fix: use blob URL for Firefox downloads (data: URLs access-denied) (#3)

Squash merge of feat/ismis-helper into prod.

.gitea/workflows/build.yml

@@ -6,6 +6,11 @@ on:
     tags: ['v*']
   pull_request:
 
+# Ensure the job token can create releases and upload assets.
+permissions:
+  contents: write
+  releases: write
+
 jobs:
   build:
     runs-on: ubuntu-latest
@@ -21,6 +26,12 @@ jobs:
       - name: Install dev dependencies
         run: npm ci || npm install
 
+      - name: Install PyMuPDF (PDF text assertions)
+        run: |
+          python3 -m venv /tmp/venv
+          /tmp/venv/bin/pip install --quiet pymupdf
+          echo "/tmp/venv/bin" >> "$GITHUB_PATH"
+
       - name: Run tests
         run: npm test
 
@@ -36,11 +47,27 @@ jobs:
 
       - name: Publish release on version tag
         if: startsWith(github.ref, 'refs/tags/v')
-        # softprops/action-gh-release works on Gitea Actions: it uses the
-        # automatically-provided GITHUB_TOKEN to create a Release for the tag
-        # and attaches the supplied files.
-        uses: softprops/action-gh-release@v2
-        with:
-          files: dist/*.zip
-          name: ${{ github.ref_name }}
-          generate_release_notes: true
+        # Gitea-native release: softprops/action-gh-release targets GitHub's API
+        # and fails on Gitea (405). Instead we call the Gitea API directly with
+        # the runner-provided GITHUB_TOKEN and GITHUB_API_URL.
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          set -euo pipefail
+          REPO="${{ github.repository }}"
+          TAG="${{ github.ref_name }}"
+          API="${GITHUB_API_URL:-https://git.ranio.xyz/api/v1}"
+          # Create the release for the tag
+          RELEASE_JSON=$(curl -sS -X POST \
+            -H "Authorization: token ${GITHUB_TOKEN}" \
+            -H "Content-Type: application/json" \
+            -d "{\"tag_name\":\"${TAG}\",\"name\":\"${TAG}\",\"body\":\"ISMIS Helper ${TAG}\"}" \
+            "${API}/repos/${REPO}/releases")
+          RELEASE_ID=$(echo "${RELEASE_JSON}" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
+          # Attach both zips as release assets
+          for f in dist/*.zip; do
+            curl -sS -X POST \
+              -H "Authorization: token ${GITHUB_TOKEN}" \
+              -F "attachment=@${f}" \
+              "${API}/repos/${REPO}/releases/${RELEASE_ID}/assets?name=$(basename "${f}")"
+          done

README.md

@@ -10,17 +10,6 @@ extension for the USC ISMIS student portal that:
 
 One codebase builds for **both browsers** (see [Build](#build)).
 
-## Features
-
-- **Saved login profiles** (stored in `browser.storage.local`) — only one enabled at
-  a time, with a master on/off toggle.
-- **One-click Download PDF** button on the Study Load page.
-- PDF settings in the popup:
-  - Truncate long course descriptions to match the official PDF (default on).
-  - Show/hide `Year Level`, and set a custom Year Level value.
-- **Bundled Segoe UI fonts** for a 1:1 look with the official document.
-- Builds and packages automatically via Gitea Actions (zips + versioned releases).
-
 ## Install
 
 > Temporary/development install. For distribution, publish the built zips to
@@ -54,6 +43,54 @@ One codebase builds for **both browsers** (see [Build](#build)).
 > For a persistent install, package and publish via Firefox AMO / Chrome Web Store
 > with the same built zips.
 
+### Distribute to real users
+
+#### Firefox (self-hosted, temporary)
+
+> **Note:** Firefox blocks unsigned extensions for normal installs. The proper
+> fix is to sign the add-on via **addons.mozilla.org (AMO)** and publish it there
+> (or self-host a signed `.xpi`). AMO is currently **under maintenance**, so until
+> it returns, the only way to install the unsigned build is to lower Firefox's
+> signing check on each user's machine. This is a per-user browser setting and is
+> **not** suitable for a broad audience — use it for personal / small-group use.
+
+1. Build the Firefox variant (or download `ismis-helper-firefox.zip` from a release):
+   ```bash
+   node scripts/build.js firefox
+   ```
+2. In Firefox, open `about:config` and accept the warning.
+3. Search for `xpinstall.signatures.required` and set it to **`false`**.
+4. **Restart Firefox** for the change to take effect.
+5. Install the add-on from the release zip: drag `ismis-helper-firefox.zip` into
+   Firefox (or File → Open File… and select it).
+
+> When AMO is back online, prefer the signed route: publish the same zip to AMO
+> (or sign it with `web-ext sign`) so users install it normally without touching
+> `about:config`.
+
+#### Chrome (self-hosted)
+
+Chrome does **not** require signing for normal installs. Users can install the
+built `ismis-helper-chrome.zip` directly:
+
+1. Download `ismis-helper-chrome.zip` from a release and unzip it.
+2. Open `chrome://extensions`, enable **Developer mode**, and click **Load unpacked**.
+3. Select the unzipped folder.
+
+> For a store experience, publish the same zip to the Chrome Web Store (one-time
+> $5 developer registration).
+
+## Features
+
+- **Saved login profiles** (stored in `browser.storage.local`) — only one enabled at
+  a time, with a master on/off toggle.
+- **One-click Download PDF** button on the Study Load page.
+- PDF settings in the popup:
+  - Truncate long course descriptions to match the official PDF (default on).
+  - Show/hide `Year Level`, and set a custom Year Level value.
+- **Bundled Segoe UI fonts** for a 1:1 look with the official document.
+- Builds and packages automatically via Gitea Actions (zips + versioned releases).
+
 ## Build
 
 Requires **Node.js** (the `build.js` script itself is dependency-free; `npm install`

background.js

@@ -10,6 +10,27 @@ function errMsg(err) {
   return String((err && err.message) || err);
 }
 
+// Build a download URL from PDF bytes. Firefox's background is a real page
+// (URL.createObjectURL available) and supports blob: URLs in downloads.
+// Chrome MV3 runs the background as a service worker where URL.createObjectURL
+// is NOT available, but its downloads API accepts base64 data: URLs.
+// Firefox rejects data: URLs in browser.downloads.download with
+// "Access denied for URL data:...", so pick per-browser.
+function buildDownloadUrl(bytes) {
+  if (typeof URL !== 'undefined' && typeof URL.createObjectURL === 'function') {
+    const blob = new Blob([bytes], { type: 'application/pdf' });
+    return URL.createObjectURL(blob);
+  }
+  // Fallback: Chrome MV3 service worker. PDFs here are small (~150KB), well
+  // within data-URL length limits.
+  let binary = '';
+  const chunk = 0x8000;
+  for (let i = 0; i < bytes.length; i += chunk) {
+    binary += String.fromCharCode.apply(null, bytes.subarray(i, i + chunk));
+  }
+  return 'data:application/pdf;base64,' + btoa(binary);
+}
+
 browser.runtime.onMessage.addListener((msg, sender, sendResponse) => {
   if (msg && msg.type === 'generate-pdf') {
     generatePdf(msg.data)
@@ -32,8 +53,9 @@ function formatPeriod(period) {
 
 async function generatePdf(data) {
   const bytes = await ISMIStools.buildStudyLoadPdf(data);
-  const blob = new Blob([bytes], { type: 'application/pdf' });
-  const url = URL.createObjectURL(blob);
+  // Firefox accepts blob: object URLs; Chrome MV3 service worker has no
+  // URL.createObjectURL, so it falls back to a base64 data: URL there.
+  const url = buildDownloadUrl(bytes);
   const header = data.header || {};
   const year = parseInt(data.year, 10);
   const period = formatPeriod(data.period || header.semester || '');
@@ -55,12 +77,10 @@ async function generatePdf(data) {
     const onChanged = (delta) => {
       if (delta.id === id && delta.state && (delta.state.current === 'complete' || delta.state.current === 'interrupted')) {
         browser.downloads.onChanged.removeListener(onChanged);
-        URL.revokeObjectURL(url);
       }
     };
     browser.downloads.onChanged.addListener(onChanged);
   } catch (err) {
-    URL.revokeObjectURL(url);
     throw err;
   }
 }

manifest.json

@@ -1,7 +1,7 @@
 {
   "manifest_version": 3,
   "name": "ISMIS Helper",
-  "version": "1.2.0",
+  "version": "1.3.0",
   "description": "Auto-login, auto-accept privacy policy, and Study Load PDF for the USC ISMIS portal.",
   "permissions": ["storage", "downloads"],
   "host_permissions": ["https://ismis.usc.edu.ph/*"],

test/pdf.test.js

@@ -38,7 +38,19 @@ function extractPdfText(bytes) {
   return execFileSync('python3', ['-c', script], { encoding: 'utf8' });
 }
 
-test('buildStudyLoadPdf returns valid PDF bytes', async () => {
+// PyMuPDF (fitz) is only needed for the text-extraction assertions below. It is
+// installed in CI, but if it is missing locally we skip those assertions rather
+// than hard-failing.
+function fitzAvailable() {
+  try {
+    execFileSync('python3', ['-c', 'import fitz'], { stdio: 'ignore' });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+test('buildStudyLoadPdf returns valid PDF bytes', async (t) => {
   const bytes = await buildStudyLoadPdf(sampleData);
   assert.ok(bytes instanceof Uint8Array);
   const head = Buffer.from(bytes.slice(0, 5)).toString('latin1');
@@ -46,6 +58,11 @@ test('buildStudyLoadPdf returns valid PDF bytes', async () => {
   const doc = await PDFLib.PDFDocument.load(bytes);
   assert.strictEqual(doc.getPageCount(), 1);
 
+  if (!fitzAvailable()) {
+    t.skip('PyMuPDF (fitz) not available; skipping PDF text assertions');
+    return;
+  }
+
   const text = extractPdfText(bytes);
   assert.ok(text.includes('CIS 2101'), 'PDF text should contain course code CIS 2101');
   assert.ok(text.includes("STUDENT'S STUDY LOAD"), 'PDF text should contain the title');