Spec compliance report

Details

l1a Implementation

• Status: ⚠️ PARTIAL
• Files:
  • core/src/core/org/jnode/vm/x86/compiler/l1a/X86BytecodeVisitor.java:1564 (visit_dstore)
  • core/src/core/org/jnode/vm/x86/compiler/l1a/X86BytecodeVisitor.java:455 (dwstore)
  • core/src/core/org/jnode/vm/x86/compiler/l1a/X86StackFrame.java:351 (getWideEbpOffset)
  • core/src/core/org/jnode/vm/x86/compiler/l1a/VirtualStack.java:135 (pop(int type))
• Issues:
  1. 32-bit mode (PASS): Correctly handles dstore_3 (opcode 0x4A). BytecodeParser (line 413) dispatches to visit_dstore(3). dwstore computes offset via getWideEbpOffset(3) → getEbpOffset(4) (slot 4, the high slot of the double). In 32-bit mode (slotSize=4), stores LSB to disp+LSB (slot 4, lower address) and MSB to disp+MSB (slot 3, higher address) — matching little-endian layout with downward-growing stack. Type validated by vstack.pop(JvmType.DOUBLE).
  2. 64-bit mode (FAIL): getWideEbpOffset still returns getEbpOffset(index+1) (offset for slot 4). The 64-bit path (os.writeMOV(BITS64, helper.BP, disp, reg)) writes 8 bytes at slot 4's location only, covering slots 4–5 instead of the required slots 3–4. The commented code at X86StackFrame.java:353-357 shows the intended fix: use getEbpOffset(index) for 64-bit mode.

l1b Implementation

• Status: ⚠️ PARTIAL
• Files:
  • core/src/core/org/jnode/vm/x86/compiler/l1b/X86BytecodeVisitor.java:1806 (visit_dstore)
  • core/src/core/org/jnode/vm/x86/compiler/l1b/X86BytecodeVisitor.java:461 (dwstore)
  • core/src/core/org/jnode/vm/x86/compiler/l1b/X86StackFrame.java:343 (getWideEbpOffset)
  • core/src/core/org/jnode/vm/x86/compiler/l1b/VirtualStack.java:135 (pop(int type))
• Issues: Identical to l1a. 32-bit correct; 64-bit has same offset bug for category-2 locals.

l2 Implementation

• Status: ❌ FAIL
• Files:
  • core/src/core/org/jnode/vm/compiler/ir/IRGenerator.java:408 (visit_dstore)
  • core/src/core/org/jnode/vm/x86/compiler/l2/GenericX86CodeGenerator.java:278 (generateCodeFor(VariableRefAssignQuad))
  • core/src/core/org/jnode/vm/x86/compiler/l2/L2ByteCodeSupportChecker.java (no visit_dstore override → supported by default)
• Issues:
  1. IR Generation: IRGenerator.visit_dstore correctly decrements stackOffset by 2 (category-2) and marks both variables[index] and variables[stackOffset] as Operand.DOUBLE. However, it creates only a single VariableRefAssignQuad from stackOffset to index, referencing only the first of the two stack slots.
  2. Code Generation: GenericX86CodeGenerator.generateCodeFor(VariableRefAssignQuad) handles LongConstant correctly (emits two 32-bit MOV at disp1 - SLOTSIZE and disp1, lines 320-325). For non-constant values (REGISTER, STACK, TOPS addressing modes), it emits only a single 32-bit MOV/PUSH/POP (lines 329-340), truncating the double to 32 bits. No type-based branching exists for Operand.DOUBLE/Operand.LONG.
  3. Support Checker: L2ByteCodeSupportChecker does not override visit_dstore, so it inherits the no-op implementation from BytecodeVisitorSupport (line 284), marking dstore as supported. The L2 compiler will attempt to compile methods containing dstore_3 but produce incorrect native code for non-constant double values.

JVM Spec References:

• dstore_3 format: opcode 74 (0x4A), no operands, implicit index 3
• Operand stack: ..., value → ... (pops one category-2 double value)
• Local variables: indices 3 and 4 both set (category-2 occupies two consecutive slots)
• Value set conversion applied before store

Closes #252

opencode session  |  github run