fix(ci): correct Bencher reporting and benchmark failures

.github/workflows/internal-benchmark.yml

@@ -23,12 +23,20 @@ jobs:
       - name: Setup Deno
         uses: denoland/setup-deno@v2
         with:
-          deno-version: v2.x
+          deno-version: latest
       - uses: bencherdev/bencher@main
       - name: Run benchmark
+        run: deno task benchmark:tinybench
+      - name: Publish benchmark
+        shell: bash
         run: |
+          args=()
+          if [[ "$GITHUB_REF" != refs/heads/master ]]; then
+            args+=(--dry-run)
+          fi
           bencher run \
-          --branch master \
+          "${args[@]}" \
+          --branch "$GITHUB_REF_NAME" \
           --threshold-measure latency \
           --threshold-test t_test \
           --threshold-max-sample-size 64 \
@@ -41,6 +49,4 @@ jobs:
           --threshold-upper-boundary _ \
           --thresholds-reset \
           --file benchmark-report.json \
-          --err \
-          --github-actions ${{ secrets.GITHUB_TOKEN }} \
-          "deno task benchmark:tinybench"
+          --err

benchmarks/benchmarks-utils.mjs

@@ -7,7 +7,7 @@ import {
   rmSync,
   writeFileSync,
 } from 'node:fs'
-import { Bench } from 'tinybench'
+import { Bench, mToNs } from 'tinybench'
 import {
   DynamicThreadPool,
   FixedThreadPool,
@@ -292,7 +292,7 @@ export const runPoolifierBenchmarkTinyBench = async (
   const bmfResults = {}
   let pool
   try {
-    const bench = new Bench()
+    const bench = new Bench({ throws: true })
     pool = buildPoolifierPool(workerType, poolType, poolSize)
 
     for (const workerChoiceStrategy of Object.values(WorkerChoiceStrategies)) {
@@ -364,30 +364,24 @@ export const runPoolifierBenchmarkTinyBench = async (
     console.table(bench.table())
 
     for (const task of tasks) {
-      if (
-        task.result?.state === 'completed' ||
-        task.result?.state === 'aborted-with-statistics'
-      ) {
-        bmfResults[task.name] = {
-          latency: {
-            value: task.result.latency.mean,
-            lower_value: task.result.latency.mean - task.result.latency.sd,
-            upper_value: task.result.latency.mean + task.result.latency.sd,
-          },
-          throughput: {
-            value: task.result.throughput.mean,
-            lower_value: task.result.throughput.mean -
-              task.result.throughput.sd,
-            upper_value: task.result.throughput.mean +
-              task.result.throughput.sd,
-          },
-        }
+      if (task.result?.state !== 'completed') {
+        throw new Error(`Benchmark did not complete: ${task.name}`)
+      }
+      bmfResults[task.name] = {
+        latency: {
+          // Tinybench reports milliseconds; Bencher latency uses nanoseconds.
+          value: mToNs(task.result.latency.mean),
+          lower_value: mToNs(task.result.latency.mean - task.result.latency.sd),
+          upper_value: mToNs(task.result.latency.mean + task.result.latency.sd),
+        },
+        throughput: {
+          value: task.result.throughput.mean,
+          lower_value: task.result.throughput.mean - task.result.throughput.sd,
+          upper_value: task.result.throughput.mean + task.result.throughput.sd,
+        },
       }
     }
     return bmfResults
-  } catch (error) {
-    console.error(error)
-    return bmfResults
   } finally {
     if (pool != null) {
       await pool.destroy()

renovate.json

@@ -7,6 +7,12 @@
     ":maintainLockFilesWeekly"
   ],
   "packageRules": [
+    {
+      "matchDatasources": ["npm"],
+      "matchFileNames": ["**/deno.json"],
+      "minimumReleaseAge": "1 day",
+      "internalChecksFilter": "strict"
+    },
     {
       "matchFileNames": ["**/package.json"],
       "matchDepTypes": [