PlzNavigate: fix regression on the FYI bot.

third_party/WebKit/Source/core/loader/FrameLoader.cpp

Index: third_party/WebKit/Source/core/loader/FrameLoader.cpp
diff --git a/third_party/WebKit/Source/core/loader/FrameLoader.cpp b/third_party/WebKit/Source/core/loader/FrameLoader.cpp
index dafe916b9563d86da938dbe4cea8e5ddcac4a5ee..04e80206bc7d59b8909f0015f440d7019b3ac3a3 100644
--- a/third_party/WebKit/Source/core/loader/FrameLoader.cpp
+++ b/third_party/WebKit/Source/core/loader/FrameLoader.cpp
@@ -1461,6 +1461,7 @@ void FrameLoader::detach() {
 
 void FrameLoader::loadFailed(DocumentLoader* loader,
                              const ResourceError& error) {
+  m_isNavigationHandledByClient = false;

[Nate Chapin, 2017/01/11 20:30:46]

This isn't obvious to me. A blink-internal navigation failed, why would an
external navigation also fail?

[arthursonzogni, 2017/01/12 12:33:29]

If an external navigation is in progress, what other kind of load could fail? I
don't know how it works, but I expected that there could be only one thing to
load at a time.

If this is not a good idea to put it here, I can move this statement in the
upper level:
* 0) FrameLoader::loadFailed()
* 1) WebLocalFrameImpl::maybeRenderFallbackContent()
* 2) RenderFrameImpl::OnFailedNavigation()

But I will need to create functions to set m_isNavigationHandledByClient from
outside of the FrameLoader. Do you think to put this line of code in 1) is okay?

Note: In the future, I should try to remove the m_isNavigationHandledByClient as
soon as I understood better how the frameloader works and how it is connected to
the testrunner.

   if (!error.isCancellation() && m_frame->owner()) {
     // FIXME: For now, fallback content doesn't work cross process.
     if (m_frame->owner()->isLocal())