fix: add encoding='utf-8' to subprocess.run() calls to fix Windows non-UTF-8 locale

[suainam]

Fix Windows Encoding Issue in subprocess.run() Calls

Problem

The code-review-graph build command fails on Windows systems with non-UTF-8 default encoding (e.g., Chinese Windows with GBK, Japanese Windows with Shift-JIS).

Error Message

UnicodeDecodeError: 'gbk' codec can't decode byte 0x80 in position 3921: illegal multibyte sequence
AttributeError: 'NoneType' object has no attribute 'splitlines'

Root Cause

In incremental.py, all subprocess.run() calls use text=True without specifying encoding='utf-8':

result = subprocess.run(
    cmd,
    capture_output=True,
    text=True,  # ❌ Uses system default encoding (GBK on Chinese Windows)
    cwd=str(repo_root),
    timeout=_GIT_TIMEOUT,
)

When text=True is used without an explicit encoding:

1. Python uses the system's default encoding (e.g., GBK on Chinese Windows)
2. Git outputs UTF-8 encoded content
3. Python tries to decode UTF-8 bytes as GBK → UnicodeDecodeError
4. The exception is caught, but result.stdout becomes None
5. Code crashes when trying to call .splitlines() on None

Affected Code Locations

All subprocess.run() calls in incremental.py:

• Line ~356: get_all_tracked_files() - git ls-files
• Line ~420: get_changed_files() - git diff
• Line ~450: get_file_content_at_commit() - git show
• And other git command invocations

Impact

• Affected platforms: Windows with non-UTF-8 system encoding (Chinese, Japanese, Korean, etc.)
• Affected operations: All git operations during graph building and updating
• Severity: Critical - completely blocks graph building on affected systems

Solution

Add encoding='utf-8' parameter to all subprocess.run() calls:

result = subprocess.run(
    cmd,
    capture_output=True,
    text=True,
    encoding='utf-8',  # ✅ Explicitly use UTF-8
    cwd=str(repo_root),
    timeout=_GIT_TIMEOUT,
)

Proposed Changes

File: code_review_graph/incremental.py

Replace all occurrences of:

text=True,

With:

text=True, encoding='utf-8',

This ensures consistent UTF-8 decoding across all platforms.

Testing

Before Fix

# On Chinese Windows (GBK)
$ python -m code_review_graph build
UnicodeDecodeError: 'gbk' codec can't decode byte 0x80...

After Fix

# On Chinese Windows (GBK)
$ python -m code_review_graph build
Building code-review-graph...
✓ Parsed 437 files
✓ Created 14,888 nodes
✓ Created 86,584 edges
[OK] Graph built successfully!

Workaround (Current)

Users on affected systems can use this workaround script:

#!/usr/bin/env python3
"""Temporary workaround for Windows encoding issue."""
import shutil
from pathlib import Path

def patch_incremental_py(file_path: Path):
    backup_path = file_path.with_suffix('.py.backup')
    shutil.copy2(file_path, backup_path)
    
    content = file_path.read_text(encoding='utf-8')
    patched = content.replace('text=True,', "text=True, encoding='utf-8',")
    file_path.write_text(patched, encoding='utf-8')
    
    return backup_path

# Patch, build, restore...

Additional Notes

• This is a backward-compatible change - UTF-8 works on all platforms
• Python's subprocess documentation recommends explicit encoding for cross-platform compatibility
• Similar issues may exist in other files that call subprocess.run() with text=True

References

• Python subprocess documentation: https://docs.python.org/3/library/subprocess.html#subprocess.run
• Related issue: Windows users unable to build graph on non-UTF-8 systems
• Tested on: Windows 10 Pro (Chinese), Python 3.10+

---

Environment:

• OS: Windows 10 Pro (Chinese locale, GBK encoding)
• Python: 3.10+
• code-review-graph: 2.3.2
• Git: 2.x (outputs UTF-8)

Reproduction:

1. Install code-review-graph on Chinese Windows
2. Run code-review-graph build in any git repository
3. Observe UnicodeDecodeError

Expected behavior:
Graph builds successfully regardless of system encoding.

Actual behavior:
Build fails with encoding error on non-UTF-8 Windows systems.