Stack size

I've tried both Mac OSX (my home machine) and a Solaris server (at school). Strangely, my Mac gets a deeper level of recursion before the Stack overflows (roughly 5500 levels versus only about 4500 on Solaris). Both of these are pretty pathetic (at least according to my algorithms teacher).

Todd

toddobryan92103 at 2007-7-2 1:51:59 >

Apparently that option doesn't actually do anything with HotSpot. I crash with the same number of level of recursion with

java -100M StackOverflowor

java -1b StackOverflow.

The StackOverflow.java file is just:

public final class StackOverflow {

private static int depth;

public static void main(String[] args) {

try {

depth = 0;

method();

} catch (StackOverflowError err) {

System.out.println(err);

}

System.out.println("depth: "+depth);

}

public static void method() {

depth++;

method();

}

}

Other people, using C and C++ are getting similar pathetic results so it may be that the stack size is fairly small. Anybody know of a way to increase the possible stack size on a Unix machine?

If not, I may have to do the project in Perl (kinda makes you shiver, doesn't it).

Thanks all,

Todd

toddobryan92103 at 2007-7-2 1:51:59 >

Todd,

I tested your code on Solaris 8 and JDK 1.3.1 and got it to work

for a large number of method calls by tweaking both the java

and unix stack size limits.

The java stack size depends on the system(shell) defined stack segment size.

1) using the default 8M stack size set (from ulimit)

#ulimit -s

8192

#java StackOverflow

java.lang.StackOverflowError

depth: 4446

# java -Xss8M StackOverflow

java.lang.StackOverflowError

depth: 80065

#java -Xss16M StackOverflow

Stack size of 16384 Kb exceeds current limit of 8192 Kb.

(Stack sizes are rounded up to a multiple of the system page size.)

See limit(1) to increase the stack size limit.

2) Increase the uniz(shell) stack size to 16M.

#ulimit -s 16384

#ulimit -a

....

stack(kbytes)16384

...

#java StackOverflow<

java.lang.StackOverflowError

depth: 4449

#java -Xss16M StackOverflow<

java.lang.StackOverflowError

depth: 160727

As you can see by changing the default stack size (both at the

unix shell level and java arguments), you can achieve the

desired result.

Hope this helps.

Cheers!

Sun-DTS

jdcusr1 at 2007-7-2 1:51:59 >

Note 1:

on Linux, (RH 7.2, JSE 1.4.2_02) the "limit stack size" has no effect unless you make the stack too small to even load the JVM

the -Xss flag does work (up to some limit)

On Windows, NOTHING effectively changes the stack size,

-Xss999 fails to load

-Xss1000 is the same as the default, as is:

-Xss1000000

so the only control is >= 1000

Sun should provide an explicit and well documented way of :

1) using all of the available stack on any architecture

2) limiting the stack size effectively (i.e. incrementally) via a flag

Apperson at 2007-7-2 1:51:59 >

Hey All,

I have been having a problem related to the stack and heap size (see [url=http://forum.java.sun.com/thread.jsp?forum=37&thread=486077&start=0&range=15]my posting[/url] for more details) and I have been following a number of forums relating to the -X options, but I have never successful managed to configure the -X options to run my program.

When I reduced the memory requirements of the C code (by reducing the array sizes, see [url=http://forum.java.sun.com/thread.jsp?forum=37&thread=486077&start=0&range=15#2279755]here[/url]) called by the JNI, the code worked but when I increased the JVM memory, it didn't make any difference. It doesn't add up in my head! Surely it should have worked?

This makes me wonder if the -X option work at all. I am using windows 2000 and java version "1.4.2_01" and I tried a series of combinations of the -X options. None worked.

I would apreciate any comments regarding this,

Thanks,

Stef

stefarg at 2007-7-2 1:51:59 >