Symantec JIT 3.00.039 computes VM stack depth wrongly

Name: tb29552 Date: 04/30/98


This applies to the JIT that comes with JDK-1.1.6
running under Windows 95/NT-4.0.

When the JIT is used to run hand-assembled bytecode,
it will occassionally complain about inconsistent stack
depth, such as:

A nonfatal internal JIT (3.00.039(x)) error 'Stack levels inconsistent' has occured in:
  'Foo.main ([Ljava/lang/String;)V': Interpreting method.

In this case, Foo.class has actually been verified correct by both javap and kimera.
A manual trace of the stack also proves the stack to be consistent.

This typically happens when the method has an exception handler that branches backward.

Here's the test code for Foo.j (written in jasmin syntax):
The JIT will not complain if you put the block of code labelled with "continue" after
the "sleepException" handler.

; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.source Foo.j
.class public Foo
.super java/lang/Object


.method public static main([Ljava/lang/String;)V
    .limit stack 5
    .limit locals 2

sleepStart:
    ldc2_w 100
    invokestatic java/lang/Thread/sleep(J)V
sleepEnd:
    aconst_null
    getstatic java/lang/System/out Ljava/io/PrintStream;
    ldc "sleep succeeded"

continue: ; expect stack=[String, System.out, dummy]
    invokevirtual java/io/PrintStream/println(Ljava/lang/String;)V
    return

sleepException:
    ; stack now contains Exception
    getstatic java/lang/System/out Ljava/io/PrintStream;
    ldc "sleep succeeded"
    goto continue
    
    
    .catch java/lang/InterruptedException from sleepStart to sleepEnd using sleepException
.end method


; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


.method public <init>()V
.limit stack 1
.limit locals 1
aload_0
invokespecial java/lang/Object/<init>()V
return
.end method


; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(Review ID: 29066)
======================================================================