Devart Forums

My company have just upgraded from Corelab 3.5 to Devart 5.0, and all has been going smoothly until one of the websites we run started having troubles under load.

A typical stack trace is like this:

http://vader.aftonbladet.se/

[IndexOutOfRangeException: Index was outside the bounds of the array.]
System.Collections.ArrayList.Add(Object value) +41
Devart.Common.GlobalComponentsCache.AddToGlobalList(IComponent component, String groupName) +426
Devart.Common.GlobalComponentsCache.AddToGlobalList(IComponent component) +7
Devart.Common.DbConnectionBase.set_Owner(Object value) +63
Storm.Wod.Data.DataSted.InitializeComponent() +3282
Storm.Wod.Data.DataSted..ctor() +16
Storm.Wod.BL.Sted.BLStedSimple..ctor() +99
WebAftonbladet.Common.ABPageBase.getData() +80
WebAftonbladet.Common.ABPageGoogle.Page_Init(Object sender, EventArgs e) +24
System.Web.Util.CalliHelper.EventArgFunctionCaller(IntPtr fp, Object o, Object t, EventArgs e) +15
System.Web.Util.CalliEventHandlerDelegateProxy.Callback(Object sender, EventArgs e) +33
System.Web.UI.Control.OnInit(EventArgs e) +99
System.Web.UI.Page.OnInit(EventArgs e) +9
System.Web.UI.Control.InitRecursive(Control namingContainer) +321
System.Web.UI.Page.ProcessRequestMain(Boolean includeStagesBeforeAsyncPoint, Boolean includeStagesAfterAsyncPoint) +834

From what i can see here, it crashes when it tries to set the owner of our database connection, trying to add it to some "globalcomponentscache" i have never heard of.

So, my questions are:

1. What is this GlobalComponentCache?
2. We have a few other (older) data access classes that doesnt set the owner of their objects, could this cause the problem? Or should i stop using "owner" altoghether?
3. Any other suggestions?

This issue is likely to be associated with multi-threading. It will be fixed.
GlobalComponentCache and Owner are needed for implementation of links instantiated between design-time components.
As a temporary workaround, you can remove the Owner property asssignment from your code.

Thank you for your prompt and to-the-point reply.

We will try the workaround.

AndreyR wrote:This issue is likely to be associated with multi-threading. It will be fixed.

Is this bug fixed in the current version?

Yes, it is.