Probe complete canonical WordPress state

Author: chubes4

packages/runtime-cloudflare/README.md

@@ -20,6 +20,10 @@ The WordPress server corpus is a separate deployment artifact, never a Worker-mo
 
 ## MDI Init Diagnostics
 
+Every `mdi-*` lifecycle diagnostic materializes the validated packaged canonical MDI seed and the separately managed R2 WordPress runtime artifact in a disposable PHP-WASM runtime. They never acquire a coordinator lease, read or write a canonical R2 revision, or promote state. The old three-option fixture was intentionally removed because it does not represent complete MDI WordPress state.
+
+`?phase=canonical-wordpress` performs a full normal WordPress boot with that exact seed and artifact outside the coordinator, then returns WordPress version, memory, and non-sensitive table counts. Its widget-state count includes the `widget_*` options and `sidebars_widgets`. `?phase=canonical-bootstrap-setup` additionally runs the normal bootstrap setup with a probe-only internal password and fixed probe origin, flushes only the disposable runtime, and returns changed-path counts. Neither probe returns password, option, user, or content values.
+
 The diagnostic-only `?phase=` probes `mdi-init-callbacks`, `mdi-init-exclude-scheduling`, `mdi-init-exclude-block-registration`, `mdi-init-exclude-theme-patterns-styles`, `mdi-init-exclude-widgets`, `mdi-init-exclude-rest-connectors-sitemaps`, and `mdi-init-exclude-initial-content-types` stop at `init`. The inventory probe returns sorted callback identifiers by priority. Each exclusion probe removes only its fixed callback group, runs `init` once, and returns before subsequent WordPress lifecycle work.
 
 The fixed widget probes split `wp_widgets_init` without changing its normal path: `mdi-widgets-callbacks` returns the sorted nested `widgets_init` inventory before `init`; `mdi-widgets-constructors` runs only the core default widget registrations and reports widget class count; `mdi-widgets-hooks` adds the nested action; `mdi-widgets-factory` runs only `WP_Widget_Factory::_register_widgets`; and `mdi-widgets-remaining-hooks` runs the remaining nested callbacks. The bounded direct-registration phases are `mdi-widgets-direct-basic-classic-first`, `mdi-widgets-direct-basic-classic-second`, `mdi-widgets-direct-media`, `mdi-widgets-direct-custom-html-block`, `mdi-widgets-direct-block`, and `mdi-widgets-direct-custom-html`. They remove every normal nested callback, then call `_register()` only on their fixed allowlisted factory objects and return attempted classes, count, completion, and memory. `mdi-widgets-option-reads` reports only fixed widget option names, grouped by `get_option()` presence and direct `$wpdb` SQLite-row presence; it returns no option values. `mdi-widgets-get-settings` calls only `get_settings()` on each default factory object, with `mdi-widgets-get-settings-first` and `mdi-widgets-get-settings-second` as fixed halves; each reports completed classes and array-shape dimensions only. `mdi-widgets-register-one` preloads the fixed `widget_text` option, calls that object's `get_settings()`, then calls only `_register_one()` for its fixed instance. Every phase uses fixed source identities and returns before later lifecycle work; none accepts callback names, removal rules, widget classes, or option names from request input.
@@ -30,6 +34,6 @@ The fixed widget probes split `wp_widgets_init` without changing its normal path
 2. Run `npm run provision:cloudflare-wordpress-runtime-corpus -- --local --persist-to <directory>` to verify and upload the exact content-addressed artifact into isolated local R2 storage. Use `--remote` only for an explicit deployment operation.
 3. Run `npm run test:cloudflare-runtime` for routing, canonical-state, artifact validation, source contract, and TypeScript coverage.
 4. Run `npm run cloudflare:dry-run` to compile the Worker without creating Cloudflare resources.
-5. Run `npm run cloudflare:local-gate` for isolated local workerd evidence. It generates and provisions the artifact before workerd starts, then injects stable test-only admin-password and auth-secret values, verifies the login form, cookie-authenticated admin redirect, authenticated REST post publication, public rendering, representative frontend/admin/editor assets, PHP diagnostics, an existing authenticated cookie after cold restart, and a fresh login after restart.
+5. Run `npm run cloudflare:local-gate` for isolated local workerd evidence. It generates and provisions the artifact before workerd starts, then verifies the complete-seed canonical boot and ephemeral bootstrap probes, injects stable test-only admin-password and auth-secret values, verifies the login form, cookie-authenticated admin redirect, authenticated REST post publication, public rendering, representative frontend/admin/editor assets, PHP diagnostics, an existing authenticated cookie after cold restart, and a fresh login after restart.
 
 This document describes local candidate verification only. It does not claim remote deployment.

packages/runtime-cloudflare/src/worker.ts

@@ -517,7 +517,7 @@ async function runBootProbe(phase: string, bucket: R2Bucket): Promise<Response>
     try {
       const wordpress = await materializeWordPressServerFiles(php, bucket)
       await materializeMarkdownDatabaseIntegration(php)
-      materializeRuntimeFiles(php, MARKDOWN_ROOT, initialMarkdownFiles())
+      materializeRuntimeFiles(php, MARKDOWN_ROOT, await packagedCanonicalMarkdownSeed())
       const evidence = (await php.run({
         code: "<?php echo json_encode(['dropin' => file_exists('/wordpress/wp-content/db.php'), 'storage' => file_exists('/wordpress/wp-content/plugins/markdown-database-integration/inc/class-wp-markdown-storage.php'), 'siteurl' => file_exists('/wordpress/wp-content/markdown/_options/siteurl.json')]);",
       })).text.trim()
@@ -528,7 +528,7 @@ async function runBootProbe(phase: string, bucket: R2Bucket): Promise<Response>
   }
 
   if (phase === "mdi-shortinit") {
-    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, initialMarkdownFiles(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
+    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, await packagedCanonicalMarkdownSeed(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
     try {
       const evidence = (await runtime.php.run({
         code: "<?php define('SHORTINIT', true); require '/wordpress/wp-load.php'; echo json_encode(['wordpressVersion' => $wp_version, 'markdownDropin' => defined('MARKDOWN_DB_DROPIN'), 'markdownMode' => defined('MARKDOWN_DB_MODE') ? MARKDOWN_DB_MODE : '']);",
@@ -540,7 +540,7 @@ async function runBootProbe(phase: string, bucket: R2Bucket): Promise<Response>
   }
 
   if (phase === "mdi-wordpress" || phase === "mdi-option" || phase === "mdi-insert") {
-    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, initialMarkdownFiles(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
+    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, await packagedCanonicalMarkdownSeed(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
     try {
       const operation = phase === "mdi-option"
         ? "$updated = update_option('wp_codebox_mdi_probe', 1); $result = ['updated' => $updated];"
@@ -557,7 +557,7 @@ async function runBootProbe(phase: string, bucket: R2Bucket): Promise<Response>
   }
 
   if (["mdi-includes", "mdi-embed", "mdi-textdomain", "mdi-ai-client", "mdi-plugin-constants", "mdi-muplugins", "mdi-plugins", "mdi-globals", "mdi-theme", "mdi-site-health-class", "mdi-site-health", "mdi-current-user", "mdi-init", "mdi-wp-loaded", "mdi-init-callbacks", "mdi-init-exclude-scheduling", "mdi-init-exclude-block-registration", "mdi-init-exclude-theme-patterns-styles", "mdi-init-exclude-widgets", "mdi-init-exclude-rest-connectors-sitemaps", "mdi-init-exclude-initial-content-types", "mdi-widgets-callbacks", "mdi-widgets-constructors", "mdi-widgets-hooks", "mdi-widgets-factory", "mdi-widgets-remaining-hooks", "mdi-widgets-direct-basic-classic-first", "mdi-widgets-direct-basic-classic-second", "mdi-widgets-direct-media", "mdi-widgets-direct-custom-html-block", "mdi-widgets-direct-block", "mdi-widgets-direct-custom-html", "mdi-widgets-option-reads", "mdi-widgets-get-settings", "mdi-widgets-get-settings-first", "mdi-widgets-get-settings-second", "mdi-widgets-register-one"].includes(phase)) {
-    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, initialMarkdownFiles(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
+    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, await packagedCanonicalMarkdownSeed(), new Uint8Array(markdownPrimaryBootstrapIndex), SITE_URL, {}, bucket)
     try {
       const evidence = (await runtime.php.run({ code: wordpressProbeCode(phase) })).text.trim()
       return probeResponse(phase, JSON.parse(evidence) as Record<string, string>)
@@ -566,6 +566,37 @@ async function runBootProbe(phase: string, bucket: R2Bucket): Promise<Response>
     }
   }
 
+  if (phase === "canonical-wordpress" || phase === "canonical-bootstrap-setup") {
+    const canonicalSeed = await packagedCanonicalMarkdownSeed()
+    const runtime = await bootWordPressRuntime("do-not-attempt-installing", true, true, undefined, canonicalSeed, new Uint8Array(markdownPrimaryBootstrapIndex), "https://canonical-probe.invalid", {}, bucket)
+    try {
+      if (phase === "canonical-bootstrap-setup") {
+        const passwordFile = "/tmp/wp-codebox-canonical-bootstrap-probe-password"
+        runtime.php.writeFile(passwordFile, new TextEncoder().encode("probe-only-password"))
+        const output = (await runtime.php.run({ code: canonicalBootstrapSetupCode(passwordFile, "https://canonical-probe.invalid") })).text.trim()
+        if (output !== "flushed") throw new Error("Canonical bootstrap setup probe did not confirm its ephemeral flush.")
+        const changes = canonicalChangedPathCounts(canonicalSeed, collectRuntimeFiles(runtime.php, MARKDOWN_ROOT))
+        return probeResponse(phase, { wordpressVersion: runtime.wordpressVersion, canonicalSeedFiles: canonicalSeed.length, ...changes })
+      }
+      const evidence = (await runtime.php.run({ code: `<?php
+require '/wordpress/wp-load.php';
+echo json_encode([
+  'wordpressVersion' => get_bloginfo('version'),
+  'postCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->posts}"),
+  'pageCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->posts} WHERE post_type = 'page'"),
+  'userCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->users}"),
+  'optionCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->options}"),
+  'widgetOptionCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->options} WHERE option_name LIKE 'widget\\_%'"),
+  'widgetStateOptionCount' => (int) $wpdb->get_var("SELECT COUNT(*) FROM {\$wpdb->options} WHERE option_name LIKE 'widget\\_%' OR option_name = 'sidebars_widgets'"),
+  'memoryBytes' => memory_get_usage(true),
+  'peakMemoryBytes' => memory_get_peak_usage(true),
+]);` })).text.trim()
+      return probeResponse(phase, { canonicalSeedFiles: canonicalSeed.length, ...JSON.parse(evidence) as Record<string, number | string> })
+    } finally {
+      runtime.php.exit()
+    }
+  }
+
   if (phase?.startsWith("seeded-")) {
     const runtime = await bootWordPressRuntime(
       "do-not-attempt-installing",
@@ -1006,18 +1037,6 @@ async function bootWordPressRuntime(
   return { php, requestHandler, wordpressVersion }
 }
 
-function initialMarkdownFiles(): RuntimeFile[] {
-  const options = [
-    { option_id: 1, option_name: "siteurl", option_value: SITE_URL, autoload: "on" },
-    { option_id: 2, option_name: "home", option_value: SITE_URL, autoload: "on" },
-    { option_id: 3, option_name: "blogname", option_value: "WP Codebox Cloudflare Runtime", autoload: "on" },
-  ]
-  return options.map((option) => ({
-    path: `_options/${option.option_name}.json`,
-    bytes: new TextEncoder().encode(JSON.stringify(option, null, 2)),
-  }))
-}
-
 function materializeRuntimeFiles(php: PHP, root: string, files: RuntimeFile[]): void {
   for (const file of files) {
     const destination = `${root}/${file.path}`
@@ -1043,6 +1062,21 @@ function collectRuntimeFiles(php: PHP, root: string): RuntimeFile[] {
   return files.sort((left, right) => left.path.localeCompare(right.path))
 }
 
+function canonicalChangedPathCounts(before: RuntimeFile[], after: RuntimeFile[]): { createdPathCount: number; changedPathCount: number; deletedPathCount: number } {
+  const beforeByPath = new Map(before.map((file) => [file.path, file.bytes]))
+  const afterByPath = new Map(after.map((file) => [file.path, file.bytes]))
+  let createdPathCount = 0
+  let changedPathCount = 0
+  let deletedPathCount = 0
+  for (const [path, bytes] of afterByPath) {
+    const previous = beforeByPath.get(path)
+    if (!previous) createdPathCount++
+    else if (previous.byteLength !== bytes.byteLength || previous.some((byte, index) => byte !== bytes[index])) changedPathCount++
+  }
+  for (const path of beforeByPath.keys()) if (!afterByPath.has(path)) deletedPathCount++
+  return { createdPathCount, changedPathCount, deletedPathCount }
+}
+
 async function sha256Hex(bytes: Uint8Array): Promise<string> {
   const digest = await crypto.subtle.digest("SHA-256", Uint8Array.from(bytes).buffer)
   return Array.from(new Uint8Array(digest), (byte) => byte.toString(16).padStart(2, "0")).join("")

scripts/cloudflare-local-gate.mjs

@@ -16,6 +16,7 @@ try {
   await run("npm", ["run", "generate:cloudflare-wordpress-runtime-corpus"])
   await run("npm", ["run", "provision:cloudflare-wordpress-runtime-corpus", "--", "--local", "--persist-to", stateDirectory])
   await startWorker()
+  await assertCanonicalProbes()
   await assertConcurrentMutations()
   const adminHtml = await login()
   const editorHtml = await assertPostNewEditor()
@@ -166,6 +167,20 @@ async function assertHealthResponse() {
   if (response.status !== 200 || body.schema !== "wp-codebox/cloudflare-runtime-health/v1" || body.marker !== "wp-codebox-cloudflare-runtime-health" || body.phpVersion !== "8.5.8" || typeof body.wordpressVersion !== "string" || body.execution?.status !== "ok") throw new Error(`Unexpected Cloudflare runtime health envelope: ${JSON.stringify(body)}`)
 }
 
+async function assertCanonicalProbes() {
+  const wordpress = await (await request(`${origin}/?phase=canonical-wordpress`)).json()
+  const setup = await (await request(`${origin}/?phase=canonical-bootstrap-setup`)).json()
+  const counts = ["canonicalSeedFiles", "postCount", "pageCount", "userCount", "optionCount", "widgetOptionCount", "widgetStateOptionCount", "memoryBytes", "peakMemoryBytes"]
+  if (wordpress.completed !== true || typeof wordpress.evidence?.wordpressVersion !== "string" || counts.some((key) => !Number.isInteger(wordpress.evidence?.[key])) || wordpress.evidence.widgetStateOptionCount !== 19) {
+    throw new Error(`Canonical WordPress probe did not boot the complete seed: ${JSON.stringify(wordpress)}`)
+  }
+  const changedCounts = ["createdPathCount", "changedPathCount", "deletedPathCount"]
+  if (setup.completed !== true || typeof setup.evidence?.wordpressVersion !== "string" || changedCounts.some((key) => !Number.isInteger(setup.evidence?.[key])) || setup.evidence.changedPathCount < 1) {
+    throw new Error(`Canonical bootstrap setup probe did not produce ephemeral changes: ${JSON.stringify(setup)}`)
+  }
+  console.log(`Canonical probes: WordPress ${wordpress.evidence.wordpressVersion}, ${wordpress.evidence.widgetOptionCount} widget_* options plus sidebars_widgets, ${setup.evidence.changedPathCount} changed ephemeral paths.`)
+}
+
 async function assertWordPressPage(target, label) {
   const response = await request(target)
   const body = await response.text()

tests/cloudflare-runtime.test.ts

@@ -225,6 +225,21 @@ test("Cloudflare MDI init diagnostics use fixed callback inventories and exclusi
   assert.doesNotMatch(worker, /searchParams\.get\([^)]*callback|searchParams\.get\([^)]*exclude|searchParams\.get\([^)]*widget/)
 })
 
+test("Cloudflare MDI lifecycle diagnostics use the complete packaged seed without canonical persistence", async () => {
+  const worker = await readFile(new URL("../packages/runtime-cloudflare/src/worker.ts", import.meta.url), "utf8")
+  const probes = worker.slice(worker.indexOf("async function runBootProbe"), worker.indexOf("if (phase?.startsWith(\"seeded-\"))"))
+
+  assert.equal((probes.match(/initialMarkdownFiles/g) ?? []).length, 0)
+  assert.ok((probes.match(/await packagedCanonicalMarkdownSeed\(\)/g) ?? []).length >= 4)
+  assert.match(probes, /phase === "canonical-wordpress" \|\| phase === "canonical-bootstrap-setup"/)
+  assert.match(probes, /materializeWordPressServerFiles/)
+  assert.match(probes, /canonicalBootstrapSetupCode\(passwordFile, "https:\/\/canonical-probe\.invalid"\)/)
+  assert.match(probes, /canonicalChangedPathCounts\(canonicalSeed, collectRuntimeFiles/)
+  assert.doesNotMatch(probes, /persistMarkdownRevision|commitLease|coordinatorCall/)
+  assert.match(worker, /'widgetOptionCount' => \(int\) \$wpdb->get_var/)
+  assert.match(worker, /function canonicalChangedPathCounts/)
+})
+
 test("Cloudflare keeps PHP-WASM in the entry Worker and uses the Durable Object only for leases", async () => {
   const worker = await readFile(new URL("../packages/runtime-cloudflare/src/worker.ts", import.meta.url), "utf8")
   const coordinator = await readFile(new URL("../packages/runtime-cloudflare/src/state-coordinator.ts", import.meta.url), "utf8")