JDK-4287874 : kestrel FCS Build M,Following JCK1.3-RC1 tests failed with java_plugin on win32
  • Type: Bug
  • Component: core-libs
  • Sub-Component: java.lang
  • Affected Version: 1.3.0
  • Priority: P2
  • Status: Closed
  • Resolution: Duplicate
  • OS: solaris_2.6,windows_95,windows_nt
  • CPU: x86,sparc
  • Submitted: 1999-11-04
  • Updated: 2000-02-07
  • Resolved: 2000-02-07
Related Reports
Duplicate :  
Duplicate :  
Duplicate :  
Duplicate :  
Duplicate :  
Duplicate :  
Relates :  
Relates :  
The following JCK tests are failing under untrusted mode with kestrel FCS build M on WIN32 only ,

api\java_lang\Math\desc.html#cos: cos test  
api\java_lang\StrictMath\desc.html#cos: cos test  
api\java_lang\StrictMath\desc.html#sin: sin test  
api\java_lang\Math\desc.html#sin: sin test  

For more info ,Please view the jtr files located at :





jeff.hsiao@eng 1999-11-17
These tests fails on Hotspot VM only. The tests passed if ran with classic VM. 

Due to HotSpot VM only (I'm told it's HotSpot client), I'm moving the bug to runtime.

andrew.bennett@eng 1999-11-17

EVALUATION Java Plug-in doesn't reference any of the math routines, and certainly doesn't modify their use by its clients. thomas.ball@Eng 1999-11-15 Please, see \\sqesvr\vsn\GammaBase\Bugs\4287874. Note, although the sin and cos tests from StrictMath are included they no longer fail. ----------------- This tests actually exhibited a bug in the VM and one in the test: a) the VM bug was a HandleMark problem that just has been fixed for the 1.3 win32 Kestrel HotSpot VM and will be fixed for HotSpot 2.0 Server today (see also my last putback). b) the test bug is that when using Math.sin/cos the VM does *not* guarantee StrictMath precision! The VM guarantees monotonicity, i.e., the interpreter & compiler will compute exactly the same value for the same argument, but they may use a fast hardware implementation for sin/cos depending on the machine. Thus, the error from the *true* value can be bigger than what StrictMath guarantees. This implementation is according to Spec. If one wants guaranteed precision, one needs to use StrictMath. robert.griesemer@Eng 1999-11-24 ###@###.### 2000-01-24 One bug ( a) ) was fixed by the VM team(as I understand). The other one ( b) ) is not a test bug but a jdk bug (4306749). The Math test is checking that the precision is within 1 ulp from the correct result and the test doesn't pass.