In the Shadow of Leaves

   1: namespace ConsoleApplication1

   2: {

   3:     class CTest

   4:     { 

   5:         public CTest(int i) {}

   6:     }

   7: }

I think that first confusing thing is the usage of the new operator. It may give you a feeling that something here is different. Take a look at the line 7 of this example. What do you think, is this method is thread safe?

   1: class C

   2: {

   3:     private CTest c;

   4:  

   5:     public void Foo()

   6:     {

   7:         c = new CTest(1);

   8:     }

   9: }

This is what happens behind the stage in method Foo(). Here is the Foo() code in a little details

1 .method public hidebysig instance void Foo() cil managed
2 {
3 .maxstack 8
4 L_0000: nop
5 L_0001: ldarg.0
6 L_0002: ldc.i4.1
7 L_0003: newobj instance void ConsoleApplication1.CTest::.ctor(int32)
8 L_0008: stfld class ConsoleApplication1.CTest ConsoleApplication1.CTest::c
9 L_000d: ret
10 }

It is clear that line 7 with opcode newobj creates and pushes a new instance of CTest into the stack and exactly one line later, the local variable c is assigned with a brand new value. If multiple threads were calling CTest::Foo() the result would be unpredictable.

We need to consider one more subtle issue: constructor parameters. If one of the parameters is a reference type, it means that external thread may change the referenced object and make the constructor invalid.

   1: class Test

   2: {

   3:     public Test(IList items)

   4:     {

   5:         if(items.Count > 0)

   6:         {

   7:             // do something special

   8:         }

   9:     }

  10: }

Constructor is not an exception case then we deal with multi threaded environment. I have no idea why and how this simple issue can confuse anyone but since I witnessed a long discussion regarding this issue, I contribute this post to remove any kind of hesitations. Enjoy.