Fire visibilitychange event on unload (behind the flag)

third_party/WebKit/Source/web/ChromeClientImpl.cpp

Index: third_party/WebKit/Source/web/ChromeClientImpl.cpp
diff --git a/third_party/WebKit/Source/web/ChromeClientImpl.cpp b/third_party/WebKit/Source/web/ChromeClientImpl.cpp
index e2e82f9ed76e182a84314984d44107dee468829e..bbe850b0b3611aeaf3187c8f599dec3afd036df4 100644
--- a/third_party/WebKit/Source/web/ChromeClientImpl.cpp
+++ b/third_party/WebKit/Source/web/ChromeClientImpl.cpp
@@ -864,7 +864,9 @@ bool ChromeClientImpl::shouldOpenModalDialogDuringPageDismissal(const DialogType
     ASSERT_WITH_SECURITY_IMPLICATION(0 <= dialog);
     ASSERT_WITH_SECURITY_IMPLICATION(dialog < static_cast<int>(WTF_ARRAY_LENGTH(kDialogs)));
 
-    const char* const kDismissals[] = { "beforeunload", "pagehide", "unload" };
+    // This array must be consistent with PageDismissalType enum defined in

[dcheng, 2016/03/11 04:20:38]

Do we need to preserve the relative order? This is currently logged in UMA. That
being said, I think the UMA stat is not very useful at this point. So maybe we
should just remove the UMA recording for this?

[kinuko, 2016/03/11 05:12:50]

Hmm, true. I removed the UMA part & changed these arrays into switch/case.  How
does it look to you now?

Also-- assuming that this patch's not looking terrible I'm going to add a few
more tests & start preparing intent process!

+    // Document.h.
+    const char* const kDismissals[] = { "beforeunload", "pagehide", "visibilitychange", "unload" };
     int dismissal = static_cast<int>(dismissalType) - 1; // Exclude NoDismissal.
     ASSERT_WITH_SECURITY_IMPLICATION(0 <= dismissal);
     ASSERT_WITH_SECURITY_IMPLICATION(dismissal < static_cast<int>(WTF_ARRAY_LENGTH(kDismissals)));