fix: align Java instrumentation with comparator schema

[mashraf-222]

Summary

• Fixed critical schema mismatch between Java instrumentation and comparator
• Instrumentation now writes to invocations table that comparator expects
• This enables behavior verification to work correctly for Java optimizations

Problem

The Java behavior verification pipeline was broken due to a schema mismatch:

• Instrumentation (instrumentation.py) wrote data to a test_results table
• Comparator (Comparator.java) read data from an invocations table
• Result: Comparator couldn't find instrumented test data, causing all comparisons to fail

Changes

instrumentation.py (lines 405-420)

• Changed table from test_results to invocations
• Updated schema to match comparator expectations:
  • Old: test_module_path, test_class_name, test_function_name, function_getting_tested, loop_index, iteration_id, runtime, return_value, verification_type
  • New: call_id, method_id, args_json, result_json, error_json, start_time, end_time
• Generate unique call_id from (loop_index * 10000) + iteration_id
• Map method_id as className.functionName
• Store JSON result in result_json column
• Store timing data in start_time and end_time columns

test_instrumentation.py

• Updated test assertions to expect invocations table instead of test_results

Testing

• All instrumentation tests pass (25 passed)
• All comparator tests pass (6 passed)
• Total: 322 of 330 Java tests pass (1 failure is pre-existing fixture issue)

Impact

This fix is critical for Java optimization workflow:

• Enables behavior verification to compare original vs optimized code
• Allows detection of correctness issues in optimizations
• Required for Java optimization features to work end-to-end

🤖 Generated with Claude Code

[mashraf-222]

we want to use the same sqlite schema across all languages. WILL NOT MERGE IT