Skip to content

All 3 compilers FAIL: l1a/l1b miss returnAddress push/type; L2 explicitly unsupported #561

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
opencode-agent wants to merge 1 commit into
base: master
Choose a base branch
from
Open

All 3 compilers FAIL: l1a/l1b miss returnAddress push/type; L2 explicitly unsupported #561

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
64 changes: 64 additions & 0 deletions jsr_w_compliance_report.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,64 @@
## Spec compliance report

| Compiler | Status | Issues |
|----------|--------|--------|
| l1a | ❌ FAIL | Missing operand stack push of returnAddress; no returnAddress type support; ret/jsr asymmetry not modeled |
| l1b | ❌ FAIL | Missing operand stack push of returnAddress; no returnAddress type support; ret/jsr asymmetry not modeled |
| l2 | ❌ FAIL | Explicitly unsupported in L2ByteCodeSupportChecker and IRGenerator |

---

## Details

### l1a Implementation
### l1a Implementation
- **Status:** ❌ FAIL
- **Files:** `core/src/core/org/jnode/vm/x86/compiler/l1a/X86BytecodeVisitor.java:2960-2962`, plus stack/local helper files
- **Issues:**
1. **Missing operand stack push** (line 2960-2962): `visit_jsr` only emits `os.writeCALL(helper.getInstrLabel(address))`. Per JVM spec (§jsr_w), the address of the *next* instruction (return address) must be pushed onto the operand stack as a `returnAddress` type value. The x86 CALL pushes to hardware stack, not JVM operand stack.
2. **No returnAddress type support**: The JVM `returnAddress` type is not represented in `JvmType`, `Item` hierarchy (`VirtualStack`, `ItemFactory`), or type system. No `ReturnAddressItem` class exists.
3. **ret/jsr asymmetry not modeled** (line 3886-3892): `visit_ret` does `os.writeJMP(helper.BP, ebpOfs)` — jumps to address stored in local variable — but `jsr` never pushed that address to the operand stack for subsequent `astore` by the subroutine.
4. **VirtualStack inconsistency**: The virtual stack is not updated to reflect the pushed return address, breaking stack depth tracking for subsequent bytecodes in the subroutine.
5. **BytecodeParser correct** (core/src/core/org/jnode/vm/bytecode/BytecodeParser.java:890-891): Properly decodes 32-bit signed offset via `gets4()` and computes absolute target `address + gets4()`.

### l1b Implementation
- **Status:** ❌ FAIL
- **Files:** `core/src/core/org/jnode/vm/x86/compiler/l1b/X86BytecodeVisitor.java:3687-3692`, plus stack/local helper files
- **Issues:**
1. **Identical to l1a**: Same `os.writeCALL(helper.getInstrLabel(address))` implementation at line 3691.
2. **Same missing operand stack push** of return address.
3. **Same missing returnAddress type** in `Item` hierarchy and `VirtualStack`.
4. **Same ret/jsr asymmetry** issue at line 4830-4839 (`visit_ret`).
5. **BytecodeParser** correctly handles 32-bit offset (shared parser).

### l2 Implementation
- **Status:** ❌ FAIL
- **Files:**
- `core/src/core/org/jnode/vm/x86/compiler/l2/L2ByteCodeSupportChecker.java:163-165`
- `core/src/core/org/jnode/vm/compiler/ir/IRGenerator.java:1321-1323`
- `core/src/core/org/jnode/vm/compiler/ir/IRBasicBlockFinder.java:216-220`
- **Issues:**
1. **Explicitly unsupported**: `L2ByteCodeSupportChecker.visit_jsr` throws `UnsupportedOperationException` (line 164).
2. **IR generation unsupported**: `IRGenerator.visit_jsr` throws `UnsupportedOperationException` (line 1322).
3. **Basic block analysis incomplete** (line 216-220): `IRBasicBlockFinder.visit_jsr` has a comment acknowledging uncertainty but does not add the branch target as a successor, nor the fall-through (next instruction) as a successor. Per spec, both the subroutine target and the instruction after jsr_w are reachable.
4. **No returnAddress type in IR**: The IR `Operand` type system has no `RETURN_ADDRESS` category.

---

## JVM Spec References
- **jsr_w spec**: https://docs.oracle.com/javase/specs/jvms/se6/html/Instructions2.doc7.html
- **Operand Stack Effect**: `... → ..., address` (pushes `returnAddress`)
- **Branch Offset**: Signed 32-bit (`(branchbyte1 << 24) | (branchbyte2 << 16) | (branchbyte3 << 8) | branchbyte4`)
- **Target**: Must be opcode within same method
- **Notes**: Used with `ret` for `finally`; asymmetry intentional (push to stack, retrieve from local)

## JNode File References
- **Parser**: `core/src/core/org/jnode/vm/bytecode/BytecodeParser.java:890-891` ✅
- **Visitor Contract**: `core/src/core/org/jnode/vm/bytecode/BytecodeVisitor.java:296` ✅
- **L1A jsr**: `core/src/core/org/jnode/vm/x86/compiler/l1a/X86BytecodeVisitor.java:2960-2962` ❌
- **L1A ret**: `core/src/core/org/jnode/vm/x86/compiler/l1a/X86BytecodeVisitor.java:3886-3892` ❌
- **L1B jsr**: `core/src/core/org/jnode/vm/x86/compiler/l1b/X86BytecodeVisitor.java:3687-3692` ❌
- **L1B ret**: `core/src/core/org/jnode/vm/x86/compiler/l1b/X86BytecodeVisitor.java:4830-4839` ❌
- **L2 Support**: `core/src/core/org/jnode/vm/x86/compiler/l2/L2ByteCodeSupportChecker.java:163-165` ❌
- **L2 IR Gen**: `core/src/core/org/jnode/vm/compiler/ir/IRGenerator.java:1321-1323` ❌
- **L2 BB Finder**: `core/src/core/org/jnode/vm/compiler/ir/IRBasicBlockFinder.java:216-220` ⚠️
Loading