Replaced EXPECT_TRUE with a LOG(WARNING) for automation commands to provide more information...

chrome/test/pyautolib/pyautolib.cc

Index: chrome/test/pyautolib/pyautolib.cc
diff --git a/chrome/test/pyautolib/pyautolib.cc b/chrome/test/pyautolib/pyautolib.cc
index 33814ac6a2d82a14a1d03041b42812c1375ab53b..91ba9712acb7d89033e137f067dfa901ea46fdd0 100644
--- a/chrome/test/pyautolib/pyautolib.cc
+++ b/chrome/test/pyautolib/pyautolib.cc
@@ -363,16 +363,22 @@ std::string PyUITestBase::_SendJSONRequest(int window_index,
                                            const std::string& request,
                                            int timeout) {
   std::string response;
+  bool response_status;
   if (window_index < 0) {  // Do not need to target a browser window.
-    EXPECT_TRUE(automation()->SendJSONRequest(request, timeout, &response));
+    response_status = automation()->SendJSONRequest(request, timeout,
+                                                    &response);
   } else {
     scoped_refptr<BrowserProxy> browser_proxy =
         automation()->GetBrowserWindow(window_index);
     EXPECT_TRUE(browser_proxy.get());
     if (browser_proxy.get()) {
-      EXPECT_TRUE(browser_proxy->SendJSONRequest(request, timeout, &response));
+      response_status = browser_proxy->SendJSONRequest(request, timeout,
+                                                       &response);
     }
   }
+  if (!response_status) {
+    LOG(FATAL) << "Automation failed: " << request;

[Nirnimesh, 2012/05/16 19:55:17]

No, do not make it fatal. There are tests to verify that a certain automation
command _does not_ succeed (like policy tests), in which case they catch the
JSONInterfaceError exception. That's a valid scenario.

A single failure does not mean the end of the world. It would be better to just
VLOG the request, and possibly output the time it took, since browser crashes
happen almost immediately whereas automation timeouts would take like 30 secs,
so the time's a clue.

+  }
   return response;
 }