Memory leak in Microsoft Java VM

[Mickey Segal]

The applet and source code at www.segal.org/java/leak2/ illustrate a memory
leak in the Microsoft Java Virtual Machine. The same applet shows no memory
leak using the Sun JVM.
Although the Microsoft VM has been discontinued, it would be helpful to know
how to avoid such leaks since many browsers still use the Microsoft VM.

Does anyone know of good workarounds other than pestering users to switch to
Sun's Java? Most of our users are constrained by their IT departments to
keep to one JVM or another.

 

[Roedy Green]

Does anyone know of good workarounds other than pestering users to switch to
Sun's Java? Most of our users are constrained by their IT departments to
keep to one JVM or another.

the discontinued MS JVM is 10 year old technology. It is utterly
ridiculous it is still holding back Java. That is a century in
computer years.

 

[Chris Smith]

The applet and source code at www.segal.org/java/leak2/ illustrate a memory
leak in the Microsoft Java Virtual Machine.
So?

Does anyone know of good workarounds other than pestering users to switch to
Sun's Java? Most of our users are constrained by their IT departments to
keep to one JVM or another.

If your users are forced to retain a Java VM that has been EOL for years
and for which Microsoft is no longer even maintaining basic security
fixes, then the ONLY sensible decision is for you to stop providing them
with software in Java. Perhaps you could manage a JavaScript client, or
Flash, or even an ActiveX plugin. Absolutely anything would be better
than an EOL Microsoft JVM.

When you talk to their IT department to decide what other technologies
are acceptable, make sure to mention the situation with the Microsoft
JVM, and strongly suggest embarking on an effort to remove that
dangerous piece of code from any systems where it remains. If you can
convince them to install the Sun JVM at the same time or allow users to
do so, then that makes your life easier as well.

--
www.designacourse.com
The Easiest Way To Train Anyone... Anywhere.

Chris Smith - Lead Software Developer/Technical Trainer
MindIQ Corporation

 

[Mickey Segal]

the discontinued MS JVM is 10 year old technology. It is utterly
ridiculous it is still holding back Java. That is a century in
computer years.

Is someone keeping statistics on use of Sun vs. Microsoft JVMs for applets
on the Web? We have not been collecting data systematically, but my guess
from spot determinations is that the usage is close to half-half, with the
Microsoft JVM falling pretty rapidly in the past year but still seen
commonly in the wild.

Even though the last significant update of the Microsoft JVM that I remember
was in 1998, it still remains faster at GUI display than the Sun JVM. One
can see this speed difference in various test applets (to be even-handed I
chose applets prepared for bug reports in three different environments):

1. Microsoft bug: the test applet discussed in this thread
(www.segal.org/java/leak2/; despite the memory leak)

2. Sun bug: the test applet for a Sun HotSpot crash thread
(www.segal.org/java/sun_jit/)

3. Macintosh problem: test applet for slow GUI display on Macintosh, even
slower than Sun (www.segal.org/java/CanvasTable3/).

One striking difference between Microsoft and Sun JVMs is that the Microsoft
JVM does not display intermediate views - it seems to wait until the GUI
operations are complete and then displays the end result. Is there some way
to coax the Sun JVM into doing the same thing? It produces a display with
less jiggling around of components and may explain some of the better GUI
speed of the antique Microsoft JVM.