Issues on Windows

[Antonio Cangiano]

I noticed a couple of issues on Windows 7 64 bit:

• IronRuby's actual time and memory results are not reported in the YAML file;
• Timeout is not working properly. When JRuby hangs on one test (e.g., fannkuch) the suite is stuck on that benchmark for hours.

Has anyone else noticed these issues on Windows?

--
Antonio Cangiano
http://antoniocangiano.com

[Roger Pack]

> IronRuby's actual time and memory results are not reported in the YAML file;

The time is not reported? I haven't seen that. Memory is probably not
reported because they don't have a working Win32OLE or what not.

> Timeout is not working properly. When JRuby hangs on one test (e.g.,
> fannkuch) the suite is stuck on that benchmark for hours.

Are you using jruby as the driver ruby?

[Antonio Cangiano]

On Fri, Jun 25, 2010 at 4:03 PM, Roger Pack <roger...@gmail.com> wrote:

The time is not reported? I haven't seen that.

Yeah. For the record I'm using ir64.

Memory is probably not
reported because they don't have a working Win32OLE or what not.

That's ok.

 

Are you using jruby as the driver ruby?

Nope. MRI 1.8.7.

-- 

Antonio Cangiano
http://antoniocangiano.com

[Jim Deville]

Hmm… ir64 and ir should both still work… As for WIN32OLE, we have a working one, it’s just not complete. Let me know what you need implemented and I’ll work on it.

 

JD

--
The GitHub project is located at http://github.com/acangiano/ruby-benchmark-suite
 
You received this message because you are subscribed to the Google
Groups "Ruby Benchmark Suite" group.
To post to this group, send email to
ruby-bench...@googlegroups.com
To unsubscribe from this group, send email to
ruby-benchmark-s...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/ruby-benchmark-suite?hl=en

[Roger Pack]

>> The time is not reported? I haven't seen that.
>
> Yeah. For the record I'm using ir64.

I'm able to get time results
http://gist.github.com/453417
What does yours look like?

>> Are you using jruby as the driver ruby?
>
> Nope. MRI 1.8.7.

re: jruby hanging

It appears that my windows time-outer (timeout2.rb) doesn't actually
work, as $? isn't set until after the process exits.

I could get around it by either requiring the use of jruby as the
runner on windows (which has IO.popen4), or by requiring the
installation and use of the popen4 gem on the host ruby.
Any objections to either of those, or preferences?

-r

[Antonio Cangiano]

On Fri, Jun 25, 2010 at 5:30 PM, Roger Pack <roger...@gmail.com> wrote:

I'm able to get time results

http://gist.github.com/453417
What does yours look like?

http://gist.github.com/453488

I could get around it by either requiring the use of jruby as the

runner on windows (which has IO.popen4), or by requiring the
installation and use of the popen4 gem on the host ruby.
Any objections to either of those, or preferences?

Let's go with the popen4 gem.

Thanks,

Antonio

-- 

Antonio Cangiano
http://antoniocangiano.com

[Roger Pack]

>> I could get around it by either requiring the use of jruby as the
>>
>> runner on windows (which has IO.popen4), or by requiring the
>> installation and use of the popen4 gem on the host ruby.
>> Any objections to either of those, or preferences?
>
> Let's go with the popen4 gem.

Good idea, since it appears that jruby's IO.popen4,though it "should"
work, currently doesn't yet for windows.
http://jira.codehaus.org/browse/JRUBY-4907

So the gem's our only option it appears.
-r

[Antonio Cangiano]

On Fri, Jun 25, 2010 at 5:06 PM, Jim Deville <jdev...@microsoft.com> wrote:

Hmm… ir64 and ir should both still work…

They do. I didn't provide the full path to the executable. That caused the issue.

[rogerdpack]

>    - Timeout is not working properly. When JRuby hangs on one test (e.g.,

>    fannkuch) the suite is stuck on that benchmark for hours.

Ok I did notice that if I run jruby through a batch file, the kill
part doesn't work correctly so the background process keeps running.
So the work around for that is to pass the full path to the java exe,
not to a batch file for now. It would take some WMI magic to
determine the true child process, which probably isn't going to
happen.

I haven't seen any hanging though, and the fannkuch benchmark seems to
terminate normally. Maybe you can add a -rtracer to the timeout2.rb
script to see where it hangs?

-r

[Antonio Cangiano]

On Sat, Jun 26, 2010 at 11:47 PM, rogerdpack <rogerp...@gmail.com> wrote:

Ok I did notice that if I run jruby through a batch file, the kill
part doesn't work correctly so the background process keeps running.
So the work around for that is to pass the full path to the java exe,
not to a batch file for now.

I figured as much. I collected all the data and I will publish the shootout on Monday.

-- 

Antonio Cangiano
http://antoniocangiano.com