JDK-4832371 : Illegal Load Factor 0.75 exception in some hardware
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 1.4.1
  • Priority: P4
  • Status: Closed
  • Resolution: Cannot Reproduce
  • OS: windows_98
  • CPU: x86
  • Submitted: 2003-03-14
  • Updated: 2006-09-08
  • Resolved: 2006-09-08
Related Reports
Relates :  
Description
This is a new bug to track the issue in 4760131, which the customer has recently
indicated is still present as of 1.4.2. There is only one report of such a
problem and it isn't clear whether we have Cyrix machines available to test on,
so this bug is being assigned a relatively low priority.

Comments
EVALUATION A submitter on the SDN forum left the following test case which on their system reports an error. public class Test implements Runnable{ int n; public Test( int i ){ n=i; } public static void main( String args[] ){ for( int i = 0; i < 100; i++ ){ Thread t = new Thread( new Test( i ) ); t.start(); } } public void run(){ long i=0; while( true ){ Test2 test = new Test2(); i++; if( i %10000000 == 0 ) System.out.print( '.' ); } } } public class Test2 { float lf; public Test2(){ this( 11, 0.75f ); } public Test2( int x, float f ){ if( f <= 0 || Float.isNaN( f ) ) System.out.println( f + " < 0 " ); this.lf = f; } } The submitter reports that the println is triggered on their system running Linux/XEN on a 2 GHz HyperThreaded P4. It isn't clear whether hyperthreading is enabled at the BIOS/chip/OS level. I've run this test case in house on a similar CPU (Intel P4 Xeon, 3.2 GHz, CPU family 15, stepping 5) on Red Hat Enterprise Linux AS 4 Update 2 with no problems. I can only guess that this is a bug probably related to context switching, either in the hardware, in Linux, or in Xen, probably related to incorrect saving/restoring of FPU state. Since there is no contact information for the submitter I'm closing this bug as "not reproducible". If the submitter wants to work with us to try to isolate the bug further on their system, please either post contact information, post on the java.sun.com JVM forum, or open a new bug.
2006-09-08

EVALUATION Not a showstopper for 1.4.2. Will investigate for 1.5. ###@###.### 2003-03-13
2003-03-13