[Ironruby-core] Code Review: Thread.critical

Curt Hagenlocher curth at microsoft.com
Mon Jan 5 16:32:56 EST 2009


The implementation of ThreadOps._CriticalMonitor and ThreadOps._IsInCriticalRegion as class-level fields will cause state information to leak between ScriptRuntimes.  These should probably be put on the RubyContext, which can either be done directly or by using        RubyContext.GetOrCreateLibraryData to create it dynamically.

Looks good otherwise.

-----Original Message-----
From: Shri Borde
Sent: Monday, January 05, 2009 10:18 AM
To: IronRuby External Code Reviewers; DLR Code Reviews
Cc: ironruby-core at rubyforge.org
Subject: Code Review: Thread.critical


  tfpt review "/shelveset:critical;REDMOND\sborde"

Microsoft.Scripting.dll:

  Fix to interpreter to get better stack trace even when call-site caching kicks in. The try-catch needs to be added to all code that can throw an exception in the interpreter (or preferably in one function like Interpreter.Interpret), but I have not done that here.

IronRuby:

  Implements Thread.critical= by using Monitor.Enter/Exit. This will work for 90% or more of scenarios IMO. MRI does not schedule other threads while Thread.critical is true (while IronRuby will), but this is more an implementation detail that falls out with the green threads scheduler, than something that should be by spec. In fact, other threads do get scheduled anyway if the critical thread does Thread.pass etc (added tests to show this behavior if you want to play with it).
  Added tests
  Added irtests.bat to run all IronRuby tests with a single command, in four processes




More information about the Ironruby-core mailing list