Merge pull request #8979 from timvandermeij/downloads

Transform Web Archive URLs to avoid downloading an HTML page instead of the PDF file

test/downloadutils.js

@@ -22,7 +22,22 @@ var crypto = require('crypto');
 var http = require('http');
 var https = require('https');
 
+function rewriteWebArchiveUrl(url) {
+  // Web Archive URLs need to be transformed to add `if_` after the ID.
+  // Without this, an HTML page containing an iframe with the PDF file
+  // will be served instead (issue 8920).
+  var webArchiveRegex =
+    /(^https?:\/\/web\.archive\.org\/web\/)(\d+)(\/https?:\/\/.+)/g;
+  var urlParts = webArchiveRegex.exec(url);
+  if (urlParts) {
+    return urlParts[1] + (urlParts[2] + 'if_') + urlParts[3];
+  }
+  return url;
+}
+
 function downloadFile(file, url, callback, redirects) {
+  url = rewriteWebArchiveUrl(url);
+
   var completed = false;
   var protocol = /^https:\/\//.test(url) ? https : http;
   protocol.get(url, function (response) {

test/pdfs/geothermal.pdf.link

@@ -1 +1 @@
-http://web.archive.org/web/20150212141833/http://geothermal.inel.gov/publications/future_of_geothermal_energy.pdf
+https://web.archive.org/web/20170930174755/https://www.pdf-archive.com/2017/09/30/future-of-geothermal-energy/future-of-geothermal-energy.pdf

test/pdfs/issue8169.pdf.link

@@ -1 +1 @@
-http://web.archive.org/save/_embed/http://210.243.166.143/prob1.pdf
+https://web.archive.org/web/20170930161657/http://210.243.166.143/prob1.pdf