
-            await persistDeliveryAttempts(paths.state_dir, staging, uploadOutcomes, attemptedDate);
-        }
-        const response = await postEnvelopeToIngest({
-            fetchImpl,
-            built,
-            envelope: applyRawEvidenceUploadOutcomes(built.envelope, uploadOutcomes),
-        });
-        advanceRawEvidenceCursor(cursor, uploadOutcomes, attemptedAt, built.raw_evidence_upload_files);
-        await writeRawEvidenceCursor(paths, cursor);
-        await reportAgentImageArtifacts({
-            fetchImpl,
-            dashboardUrl: built.dashboard_url,
-            deviceToken: built.device_token,
-            provenance,
-            generatedAt: built.envelope.generated_at,
-            ticketId: built.ticket_id,
-            repoFingerprint: built.envelope.work_context.repo_fingerprint,
-            worktreeFingerprint: built.envelope.work_context.worktree_fingerprint,
-            outcomes: uploadOutcomes,
-            reused: built.raw_evidence_facts?.reused ?? [],
-            cursorObjects: cursor.objects,
-        }).catch((error) => {
-            // The whole point of `agent_artifact_report_local_error` is that it is
-            // NOT the network error the reporter already labels itself — it is a
-            // throw from our own code on the way in. That distinction is only
-            // useful if the throw says what it was (BLI-3238).
-            console.error("[cockpit-sync] agent image artifact report threw locally", JSON.stringify({
-                reason: "agent_artifact_report_local_error",
-                ...describeError(error),
-            }));
-            return {
-                posted: false,
-                reason: "agent_artifact_report_local_error",
-                recorded_count: 0,
-            };
-        });
-        await recordIngestedSyncOutcome({
-            paths,
-            options,
-            built,
-            uploadOutcomes,
-            attemptedAt,
-        });
-        // The success branch logs too. A line that only fires on failure cannot
-        // answer "did anything land at all today?", which is the question that
-        // would have caught BLI-2528 in a week instead of 57 days.
-        console.error("[cockpit-sync] ingest accepted the envelope", JSON.stringify({
-            reason: "ingest_accepted",
-            http_status: response.status,
-            event_count: built.event_count,
-            raw_evidence_file_count: built.raw_evidence_upload_files.length,
-            uploaded_chunk_count: uploadedChunkCount,
-            failed_object_count: uploadOutcomes.filter((outcome) => outcome.upload_state === "upload_failed").length,
-        }));
-        return {
-            status: "uploaded",
-            dashboard_url: built.dashboard_url,
-            ticket_id: built.ticket_id,
-            work_context_id: built.envelope.work_context.work_context_id,
-            head_sha: built.head_sha,
-            event_count: built.event_count,
-            source_scan_count: built.source_scan_count,
-            risk_flag_count: built.risk_flag_count,
-            http_status: response.status,
-            ...summarizeRawEvidenceDelivery(built, uploadOutcomes, uploadedChunkCount, cursor, staging, attemptedDate),
-        };
+        await deliverSyncEvidence(run);
+        return await ingestDeliveredSync(run);
     }
     catch (error) {
-        // The spool row used to store `error.message` verbatim, so every distinct
-        // server sentence became its own "reason" and nothing could count how many
-        // machines were failing the same way (BLI-3483). It now stores a label from
-        // a closed set plus a bounded, redacted cause.
-        const classified = classifySyncFailure(error);
-        // A committed content-addressed object may be shared by concurrent syncs.
-        // Never delete it from this error path: an ingest retry can reuse it, while
-        // client-side cleanup cannot prove exclusive ownership without racing a
-        // second sync that is about to index the same object.
-        const spooledFailureReason = classified.failure_reason;
-        console.error("[cockpit-sync] sync spooled for retry", JSON.stringify({
-            reason: classified.reason,
-            ...(classified.http_status === null
-                ? {}
-                : { http_status: classified.http_status }),
-            ...(classified.detail ? { detail: classified.detail } : {}),
-            event_count: built.event_count,
-            raw_evidence_file_count: built.raw_evidence_upload_files.length,
-            uploaded_chunk_count: uploadedChunkCount,
+        return spoolFailedSync(run, error);
+    }
+}
+async function deliverSyncEvidence(run) {
+    const { options, paths, built, fetchImpl, provenance, attemptedAt, attemptedDate, staging } = run;
+    if (built.raw_evidence_upload_files.length > 0) {
+        // Objects still inside their backoff window are not offered at all. They
+        // become held outcomes below, which keeps them out of the envelope, keeps
+        // the sync in retry_pending, and names the reason — the 559th identical
+        // attempt is what BLI-3066 made impossible. An operator-invoked retry
+        // (`cockpit backfill`, doctor) offers them anyway and says so: the window
+        // is the scheduler's cadence, not an answer to a person asking now
+        // (BLI-3118).
+        const { deliverable, held, bypassed } = partitionHeldEvidenceFiles(built.raw_evidence_upload_files, staging, attemptedDate, options.evidenceDeliveryMode);
+        logEvidenceHeldByBackoff(held, attemptedAt);
+        logEvidenceBackoffBypassed(bypassed, attemptedAt);
+        run.uploadOutcomes = held;
+        if (deliverable.length > 0) {
+            const upload = await uploadRawEvidenceFilesChunked({
+                fetchImpl,
+                dashboardUrl: built.dashboard_url,
+                deviceToken: built.device_token,
+                provenance,
+                generatedAt: built.envelope.generated_at,
+                files: deliverable,
+            });
+            run.uploadOutcomes = [...held, ...upload.outcomes];
+            run.uploadedChunkCount = upload.uploaded_chunk_count;
+        }
+        await persistDeliveryAttempts(paths.state_dir, staging, run.uploadOutcomes, attemptedDate);
+    }
+}
+async function ingestDeliveredSync(run) {
+    const { paths, options, built, fetchImpl, cursor, attemptedAt, attemptedDate, staging, uploadOutcomes, uploadedChunkCount } = run;
+    const response = await postEnvelopeToIngest({
+        fetchImpl,
+        built,
+        envelope: applyRawEvidenceUploadOutcomes(built.envelope, uploadOutcomes),
+    });
+    advanceRawEvidenceCursor(cursor, uploadOutcomes, attemptedAt, built.raw_evidence_upload_files);
+    await writeRawEvidenceCursor(paths, cursor);
+    await reportSyncArtifacts(run);
+    await recordIngestedSyncOutcome({
+        paths,
+        options,
+        built,
+        uploadOutcomes,
+        attemptedAt,
+    });
+    logAcceptedSync(run, response);
+    return {
+        status: "uploaded",
+        dashboard_url: built.dashboard_url,
+        ticket_id: built.ticket_id,
+        work_context_id: built.envelope.work_context.work_context_id,
+        head_sha: built.head_sha,
+        event_count: built.event_count,
+        source_scan_count: built.source_scan_count,
+        risk_flag_count: built.risk_flag_count,
+        http_status: response.status,
+        ...summarizeRawEvidenceDelivery(built, uploadOutcomes, uploadedChunkCount, cursor, staging, attemptedDate),
+    };
+}
+async function reportSyncArtifacts(run) {
+    const { built, fetchImpl, provenance, cursor, uploadOutcomes } = run;
+    await reportAgentImageArtifacts({
+        fetchImpl,
+        dashboardUrl: built.dashboard_url,
+        deviceToken: built.device_token,
+        provenance,
+        generatedAt: built.envelope.generated_at,
+        ticketId: built.ticket_id,
+        repoFingerprint: built.envelope.work_context.repo_fingerprint,
+        worktreeFingerprint: built.envelope.work_context.worktree_fingerprint,
+        outcomes: uploadOutcomes,
+        reused: built.raw_evidence_facts?.reused ?? [],
+        cursorObjects: cursor.objects,
+    }).catch((error) => {
+        // The whole point of `agent_artifact_report_local_error` is that it is
+        // NOT the network error the reporter already labels itself — it is a
+        // throw from our own code on the way in. That distinction is only
+        // useful if the throw says what it was (BLI-3238).
+        console.error("[cockpit-sync] agent image artifact report threw locally", JSON.stringify({
+            reason: "agent_artifact_report_local_error",
+            ...describeError(error),
         }));
-        const entry = await recordUploadFailure(paths, {
-            last_attempt_at: attemptedAt,
-            dashboard_url: built.dashboard_url,
-            work_context_id: built.envelope.work_context.work_context_id,
-            ticket_id: built.ticket_id,
-            repo_label: built.repo_label,
-            branch: built.envelope.work_context.branch,
-            event_count: built.event_count,
-            source_scan_count: built.source_scan_count,
-            risk_flag_count: built.risk_flag_count,
-            raw_evidence_file_count: built.raw_evidence_upload_files.length,
-            retry_sources: retrySourcesForFailedSync(options, built.raw_evidence_facts),
-            failure_reason: spooledFailureReason,
-            retry_command: "cockpit sync",
-        });
         return {
-            status: "spooled",
-            dashboard_url: built.dashboard_url,
-            ticket_id: built.ticket_id,
-            work_context_id: built.envelope.work_context.work_context_id,
-            head_sha: built.head_sha,
-            event_count: built.event_count,
-            source_scan_count: built.source_scan_count,
-            risk_flag_count: built.risk_flag_count,
-            failure_reason: spooledFailureReason,
-            failure_class: classified.reason,
-            failure_http_status: classified.http_status,
-            spool_entry_id: entry.spool_id,
-            retry_command: entry.retry_command,
-            ...summarizeRawEvidenceDelivery(built, uploadOutcomes, uploadedChunkCount, cursor, staging, attemptedDate),
+            posted: false,
+            reason: "agent_artifact_report_local_error",
+            recorded_count: 0,
         };
-    }
+    });
+}
+function logAcceptedSync(run, response) {
+    const { built, uploadedChunkCount, uploadOutcomes } = run;
+    // The success branch logs too. A line that only fires on failure cannot
+    // answer "did anything land at all today?", which is the question that
+    // would have caught BLI-2528 in a week instead of 57 days.
+    console.error("[cockpit-sync] ingest accepted the envelope", JSON.stringify({
+        reason: "ingest_accepted",
+        http_status: response.status,
+        event_count: built.event_count,
+        raw_evidence_file_count: built.raw_evidence_upload_files.length,
+        uploaded_chunk_count: uploadedChunkCount,
+        failed_object_count: uploadOutcomes.filter((outcome) => outcome.upload_state === "upload_failed").length,
+    }));
+}
+async function spoolFailedSync(run, error) {
+    const { paths, options, built, attemptedAt, attemptedDate, cursor, staging, uploadOutcomes, uploadedChunkCount } = run;
+    // The spool row used to store `error.message` verbatim, so every distinct
+    // server sentence became its own "reason" and nothing could count how many
+    // machines were failing the same way (BLI-3483). It now stores a label from
+    // a closed set plus a bounded, redacted cause.
+    const classified = classifySyncFailure(error);
+    // A committed content-addressed object may be shared by concurrent syncs.
+    // Never delete it from this error path: an ingest retry can reuse it, while
+    // client-side cleanup cannot prove exclusive ownership without racing a
+    // second sync that is about to index the same object.
+    const spooledFailureReason = classified.failure_reason;
+    logSpooledSyncFailure(run, classified);
+    const entry = await recordUploadFailure(paths, {
+        last_attempt_at: attemptedAt,
+        dashboard_url: built.dashboard_url,
+        work_context_id: built.envelope.work_context.work_context_id,
+        ticket_id: built.ticket_id,
+        repo_label: built.repo_label,
+        branch: built.envelope.work_context.branch,
+        event_count: built.event_count,
+        source_scan_count: built.source_scan_count,
+        risk_flag_count: built.risk_flag_count,
+        raw_evidence_file_count: built.raw_evidence_upload_files.length,
+        retry_sources: retrySourcesForFailedSync(options, built.raw_evidence_facts),
+        failure_reason: spooledFailureReason,
+        retry_command: "cockpit sync",
+    });
+    return {
+        status: "spooled",
+        dashboard_url: built.dashboard_url,
+        ticket_id: built.ticket_id,
+        work_context_id: built.envelope.work_context.work_context_id,
+        head_sha: built.head_sha,
+        event_count: built.event_count,
+        source_scan_count: built.source_scan_count,
+        risk_flag_count: built.risk_flag_count,
+        failure_reason: spooledFailureReason,
+        failure_class: classified.reason,
+        failure_http_status: classified.http_status,
+        spool_entry_id: entry.spool_id,
+        retry_command: entry.retry_command,
+        ...summarizeRawEvidenceDelivery(built, uploadOutcomes, uploadedChunkCount, cursor, staging, attemptedDate),
+    };
+}
+function logSpooledSyncFailure(run, classified) {
+    const { built, uploadedChunkCount } = run;
+    console.error("[cockpit-sync] sync spooled for retry", JSON.stringify({
+        reason: classified.reason,
+        ...(classified.http_status === null
+            ? {}
+            : { http_status: classified.http_status }),
+        ...(classified.detail ? { detail: classified.detail } : {}),
+        event_count: built.event_count,
+        raw_evidence_file_count: built.raw_evidence_upload_files.length,
+        uploaded_chunk_count: uploadedChunkCount,
+    }));
 }
 /**
  * Assemble the envelope, and if this machine simply cannot upload, say so in
@@ -225,18 +249,8 @@ function advanceRawEvidenceCursor(cursor, outcomes, attemptedAt, files = []) {
     cursor.updated_at = attemptedAt;
 }
 /**
- * What the operator is owed after ingest accepted the envelope, as a decision
- * table on the evidence this sync left behind:
- *
- * | evidence state                        | ledger                              |
- * |---------------------------------------|-------------------------------------|
- * | a gap a later sync could rescue        | success, then a queued retry naming why |
- * | a gap no retry can rescue             | success, then a blocker naming why, plus a stderr record |
- * | nothing missing                       | success, pending state cleared      |
- *
- * The middle row is the one that matters. A permanently rejected object gets no
- * retry — nothing would change — but it must still be named, or `cockpit status`
- * reads clean on a machine with missing collection (BLI-2528).
+ * Record success before queuing a rescuable gap or blocking a permanent one.
+ * A permanently rejected object gets no retry but remains visible in status.
  */
 async function recordIngestedSyncOutcome(options) {
     const { paths, built, uploadOutcomes, attemptedAt } = options;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bli-cockpit/cli",
-  "version": "0.2.237",
+  "version": "0.2.239",
   "private": false,
   "type": "module",
   "bin": {
@@ -27,9 +27,9 @@
     "test": "node dist/cli.js --help && node ../../scripts/assert-public-cli-routing.mjs && node ../../scripts/assert-public-cli-verb-help.mjs && node ../../scripts/assert-public-cli-surface-doors.mjs && node ../../scripts/assert-public-cli-runtime-files.mjs && node ../../scripts/assert-public-cli-exit-contract.mjs && node ../../scripts/assert-public-cli-no-fleet-posts.mjs && node ../../scripts/assert-public-cli-doctor-json.mjs && node ../../scripts/assert-public-package-pack.mjs --workspace=@bli-cockpit/cli"
   },
   "dependencies": {
-    "@bli-cockpit/memory-mcp": "0.1.69",
-    "@bli-cockpit/mcp": "0.1.142",
-    "@bli-cockpit/telemetry-core": "0.1.76"
+    "@bli-cockpit/memory-mcp": "0.1.70",
+    "@bli-cockpit/mcp": "0.1.144",
+    "@bli-cockpit/telemetry-core": "0.1.77"
   },
   "repository": {
     "type": "git",