PDF: Do more robust page number parsing in OOP plugin.

chrome/browser/resources/pdf/pdf.js

Index: chrome/browser/resources/pdf/pdf.js
diff --git a/chrome/browser/resources/pdf/pdf.js b/chrome/browser/resources/pdf/pdf.js
index 70f53b7afc0068642049aacaeae66f4aae1c5929..c5920f877dc10c80a416c95c23b8fb7ab4ebc105 100644
--- a/chrome/browser/resources/pdf/pdf.js
+++ b/chrome/browser/resources/pdf/pdf.js
@@ -287,8 +287,10 @@ PDFViewer.prototype = {
     // Order is important as later actions can override the effects
     // of previous actions.
     if ('page' in paramsDictionary) {
-      // value is 1-based.
-      this.viewport_.goToPage(paramsDictionary['page'] - 1);
+      // |pageNumber| is 1-based, but goToPage() take a zero-based page number.
+      var pageNumber = parseInt(paramsDictionary['page']);
+      if (!isNaN(pageNumber))
+        this.viewport_.goToPage(pageNumber - 1);

[raymes, 2014/08/20 00:53:54]

What happens without this?

[Lei Zhang, 2014/08/20 00:57:19]

Without this CL, in goToPage(), we blow past the page bounds checks and try to
get the dimensions for page NaN.

     }
   },