[ruby-core:27741] Re: [Feature #2619] Proposed method: Process.fork_supported?

On Sat, Jan 23, 2010 at 1:50 PM, Caleb Clausen <[e-mail]> wrote:
> I don't know much about jruby, but I can't see why you say this is so
> difficult. It seems to me that fork_supported? (or whatever it should
> be called) could be implemented like this:

Here's fork_supported? on JRuby:

def Process.fork_supported?; false; end

There's no way to fork a JVM properly; too many signal handlers and
threads that just don't propagate properly. In fact, I'd *prefer*
fork_supported? over the respond_to? behavior change.

> Does that not work right in jruby? In the general case, there are many
> such system methods that need to be checked; you wouldn't want to
> repeat the above for each one. But it can be generalized and dry'd up
> a little:
>
> module Platform
>  SYSCALLNAME2TEST={
>    :fork => "!!fork{}",
>    :ppid => "res=Process.ppid; res and res !=0",
>    #etc
>  }
>
>  eval SYSCALLNAME2TEST.map{|name,test|
>     "
>     def #{name}_supported?
>       return @@#{name}_supported unless @@#{name}_supported==nil
>      @@#{name}_supported= (#{test})
>      return @@#{name}_supported
>    rescue Exception
>      return @@#{name}_supported=false
>    end
>    "
>  }.join("\n")
> end

Except that calling many methods will have side effects we don't want
to happen, or may have overcomplicated tests, or may not be testable
at all and you don't know until you call them with valid inputs if
they'll work. The general case is basically the halting problem...you
can't determine from inspection whether calling a given
method/function will behave well enough to claim it is "implemented".

> Essentially, you're doing the same kind of tests that configure does,
> but at runtime instead of compile time. Maybe you have to do a
> different special test for each system call, but that's one line, or
> at most a few, for each one. So in all we're talking about, what,
> several hundred lines of (ruby) code? I don't see how any of this is
> difficult. Tedious, perhaps, but not that difficult. I wouldn't even
> characterize it as "a very large amount of work".

Great, write it for all the system calls that 1.9 reports respond_to?
as false for :) And make sure it works correctly on all the platforms
where people use JRuby. You do have AIX, zLinux, and OpenVMS systems
available, right?

- Charlie