misc: avoid splitting node labels in the middle of unicode surrogate pairs

.github/scripts/git-get-shared-history.sh

@@ -20,6 +20,13 @@ set -euxo pipefail
 
 # We can always use some more history
 git -c protocol.version=2 fetch --deepen=100
+
+# Find out if the PR is coming from a fork
+base_clone_url=$(jq --raw-output '.pull_request.head.repo.clone_url' $GITHUB_EVENT_PATH)
+# If it is, we need the fork's history, too.
+if [[ $base_clone_url != "GoogleChrome/lighthouse.git" ]]; then
+  git -c protocol.version=2 fetch --deepen=100 "$base_clone_url"
+fi
 echo "History is deepened."
 
 if git merge-base HEAD origin/master > /dev/null; then

lighthouse-core/lib/page-functions.js

@@ -385,7 +385,9 @@ function getNodeLabel(node) {
     if (str.length <= maxLength) {
       return str;
     }
-    return str.slice(0, maxLength - 1) + '…';
+    // Take advantage of string iterator multi-byte character awareness.
+    // Regular `.slice` will ignore unicode character boundaries and lead to malformed text.
+    return Array.from(str).slice(0, maxLength - 1).join('') + '…';
   }
   const tagName = node.tagName.toLowerCase();
   // html and body content is too broad to be useful, since they contain all page content

lighthouse-core/test/lib/page-functions-test.js

@@ -116,6 +116,14 @@ describe('Page Functions', () => {
       assert.equal(pageFunctions.getNodeLabel(el).length, 80);
     });
 
+    it('Truncates long text containing unicode surrogate pairs', () => {
+      const el = dom.createElement('div');
+      // `getNodeLabel` truncates to 80 characters internally.
+      // We want to test a unicode character on the boundary.
+      el.innerText = Array(78).fill('a').join('') + '💡💡💡';
+      assert.equal(pageFunctions.getNodeLabel(el), Array(78).fill('a').join('') + '💡…');
+    });
+
     it('Uses tag name for html tags', () => {
       const el = dom.createElement('html');
       assert.equal(pageFunctions.getNodeLabel(el), 'html');