Re: Interest in a JIT on Linux

1998-11-03 Thread Context Grey

A better JIT for Linux is definitely needed.

TYA may well become such, but it has a long ways to go-
I've timed it on several applications and gotten
between 0 and 30% speedup -- not very impressive
yet.



*strange* rmi/jar related problem

1998-12-17 Thread Context Grey

Hi, 

I have a strange problem using rmi that is somehow related
to .jar files.  This is 1.1.7a on Redhat intel 5.1.

Here's what seems to happen:
- remove all .class & .jar files
- make (compiles and then produces a .jar file)
- start rmiregistry
- run the server, get this error:
  java.rmi.ServerException: Server RemoteException; nested exception is: 
java.rmi.UnmarshalException: error unmarshalling arguments;
nested excep
  tion is: 
java.lang.ClassNotFoundException: 
- jar -xvf the .jar file
- run the server, works this time.  So it seems that the
  CLASSPATH does not mention the .jar file?  wrong:
- jar the .class files into a jar again
- delete the .class files
- run the server, it works...running from a jar this time.

Here's an annotated screen capture of this:
40  4:43killall java# kill previous rmiregistry & server
41  4:43psg java# ps aux|grep.  yes, no java
42  4:43rmiregistry &
43  4:43xfilesserver.sh # fails
44  4:44jar -xvf XF.jar
45  4:44rm XF.jar
46  4:44xfilesserver.sh # works
47  4:44jar -cvf XF.jar *.class   # has *the same* class files as
before!
48  4:44rm *.class
49  4:44newh
50  4:44xfilesserver.sh # works
51  4:45rm -rf META-INF/
52  4:45xfilesserver.sh # works



Re: Java Shell

1999-01-04 Thread Context Grey

I think this is a good idea-

I've wanted to write little Unix-style utilities in java, but the
startup overhead is quite large.  As such a script written in
tcl or something else "feels" much faster, even though in fact
Java is considerably faster (some things like tcl are *thousands*
of times slower than C, rather than being merely 10x).

Some once told me this is the sort of thing you could do in Lisp &
Scheme-
i.e. create a new namespace, run a program in it, then discard the
space and garbage collect.

In java, one has some control over the class loader.  Does anyone know
if it is possible to write a "custom sandbox" to run a program in & then
get rid of this space?



how to remotely launch java (nt, linux) ? (off-topic...)

1999-03-07 Thread Context Grey

Hello,

this is definitely off topic, but I'm hoping someone here
might have some relevant experience.  
  (I'm not sure where the best place to post this is, 
or how to best phrase it even...the Java-linux readers
are intelligent and speak language that I understand at least)

I'd like to use a java RMI "daemon" as a process launcher for
a farm of Linux and NT machines.  In other words, I somehow run a 
java master process on each machine and can then, using RMI, 
tell this process to run commands on that box.  

Under Linux I can simply "rsh" from one master machine to start the
Daemon
on each machine.

My question: is there any way to start a java program on NT
either remotely, or automatically (e.g. is there an autoexec.bat
script that can be used to start it??)

More background: - there are a lot of machines, logging on to
each one to start the process will be tedious.
- My background is Java & some Unix, I don't know NT at all
and am using Java to mostly isolate the application from the
platform difference.  (The RMI process server thing
seems to work OK under NT once I get it launched...)


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



[Fwd: swingset/tableview does not work in 1.1.6v2]

1998-08-10 Thread Context Grey

 


1.1.6v2 glibc doesn't work very well.  Running the swingset application
demo -

-  It sometimes hangs on startup
-  Select the Tableview subdemo, try typing in a 'favorite sport' cell:
   each character except the first is inserted twice; return is inserted 
   as 'o' (rather than ending the editing as it is supposed to).

1.1.5v5-980311 works ok.

This is on redhat5.0 with upgraded glibc & ld.