From: SAKURAI Masashi <m.sakurai@...>
Date: 2005-11-08T23:48:37+09:00
Subject: Re: yajb usage queries...

Hi,

W.B.Hill wrote:

> OK - How can I pass a value *from* java to the ruby implementation of 
> the abstract class??? If the commented out code is activated it gives:
>   jbridge.RemoteRuntimeException: class=TypeError  message=can't convert
>   Fixnum into String


This error is simply thrown by ruby, not yajb.
I could run your code, appending ".to_s", as follows:

 >  puts("---> Param given: " + arg.to_s)

> Also, this code seems to suffer from a race. *Sometimes* rtest is 
> called before jtest has finished! Possibly related, one run can finish 
> fine, but the next run gives:
>
> Full thread dump Java HotSpot(TM) Client VM (1.4.2_09-b05 mixed mode):
> "Thread-16" prio=1 tid=0x081c4520 nid=0x7cb8 waiting on condition 
> [4d063000..4d0638b8]
>         at java.lang.Thread.sleep(Native Method)
>         at jbridge.BridgeServer$1.run(BridgeServer.java:248)
>         at java.lang.Thread.run(Unknown Source)
>
> How can I syncronize properly (I'm guessing this is an issue of the 
> ruby end of the bridge quitting before the java's done, but I'd prefer 
> not to have to do a wrapper on the java end of things.)

Your guess is correct. When ruby finishes, the ruby side of yajb sends a 
shutdown message to the java side. Then the java side calls System.exit 
after the 0.5 second sleep due to waiting for the completion of shutdown 
message. I think the crush dump was caused by TERM signal during 
sleeping. I will reconsider the finish scheduling and improve in the 
next release.

> BTW, really like yajb for the integration, the way there's only jars 
> and ruby, no JNI to mess around with!


Yes. I think that yajb is useful for the application deployment under 
the environment that has no compiler, such as Windows. OTOH, because the 
performance is also important, I 'm considering writing the JNI driver.

Thanks.
--
SAKURAI Masashi (family given)
m.sakurai@dream.com