Either use .NET to the max or not. Mixing the two can be quite frustrating. If you do mix them, then this is your best friend:

 

GC.Collect(GC.MaxGeneration);

GR.WaitForPendingFinalizers();

GC.Collect(GC.MaxGeneration);

 

The other good friend is System.Marshal.ReleaseComObject() but constant GC takes away the pain of knowing just how many times you need to call ReleaseComObject (one call is almost never enough).

 

We run the GC code above everytime we are about to shut down the app or a document. Failure to do so can mean that when .NET decides to run GC and the RCWs that get collected call Release(), there may not be any object in memory (kaboom). The alternative is to wait for each object's final release, which may never occur until shutdown since GC runs when MS decides it does. If your documents contain as much data as ours, its totally impractical to wait for GC to recover your memory when the final release call comes.

 

I know many elite programmers still prefer MFC over .NET (I know I do). Still, I sense it might be time to emigrate.

 

I'm involved in a project with 600k lines of 10 years old MFC-code. With GoF fresh in mind it is hard not to eager for a complete re-design. Especially since this is a project with a bright future.

 

Here's my dilemma. What should we do?

1) Rewrite all code in modern C++ with STL/Boost, together with Xtreme ToolkitPro for the GUI.

2) Emigrate to C#/.NET and use Xtreme SuitePro for some fancy GUI.

3) Other ideas?