Initializing code

Tidbit 2 « Bug Vanquisher

Tidbit 2 « Bug Vanquisher — Sun, 12 Oct 2008 17:29:26 GMT

PingBack from http://tanveerbadar.wordpress.com/2008/10/12/tidbit-2/

re: Initializing code

Chris Brumme — Fri, 16 Apr 2004 16:34:00 GMT

Excepting the bug that Jeroen mentioned above, the CLR guards execution of a .cctor so that only one thread can proceed. You do not need to perform any synchronization yourself.

re: Initializing code

Kenny — Fri, 16 Apr 2004 15:54:00 GMT

Great blog! Quick question: do we need to worry about multiple threads entering a cctor or does the CLR 'suspend' all threads accessing a type while it runs the class constructor? This kind of race condition with static initialization is an issue in C++. Is it the same in the CLR?

RE: Initializing code

Chris Brumme — Wed, 16 Apr 2003 17:12:00 GMT

Yes, this is a known bug. Unfortunately, we are too aggressive in wiring up callsites. It's on the list to get fixed.

RE: Initializing code

Jeroen Frijters — Wed, 16 Apr 2003 15:35:00 GMT

Unfortunately, there is a (small) bug in the current implementation. Here is an example that successfully instantiates a type after the .cctor threw an exception: using System; class cctor { static cctor() { new cctor(); throw new Exception("barf"); } } class Test { static void Main(string[] args) { try { new cctor(); } catch(Exception x) { Console.WriteLine(x); } try { new cctor(); Console.WriteLine("Hey, we got an instance!"); } catch(Exception x) { Console.WriteLine(x); } } }