Process null capabilities as a capability fetch failure.

chrome/browser/resources/print_preview/native_layer.js

Index: chrome/browser/resources/print_preview/native_layer.js
diff --git a/chrome/browser/resources/print_preview/native_layer.js b/chrome/browser/resources/print_preview/native_layer.js
index 22a94693450b9fed47d36c889a5c0af8b515e81e..7657d94325fb8c91a5a3dd75dd47b4b53212dbe6 100644
--- a/chrome/browser/resources/print_preview/native_layer.js
+++ b/chrome/browser/resources/print_preview/native_layer.js
@@ -550,6 +550,11 @@ cr.define('print_preview', function() {
      * @private
      */
     onUpdateWithPrinterCapabilities_: function(settingsInfo) {
+      if (settingsInfo.capabilities == null) {

[dpapad, 2017/05/01 17:39:28]

Why are we doing this in Javascript instead of C++? At [1] there is already code
that checks settingsInfo and calls the error callback directly. Here we let C++
call the success callback, and redirect to the error callback ourselves. It
seems unnecessary.

[1]
https://cs.chromium.org/chromium/src/chrome/browser/ui/webui/print_preview/pr...

[skau, 2017/05/01 22:51:26]

It crept up as a bug.  The code you referenced only checks that settingsInfo is
non-null.  But in fact, we get crashes if settingsInfo.capabilities is null. 
However, I'll make this an assert and strengthen the check in the handler. 

+        this.onFailedToGetPrinterCapabilities_(settingsInfo['printerId']);
+        return;
+      }
+
       var capsSetEvent = new Event(NativeLayer.EventType.CAPABILITIES_SET);
       capsSetEvent.settingsInfo = settingsInfo;
       this.dispatchEvent(capsSetEvent);