fix: harden public download companion sync

mindspace-public-finish-sync.mjs

@@ -2,6 +2,7 @@ import fs from 'node:fs';
 import path from 'node:path';
 
 import { extractStaticPageLinks } from './mindspace-chat-save.mjs';
+import { DOWNLOADABLE_FILE_PATTERN } from './mindspace-html-download-links.mjs';
 
 /**
  * Public HTML finish-sync invariants (regression guard — do not simplify away).
@@ -10,45 +11,56 @@ import { extractStaticPageLinks } from './mindspace-chat-save.mjs';
  */
 
 const PUBLIC_HTML_PATH_PATTERN = /(?:^|[^a-z0-9_./-])(public\/[a-z0-9][a-z0-9._/-]{0,255}\.html)\b/i;
-const PUBLIC_DOCX_HREF_PATTERN = /href=["']([^"'#?\s]+\.docx)["']/gi;
+const PUBLIC_DOWNLOAD_ATTR_PATTERN = /(?:href|src)=["']([^"'#?\s]+)["']/gi;
+const DEFAULT_DOWNLOAD_SYNC_MIN_SIZE = 1024;
 
-function normalizePublicDocxHref(href) {
+function normalizePublicDownloadHref(href) {
   const clean = String(href ?? '')
     .split('?')[0]
     .split('#')[0]
     .replace(/^\.\//, '')
     .trim();
-  if (!clean || clean.includes('://') || clean.startsWith('data:')) return null;
+  if (
+    !clean ||
+    clean.includes('://') ||
+    clean.startsWith('data:') ||
+    !DOWNLOADABLE_FILE_PATTERN.test(clean)
+  ) {
+    return null;
+  }
   if (clean.startsWith('../oa/')) return clean.slice(3);
   if (clean.startsWith('oa/')) return clean;
   if (!clean.includes('/')) return `public/${clean}`;
   return clean;
 }
 
-function extractPublicDocxReferencesFromHtml(publishDir) {
+function extractPublicDownloadReferencesFromHtml(publishDir) {
   const root = path.resolve(String(publishDir ?? ''));
   const publicDir = path.join(root, 'public');
   if (!fs.existsSync(publicDir) || !fs.statSync(publicDir).isDirectory()) return [];
 
-  const refs = new Set();
+  const refs = new Map();
   for (const file of fs.readdirSync(publicDir)) {
     if (!file.toLowerCase().endsWith('.html')) continue;
     const content = fs.readFileSync(path.join(publicDir, file), 'utf8');
-    for (const match of content.matchAll(PUBLIC_DOCX_HREF_PATTERN)) {
-      const relativePath = normalizePublicDocxHref(match[1]);
-      if (relativePath) refs.add(relativePath);
+    const htmlRelativePath = `public/${file}`;
+    for (const match of content.matchAll(PUBLIC_DOWNLOAD_ATTR_PATTERN)) {
+      const relativePath = normalizePublicDownloadHref(match[1]);
+      if (relativePath && !refs.has(relativePath)) {
+        refs.set(relativePath, { relativePath, htmlRelativePath });
+      }
     }
   }
-  return [...refs];
+  return [...refs.values()];
 }
 
-function findLatestCompleteOaDocx(publishDir, { minSize = 8000 } = {}) {
+function listOaDownloadCandidates(publishDir, extension, { minSize = DEFAULT_DOWNLOAD_SYNC_MIN_SIZE } = {}) {
   const oaDir = path.join(path.resolve(String(publishDir ?? '')), 'oa');
-  if (!fs.existsSync(oaDir) || !fs.statSync(oaDir).isDirectory()) return null;
+  if (!fs.existsSync(oaDir) || !fs.statSync(oaDir).isDirectory()) return [];
 
-  const candidates = fs
+  return fs
     .readdirSync(oaDir)
-    .filter((name) => name.toLowerCase().endsWith('.docx'))
+    .filter((name) => name.toLowerCase().endsWith(extension))
     .map((name) => {
       const absolutePath = path.join(oaDir, name);
       const stat = fs.statSync(absolutePath);
@@ -56,46 +68,105 @@ function findLatestCompleteOaDocx(publishDir, { minSize = 8000 } = {}) {
     })
     .filter((item) => item.size >= minSize)
     .sort((a, b) => b.mtimeMs - a.mtimeMs || b.size - a.size);
-
-  return candidates[0] ?? null;
 }
 
-export function syncPublicDocxDownloads({ publishDir, minCompleteSize = 8000 } = {}) {
-  const root = path.resolve(String(publishDir ?? ''));
-  if (!root) return { synced: [], skipped: [] };
+function resolvePublicDownloadSyncSource(
+  root,
+  relativePath,
+  htmlRelativePath,
+  { minSize = DEFAULT_DOWNLOAD_SYNC_MIN_SIZE } = {},
+) {
+  const extension = path.extname(relativePath).toLowerCase();
+  const basename = path.basename(relativePath);
+  const htmlStemFilename = path.posix.basename(htmlRelativePath).replace(/\.html$/i, extension);
+  const exactCandidates = [
+    path.posix.join('oa', basename),
+    path.posix.join('oa', htmlStemFilename),
+  ];
 
-  const source = findLatestCompleteOaDocx(root, { minSize: minCompleteSize });
-  if (!source) return { synced: [], skipped: [] };
+  for (const candidate of exactCandidates) {
+    const absolutePath = path.resolve(root, candidate);
+    if (absolutePath === path.resolve(root, relativePath)) continue;
+    try {
+      const stat = fs.statSync(absolutePath);
+      if (stat.isFile() && stat.size >= minSize) {
+        return {
+          name: path.basename(candidate),
+          absolutePath,
+          size: stat.size,
+          strategy: 'exact_oa_match',
+        };
+      }
+    } catch {
+      // Ignore missing exact candidates and continue.
+    }
+  }
+
+  const fallbackCandidates = listOaDownloadCandidates(root, extension, { minSize });
+  if (fallbackCandidates.length === 1) {
+    return { ...fallbackCandidates[0], strategy: 'single_oa_fallback' };
+  }
+
+  return null;
+}
+
+export function syncPublicDocxDownloads({
+  publishDir,
+  minCompleteSize = DEFAULT_DOWNLOAD_SYNC_MIN_SIZE,
+} = {}) {
+  const root = path.resolve(String(publishDir ?? ''));
+  if (!root) return { synced: [], skipped: [], missing: [], source: null, sourceByRelativePath: {} };
 
   const synced = [];
   const skipped = [];
-  for (const relativePath of extractPublicDocxReferencesFromHtml(root)) {
+  const missing = [];
+  const sourceByRelativePath = {};
+  for (const reference of extractPublicDownloadReferencesFromHtml(root)) {
+    const { relativePath, htmlRelativePath } = reference;
     const destination = path.resolve(root, relativePath);
     if (destination !== root && !destination.startsWith(`${root}${path.sep}`)) {
       skipped.push(relativePath);
       continue;
     }
-    let shouldCopy = !fs.existsSync(destination) || !fs.statSync(destination).isFile();
-    if (!shouldCopy) {
-      try {
-        shouldCopy = fs.statSync(destination).size < source.size;
-      } catch {
-        shouldCopy = true;
-      }
-    }
-    if (!shouldCopy) {
+    const source = resolvePublicDownloadSyncSource(root, relativePath, htmlRelativePath, {
+      minSize: minCompleteSize,
+    });
+    const destinationExists = fs.existsSync(destination) && fs.statSync(destination).isFile();
+    if (destinationExists && !source) {
       skipped.push(relativePath);
       continue;
     }
+    if (!source) {
+      missing.push(relativePath);
+      continue;
+    }
+    if (destinationExists) {
+      try {
+        if (fs.statSync(destination).size >= source.size) {
+          skipped.push(relativePath);
+          continue;
+        }
+      } catch {
+        // Fall through and rewrite from source.
+      }
+    }
     try {
       fs.mkdirSync(path.dirname(destination), { recursive: true });
       fs.copyFileSync(source.absolutePath, destination);
       synced.push(relativePath);
+      sourceByRelativePath[relativePath] = source.name;
     } catch {
       skipped.push(relativePath);
     }
   }
-  return { synced, skipped, source: source.name };
+  const sourceNames = [...new Set(Object.values(sourceByRelativePath))];
+  return {
+    synced,
+    skipped,
+    missing,
+    source: sourceNames.length === 1 ? sourceNames[0] : null,
+    sourceByRelativePath,
+  };
 }
 
 function messageText(message) {

mindspace-public-finish-sync.test.mjs

@@ -258,6 +258,58 @@ test('syncPublicDocxDownloads skips public docx that is already up to date', ()
   }
 });
 
+test('syncPublicDocxDownloads copies same-named oa docx even when it is small', () => {
+  const publishDir = fs.mkdtempSync(path.join(os.tmpdir(), 'public-docx-sync-'));
+  try {
+    const oaDir = path.join(publishDir, 'oa');
+    const publicDir = path.join(publishDir, 'public');
+    fs.mkdirSync(oaDir, { recursive: true });
+    fs.mkdirSync(publicDir, { recursive: true });
+    fs.writeFileSync(path.join(oaDir, 'industry-distribution-report.docx'), 'x'.repeat(3294));
+    fs.writeFileSync(
+      path.join(publicDir, 'industry-distribution-report.html'),
+      '<a href="industry-distribution-report.docx" download>download</a>',
+    );
+
+    const result = syncPublicDocxDownloads({ publishDir, minCompleteSize: 1024 });
+    assert.deepEqual(result.synced, ['public/industry-distribution-report.docx']);
+    assert.deepEqual(result.missing, []);
+    assert.equal(result.source, 'industry-distribution-report.docx');
+    assert.equal(
+      fs.readFileSync(path.join(publicDir, 'industry-distribution-report.docx'), 'utf8').length,
+      3294,
+    );
+  } finally {
+    fs.rmSync(publishDir, { recursive: true, force: true });
+  }
+});
+
+test('syncPublicDocxDownloads reports missing public docx when no safe source exists', () => {
+  const publishDir = fs.mkdtempSync(path.join(os.tmpdir(), 'public-docx-sync-'));
+  try {
+    const oaDir = path.join(publishDir, 'oa');
+    const publicDir = path.join(publishDir, 'public');
+    fs.mkdirSync(oaDir, { recursive: true });
+    fs.mkdirSync(publicDir, { recursive: true });
+    fs.writeFileSync(path.join(oaDir, '暑假计划.docx'), 'x'.repeat(4084));
+    fs.writeFileSync(path.join(oaDir, '韩国旅游攻略.docx'), 'x'.repeat(5889));
+    fs.writeFileSync(
+      path.join(publicDir, 'industry-distribution-report.html'),
+      '<a href="industry-distribution-report.docx" download>download</a>',
+    );
+
+    const result = syncPublicDocxDownloads({ publishDir, minCompleteSize: 1024 });
+    assert.deepEqual(result.synced, []);
+    assert.deepEqual(result.missing, ['public/industry-distribution-report.docx']);
+    assert.equal(
+      fs.existsSync(path.join(publicDir, 'industry-distribution-report.docx')),
+      false,
+    );
+  } finally {
+    fs.rmSync(publishDir, { recursive: true, force: true });
+  }
+});
+
 test('materializeMissingPublicHtmlWrites replaces existing real html when content changed', () => {
   const publishDir = fs.mkdtempSync(path.join(os.tmpdir(), 'public-finish-sync-'));
   try {

server.mjs

@@ -4276,7 +4276,7 @@ api.get('/sessions/:sessionId/events', async (req, res, next) => {
         messages = null;
       }
     }
-    await syncPublicHtmlAfterFinish({
+    const syncResult = await syncPublicHtmlAfterFinish({
       messages,
       currentUser: req.currentUser,
       publishDir,
@@ -4285,6 +4285,11 @@ api.get('/sessions/:sessionId/events', async (req, res, next) => {
           ? (userId, options) => mindSpaceAssets.syncWorkspaceAssets(userId, options)
           : null,
     });
+    if (Array.isArray(syncResult?.docxSync?.missing) && syncResult.docxSync.missing.length > 0) {
+      console.warn(
+        `[MindSpace] missing public download files after finish for user ${uid}: ${syncResult.docxSync.missing.join(', ')}`,
+      );
+    }
     await syncUserGeneratedPages(uid);
   };
   return tkmindProxy.proxySessionEvents(req, res, sessionId, {