Ok...RTFM got me again: http://msdn.microsoft.com/library/en-us/cpref/html/frlrfSystemObjectClas sFinalizeTopic.asp
The problem we were having was that the # of finalizers during teardown of our process was going up. In that case (as the docs say), the finalization gives up (since it may never finish if the finalization has recursion or something else nasty). Here is an example, obviously something you'd never really do, that demonstrated the behavior[1]. Thanks, Shawn Wildermuth [EMAIL PROTECTED] http://adoguy.com http://adoguy.com/book http://shawnwildermuth.com [1] using System; using System.IO; namespace TestMissingFinalizers { class Class1 { static public int iterations = 1000; [STAThread] static void Main(string[] args) { int iter = iterations; for (int x = 0; x < iter; ++x) { // Create a new Instance // to be collected as soon // as possible new Foo(x); } } } public class Foo { public Foo(int number) { _number = number; } int _number; ~Foo() { // The tolerance is pretty high // so it takes quite a lot of // new finalizers for it to // bail out. for (int x = 0; x < 10000; ++x) { new Bar(); } // Write out to see our results // Write out the number of collections // that happen Console.WriteLine("Number Uncollected: {0}", --Class1.iterations); } } public class Bar { ~Bar() { } int x = 0; } } You can read messages from the Advanced DOTNET archive, unsubscribe from Advanced DOTNET, or subscribe to other DevelopMentor lists at http://discuss.develop.com.
