Description of problem: Runtime JavaScript error that occurs in Eldad's large scale env. when using Firefox 37. Doesn't occur in Chrome. Seems like an application-level null pointer exception triggered from our GWT data table widget in response to some DOM event. Version-Release number of selected component (if applicable): 3.5.1 VT14.1 Steps to Reproduce: "login to rhevm webadmin > after a while the error appears (if not, click on hosts tab and wait for a while). bear in mind it was reproduced over WAN." Actual results: Unresponsive script dialog appears: "A script on this page may be busy, or it may have stopped responding. You can stop the script now, open the script in the debugger, or let the script continue." Expected results: Unresponsive script dialog does not appear. Additional info - JS stack trace: "Sun Apr 26 17:58:06 GMT+300 2015 com.google.gwt.logging.client.LogConfiguration SEVERE: (TypeError) : elem is null com.google.gwt.core.client.JavaScriptException: (TypeError) : elem is null at Unknown.com_google_gwt_dom_client_DOMImpl_$getNextSiblingElement__Lcom_google_gwt_dom_client_DOMImpl_2Lcom_google_gwt_dom_client_Element_2Lcom_google_gwt_dom_client_Element_2(Unknown Source) at Unknown.com_google_gwt_dom_client_Element_$getNextSiblingElement__Lcom_google_gwt_dom_client_Element_2Lcom_google_gwt_dom_client_Element_2(Unknown Source) at Unknown.com_google_gwt_cell_client_CompositeCell_isEditing__Lcom_google_gwt_cell_client_Cell$Context_2Lcom_google_gwt_dom_client_Element_2Ljava_lang_Object_2Z(Unknown Source) at Unknown.com_google_gwt_user_cellview_client_AbstractCellTable_$fireEventToCell__Lcom_google_gwt_user_cellview_client_AbstractCellTable_2Lcom_google_gwt_user_client_Event_2Ljava_lang_String_2Lcom_google_gwt_dom_client_Element_2Ljava_lang_Object_2Lcom_google_gwt_cell_client_Cell$Context_2Lcom_google_gwt_cell_client_HasCell_2V(Unknown Source) at Unknown.com_google_gwt_user_cellview_client_AbstractCellTable_$onBrowserEvent2__Lcom_google_gwt_user_cellview_client_AbstractCellTable_2Lcom_google_gwt_user_client_Event_2V(Unknown Source) at Unknown.org_ovirt_engine_ui_common_widget_table_AbstractActionTable$1_onBrowserEvent2__Lcom_google_gwt_user_client_Event_2V(Unknown Source) at Unknown.com_google_gwt_user_cellview_client_AbstractHasData_onBrowserEvent__Lcom_google_gwt_user_client_Event_2V(Unknown Source) at Unknown.com_google_gwt_user_client_DOM_dispatchEventImpl__Lcom_google_gwt_user_client_Event_2Lcom_google_gwt_user_client_Element_2Lcom_google_gwt_user_client_EventListener_2V(Unknown Source) at Unknown.com_google_gwt_user_client_DOM_dispatchEvent__Lcom_google_gwt_user_client_Event_2Lcom_google_gwt_user_client_Element_2Lcom_google_gwt_user_client_EventListener_2V(Unknown Source) at Unknown.com_google_gwt_user_client_impl_DOMImplStandard_$initEventSystem__Lcom_google_gwt_user_client_impl_DOMImplStandard_2V/com_google_gwt_user_client_impl_DOMImplStandard_dispatchEvent<(Unknown Source) at Unknown.com_google_gwt_core_client_impl_Impl_apply__Ljava_lang_Object_2Ljava_lang_Object_2Ljava_lang_Object_2Ljava_lang_Object_2(Unknown Source) at Unknown.com_google_gwt_core_client_impl_Impl_entry0__Ljava_lang_Object_2Ljava_lang_Object_2Ljava_lang_Object_2Ljava_lang_Object_2(Unknown Source) at Unknown.com_google_gwt_core_client_impl_Impl_entry__Lcom_google_gwt_core_client_JavaScriptObject_2Lcom_google_gwt_core_client_JavaScriptObject_2/<(Unknown Source)
See bug 1207299 for reference and description of the original issue.
Greg, please be in touch with Eldad [emarcian] for reproducing, assistance in testing, etc. Note that this seems to happen only in FF (not in Chrome), and maybe only in WAN, rather than on LAN (strange...).
we may want to backport that into 3.5.z.
note: this reproduces only for FF versions > 36. I strongly recommend trying to test on FF38, which is about to become the standard FF version provided within RHEL.
I'm unable to reproduce on my own, but I don't think I need to since we have the stack trace. I suspect the issue is this line of code in GWT's CompositeCell: public boolean isEditing(Context context, Element parent, C value) { ** Element curChild = getContainerElement(parent).getFirstChildElement(); ** for (HasCell<C, ?> hasCell : hasCells) { ... curChild is never checked for being null later in the method, but we're seeing it be null. (It's probably null in the rare circumstance of two concurrent grid refreshes -- certainly possible over a latent network.) So I'd classify this as a GWT bug. We can easily work around by overriding "isEditing" and safely handling nulls in there.
after discussing with Greg: we need to fix this issue for 3.6 first. however, in 3.6 we have made significant changes: (a) upgraded GWT and GWT-related dependencies. (b) refactored a lot of code around table-cells for the PatternFly-tooltips changes. we don't even know whether this issue reproduces in 3.6. @Eldad - do you have a 3.6 environment (similar to your 3.5 environment) on which you can attempt to reproduce the problem?
we need to invest some time and effort for this, in progress
Met with Eldad today. He is going to try to reproduce on master. He'll follow up in a day or two.
Eldad reports the issue cannot be reproduced on master. This is good -- I think that means it was a bug in version of GWT in oVirt 3.5.x, and has since been fixed in the newer version of GWT we're using in oVirt 3.6. Einav / PM can advise on pursuing a workaround for z-stream. (Such a small issue ... doesn't seem like z-stream material to me.)