Code Review: Thread.critical

tfpt review “/shelveset:critical;REDMOND\sborde”


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.


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

I’d rather see the functionality in irtest.bat in IronRuby’s run.bat.
That way it runs with run 0.

The specs will probably need to be cleaned up to meet RubySpec style
guidelines, and to get Brian to approve them. Go ahead and commit them,
and I can clean them up, or we can work on them together later.

For future reference, style guidelines are here:


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.

Good catch. I have fixed the issue. Also updated the tests according to
Jim’s suggestions. Also, the last test using Thread.kill was wrong but
seemed to succeed since “flunk” fails silently when used on the non-main
thread. New patch included if you want to take a look. Will go ahead and
submit to Snap.