Bug 333829

Description David Williams 2011-01-10 01:04:16 EST

similar to other bugs related to 333678 the subject class produces different byte codes with latest JDT compiler. 

The reason is apparently that the compiler handles the empty switch statement in 

protected void setSortColumn(int column) 

method differently. 

Presumably more efficiently? 

For the source of 

	@Override
	protected void setSortColumn(int column) {
		switch (column) {
		// case 1:
		// sortByName();
		// break;
		// case 2:
		// sortByType();
		// break;
		}
		super.setSortColumn(column);
	}


The 3.6 compiler produced: 

  // Method descriptor #4 (I)V
  // Stack: 2, Locals: 2
  protected void setSortColumn(int column);
    0  aload_0 [this]
    1  iload_1 [column]
    2  invokespecial org.eclipse.wst.xsl.internal.debug.ui.AbstractTableBlock.setSortColumn(int) : void [261]
    5  return
      Line numbers:
        [pc: 0, line: 179]
        [pc: 5, line: 180]
      Local variable table:
        [pc: 0, pc: 6] local: this index: 0 type: org.eclipse.wst.xsl.internal.debug.ui.tabs.main.TransformsBlock
        [pc: 0, pc: 6] local: column index: 1 type: int



The new 3.7 M4 compiler produces

  // Method descriptor #4 (I)V
  // Stack: 2, Locals: 2
  protected void setSortColumn(int column);
    0  iload_1 [column]
    1  pop
    2  aload_0 [this]
    3  iload_1 [column]
    4  invokespecial org.eclipse.wst.xsl.internal.debug.ui.AbstractTableBlock.setSortColumn(int) : void [261]
    7  return
      Line numbers:
        [pc: 0, line: 171]
        [pc: 2, line: 179]
        [pc: 7, line: 180]
      Local variable table:
        [pc: 0, pc: 8] local: this index: 0 type: org.eclipse.wst.xsl.internal.debug.ui.tabs.main.TransformsBlock
        [pc: 0, pc: 8] local: column index: 1 type: int


While I can't "read" byte code instructions, the empty switch statement does look a bit odd, and I'd not be surprised if it couldn't be "compiled out" of byte codes. 

My main concern is that the bits now differ, even though version/qualifier is the same ... so ... I'd suggest simply incrementing service field by +100 re-releasing.