Bug: Temp Directory Leak When Clone Fails in agentic_change.py

[Serhan-Asad]

Summary

When pdd fix fails to clone a repository (or crashes mid-execution), the temporary directory created by _setup_repository() is not cleaned up, causing a disk space leak.

Location

File: pdd/pdd/agentic_change.py
Function: _setup_repository() (lines 111-136)
Caller: run_agentic_change() (lines 222-245)

Problem

The code creates a temporary directory using tempfile.mkdtemp() but doesn't clean it up when:

1. Clone fails (network error, auth error, etc.)
2. Orchestrator crashes mid-execution
3. Any exception occurs before cleanup

Current Code Flow

```python

Line 112: Create temp directory

temp_dir = Path(tempfile.mkdtemp(prefix=f"pdd_{repo}_"))

Lines 123-134: Try to clone

try:
result = subprocess.run(...)
if result.returncode != 0:
raise RuntimeError(...) # ❌ Exits without cleanup!
except Exception as e:
raise RuntimeError(...) # ❌ Exits without cleanup!

return temp_dir
```

Missing Cleanup

Unlike agentic_test.py which has cleanup in a finally block (lines 253-262), agentic_change.py has no cleanup code.

Impact

Disk Space Leak

• Empty folder leak: If clone fails → empty directory left behind (~0 bytes)
• Full repo leak: If orchestrator crashes → entire cloned repository left behind (could be MBs)
• Partial work leak: If crashes mid-execution → repo + partial changes left behind (could be MBs)

Real-World Scenarios

1. CI/CD with network issues: Issue fetched via API, then network drops during clone
2. Rate limiting: API works but git operations get rate-limited
3. Private repos: Issue is public but repo requires auth
4. Disk full: Clone starts but fails mid-way
5. Orchestrator crashes: Any exception during execution leaves temp folder

Evidence

Test Results

```bash

Before test

Temp folders: 0

Run: pdd fix facebook/react#1

(with test line forcing failure)

After test

Temp folders: 1
Leaked folder: /var/folders/.../T/pdd_react_04d40tvw
Status: Empty directory (clone failed)
```

Code Comparison

agentic_test.py (HAS cleanup):
```python
try:
return run_agentic_test_orchestrator(...)
except Exception as e:
return False, ...
finally:
# Cleanup if we created a temp directory
if repo_path != current_cwd and repo_path.exists():
shutil.rmtree(repo_path) # ✅ CLEANS UP!
```

agentic_change.py (MISSING cleanup):
```python
try:
work_dir = _setup_repository(...)
except RuntimeError as e:
return False, str(e), ... # ❌ Exits without cleanup!

NO FINALLY BLOCK!

return run_agentic_change_orchestrator(...)
```

Recommended Fix

Add cleanup similar to agentic_test.py:

```python

6. Setup Repository (Clone or Use Current)

current_cwd = Path.cwd()
work_dir = None
try:
work_dir = _setup_repository(owner, repo, quiet)
except RuntimeError as e:
return False, str(e), 0.0, "", []

7. Run Orchestrator

try:
if not quiet:
console.print(f"[bold green]Starting Agentic Change Orchestrator...[/bold green]")

return run_agentic_change_orchestrator(
    issue_url=issue_url,
    issue_content=issue_content,
    repo_owner=owner,
    repo_name=repo,
    issue_number=issue_number,
    issue_author=author,
    issue_title=title,
    cwd=work_dir,
    verbose=verbose,
    quiet=quiet,
    timeout_adder=timeout_adder,
    use_github_state=use_github_state
)

finally:
# Cleanup if we created a temp directory
if work_dir and work_dir != current_cwd and work_dir.exists():
if not quiet:
console.print(f"[dim]Cleaning up temporary repository at {work_dir}...[/dim]")
try:
shutil.rmtree(work_dir)
except Exception as e:
if verbose:
console.print(f"[yellow]Warning: Failed to cleanup temp dir: {e}[/yellow]")
```

Severity

Medium-High

• Blocks disk space over time
• Worse for CI/CD environments (many runs = many leaks)
• Can accumulate hundreds of MBs if orchestrator crashes after cloning large repos

Related Files

• pdd/pdd/agentic_test.py - Has proper cleanup (reference implementation)
• pdd/pdd/agentic_change.py - Missing cleanup (bug location)