Replace magic with odrcore

[andiwand]

WIP

[claude]

Claude finished @andiwand's task —— View job

---

Code Review: Replace magic with odrcore

I've completed my review of this PR that replaces the magic dependency with odrcore functionality. Here's my comprehensive feedback:

Review Checklist:

• Analyze build configuration changes
• Review native code additions (CoreWrapper)
• Examine MetadataLoader refactoring
• Check for proper error handling
• Assess performance implications
• Review security considerations
• Evaluate test coverage

Summary

This PR successfully replaces the legacy "magic" library with odrcore's built-in mimetype detection functionality. The changes are well-structured and maintain backward compatibility.

Code Quality and Best Practices ✅

Positive aspects:

1. Clean migration: The PR cleanly removes the old magic dependency and integrates odrcore's mimetype detection
2. Proper JNI implementation: The native wrapper follows JNI best practices with proper string conversion and error handling
3. Consistent error handling: The native method returns null on failure, which is properly handled in Java
4. Build configuration updates: Appropriate updates to Gradle, Conan, and NDK versions

Areas for improvement:

1. Variable naming issue: In MetadataLoader.java line 110, there's a bug - checking type instead of mimetype:

   if (type == null) {  // Should be: if (mimetype == null)

2. Potential NPE: The CoreWrapper.mimetype() method could throw exceptions that aren't specific enough - consider wrapping in a more specific exception type

Potential Bugs or Issues 🐛

1. Critical bug in MetadataLoader.java (line 110):

   if (type == null) {  // Bug: 'type' is not defined, should be 'mimetype'

   This will cause a compilation error or runtime issue.

2. Magic database path configuration: The libmagic database path is still being set in setGlobalParams even though magic is being replaced. This might be leftover code that should be removed.

Performance Considerations ⚡

1. Positive: Using odrcore's native mimetype detection should be more efficient than loading an external magic database
2. Caching opportunity: Consider caching mimetype results for frequently accessed files to avoid repeated JNI calls
3. Native call overhead: Each mimetype detection involves a JNI call - for bulk operations, consider a batch API

Security Concerns 🔒

1. Path validation: The native mimetypeNative method accepts a file path without validation. Ensure paths are sanitized before passing to native code
2. Exception information leakage: Native exceptions are caught broadly - be careful not to expose sensitive file system information in error messages
3. File access: The mimetype detection requires file read access - ensure proper permissions are checked at the Java layer

Test Coverage 🧪

Missing test coverage for:

1. The new CoreWrapper.mimetype() method needs unit tests
2. Edge cases like invalid file paths, non-existent files, and permission errors
3. Integration tests for the MetadataLoader changes
4. Performance regression tests to ensure mimetype detection doesn't slow down file loading

Additional Recommendations

1. Remove unused code: Clean up the libmagic-related configuration in setGlobalParams if it's no longer needed
2. Documentation: Add JavaDoc comments to the new mimetype method explaining its behavior and exceptions
3. Consider fallback: The current implementation has multiple fallback mechanisms for mimetype detection - document the precedence order
4. Version compatibility: Ensure the odrcore version 5.1.1 is stable and well-tested for production use

Build Configuration Notes

• NDK version updated from 26.3.11579264 to 28.1.13356709
• Gradle wrapper updated to 8.13
• Android compile/target SDK updated to 35
• New theme file added for Android v35 with edge-to-edge enforcement

Overall, this is a solid PR that successfully modernizes the mimetype detection. The main concern is the bug in MetadataLoader.java line 110 that needs to be fixed before merging.

[TomTasche]

Tested, works! Also removed the old dependency now...