to clarify: it's when clicking on the Component listbox, which causes the horrific 
bugzilla javascript to run. they have a lot of subcomponents/versions in their bugzilla, 
so the tripple-nested for loops in the script really bog things down. 
 
using CVS HEAD, konqueror takes ~12seconds of wall time to complete the script. a 
recent version of mozilla takes ~10seconds of wall time (during which period it's 
unrespsonsive as well). 
 
in CVS, however, konqueror provides the option to suspend the offending javascript 
after a few seconds via a dialog box. 
 
personally, i'd consider this more a problem with the javascript on the page, and even 
be inclined to close it since Konqueror now provides a way to abort the script. 

I tried to open the website in the Internet Explorer and clicked on the listbox around. 
There weren't any problems while clicking on the listbox.  
I hope this information helps. 

thanks for checking that Frauke...  
 
so our javascript implementation has some rather serious performance issues left to 
be addressed. i'll keep my eye on its progress for this BR... 

I tested it with KDE 3.3.2a and this bug ist still present (although you get the message that asks if you want to abort the execution of the JavaScript). However Mozilla 1.4 also shows this lag on that page...

Also got this on http://bugzilla.ubuntu.com/query.cgi?format=advanced when selection Ubuntu as product and this is a bug not a wishlist.

It's a wishlist as long as it's requesting performance improvements.

I disagree -- and I am the person likely to fix it. It's not a wishlist if it's talking about -horrible- performance

*** Bug 42707 has been marked as a duplicate of this bug. ***

Original report is down to a couple seconds or so on current tree. Not too stellar, but IMHO not unacceptable --- and the website forces us to go through 2 layers of caching(!) to make stuff reasonably fast, something that would be trivial to change in the code...

As for cartman's part: testing a patch right now.

Hmm, actually, that may be with my patched tree...

Created attachment 13896 [details]
patch 

here is the patch for the problem Ismail reported. Might make the original
faster, too

I can confirm it fixes my test case. Thanks Maksim ( once again ) !

Salı 13 Aralık 2005 19:24 tarihinde, Maksim Orlovich şunları yazmıştı: 
[bugs.kde.org quoted mail]

I can confirm it fixes my test case. Thanks Maksim ( once again ) !

SVN commit 500223 by orlovich:

Optimize appends and clears to/of selects via explicit APIs. 
BUG:62785


 M  +1 -1      ecma/kjs_html.cpp  
 M  +42 -13    html/html_formimpl.cpp  
 M  +3 -2      html/html_formimpl.h  


--- branches/KDE/3.5/kdelibs/khtml/ecma/kjs_html.cpp #500222:500223
@@ -3336,7 +3336,7 @@
     }
     else // remove elements
       while (diff-- > 0)
-        element.remove(newLen);
+        element.remove(newLen + diff);
 
     return;
   }
--- branches/KDE/3.5/kdelibs/khtml/html/html_formimpl.cpp #500222:500223
@@ -1893,6 +1893,7 @@
     // 0 means invalid (i.e. not set)
     m_size = 0;
     m_minwidth = 0;
+    m_length   = 0;
 }
 
 HTMLSelectElementImpl::~HTMLSelectElementImpl()
@@ -1946,15 +1947,9 @@
 
 long HTMLSelectElementImpl::length() const
 {
-    int len = 0;
-    uint i;
-    QMemArray<HTMLGenericFormElementImpl*> items = listItems();
-    const uint itemsSize = items.size();
-    for (i = 0; i < itemsSize; ++i) {
-        if (items[i]->id() == ID_OPTION)
-            ++len;
-    }
-    return len;
+    if (m_recalcListItems)
+        recalcListItems();
+    return m_length;
 }
 
 void HTMLSelectElementImpl::add( const HTMLElement &element, const HTMLElement &before, int& exceptioncode )
@@ -1962,8 +1957,21 @@
     if(element.isNull() || element.handle()->id() != ID_OPTION)
         return;
 
-    insertBefore(element.handle(), before.handle(), exceptioncode );
-    if (!exceptioncode)
+    HTMLOptionElementImpl* option = static_cast<HTMLOptionElementImpl*>(element.handle());;
+    //Fast path for appending an item. Can't be done if it is selected and 
+    //we're single-select, since we may need to drop an implicitly-selected item
+    bool fastAppendLast = false;
+    if (before.handle() == 0 && (m_multiple || !option->selected()) && !m_recalcListItems)
+        fastAppendLast = true;
+
+    insertBefore(option, before.handle(), exceptioncode );
+
+    if (fastAppendLast) {
+        m_listItems.resize(m_listItems.size() + 1);
+        m_listItems[m_listItems.size() - 1] = option;
+        ++m_length;
+        m_recalcListItems = false;
+    } else if (!exceptioncode)
         setRecalcListItems();
 }
 
@@ -1976,8 +1984,22 @@
     if(listIndex < 0 || index >= int(items.size()))
         return; // ### what should we do ? remove the last item?
 
+    //Fast path for last element, for e.g. clearing the box
+    //Note that if this is a single-select, we may have to recompute
+    //anyway if the item was selected, since we may want to set 
+    //a different one
+    bool fastRemoveLast = false;
+    if ((listIndex == items.size() - 1) && !m_recalcListItems && 
+        (m_multiple || !static_cast<HTMLOptionElementImpl*>(items[listIndex])->selected()))
+            fastRemoveLast = true;
+
     removeChild(items[listIndex], exceptioncode);
-    if( !exceptioncode )
+
+    if (fastRemoveLast) {
+        m_listItems.resize(m_listItems.size() - 1);
+        --m_length;
+        m_recalcListItems = false;
+    } else if( !exceptioncode)
         setRecalcListItems();
 }
 
@@ -2220,6 +2242,11 @@
     if (optionIndex < 0 || optionIndex >= itemsSize)
         return -1;
 
+    //See if we're asked for the very last item, and check whether it's an <option>
+    //to fastpath clear
+    if (optionIndex == (m_length - 1) && items[itemsSize - 1]->id() == ID_OPTION)
+        return itemsSize - 1;
+
     int listIndex = 0;
     int optionIndex2 = 0;
     for (;
@@ -2247,11 +2274,12 @@
     return optionIndex;
 }
 
-void HTMLSelectElementImpl::recalcListItems()
+void HTMLSelectElementImpl::recalcListItems() const
 {
     NodeImpl* current = firstChild();
     m_listItems.resize(0);
     HTMLOptionElementImpl* foundSelected = 0;
+    m_length = 0;
     while(current) {
         if (current->id() == ID_OPTGROUP && current->firstChild()) {
             // ### what if optgroup contains just comments? don't want one of no options in it...
@@ -2260,6 +2288,7 @@
             current = current->firstChild();
         }
         if (current->id() == ID_OPTION) {
+            ++m_length;
             m_listItems.resize(m_listItems.size()+1);
             m_listItems[m_listItems.size()-1] = static_cast<HTMLGenericFormElementImpl*>(current);
             if (!foundSelected && !m_multiple && m_size <= 1) {
--- branches/KDE/3.5/kdelibs/khtml/html/html_formimpl.h #500222:500223
@@ -431,14 +431,15 @@
     void notifyOptionSelected(HTMLOptionElementImpl *selectedOption, bool selected);
 
 private:
-    void recalcListItems();
+    void recalcListItems() const;
 
 protected:
     mutable QMemArray<HTMLGenericFormElementImpl*> m_listItems;
     short m_minwidth;
     signed short m_size : 15;
     bool m_multiple : 1;
-    bool m_recalcListItems;
+    mutable bool m_recalcListItems :1;
+    mutable unsigned int   m_length:31;
 };
 
 // -------------------------------------------------------------------------