Refactor Result Classes with Unified Data Container and Improved Extensibility

[zewei-wang]

Because the old PR: #941 is accidentally closed and cannot be reopened. I created this new PR with new nits fixed commit.

Description:

This pull request introduces a major refactor to the Result class and its associated data classes. The key improvements and changes are:

1. UML Design: Inheritance → Composition

• Moved from inheritance-based to composition-based design, eliminating overlapping functionality.
• Merged duplicated classes (BuildResult and RunResult in builder_runner.py and results.py) to streamline data containers.
  UML Diagram Link

2. Class Renaming for Clarity

• Renamed BuildResult, RunResult, and AnalysisResult to:
  • BuildInfo
  • RunInfo
  • AnalysisInfo

3. Migration to New Result Class

• Replaced the old Result classes:
  • Updated 8–10 files, including:
    • results.py
    • fuzz_target_error.py
    • builder_runner.py
    • run_one_experiment.py
    • semantic_analyzer.py
    • pipeline.py
    • evaluator.py
    • execution_stage.py
    • prototyper.py
    • enhancer.py
  • Approximately 700–800 lines of code modified.
• Migration:
  • Replaced all old class references to the new Result structure in one comprehensive PR, ensuring consistency and avoiding migration conflicts.

4. New Class Field Semantic Definition

• Specify semantic definitions in classes, remove unused fields:
  • succeeded → compiles in BuildInfo
  • succeeded → crashes in RunInfo
• Introduced read-only methods get_build_script_source() and get_fuzz_target_source() in Result for better access.

5. Crash and Coverage Adaptations

• Incorporated changes from related PRs:
  • Crash Analyzer PR
  • Coverage Analysis PR
• Added new fields to Result:
  • true_bug and insight from CrashResult
  • run_error, crash_func, crash_symptom, crash_stacks from SemanticCheckResult
• Updated CoverageAnalysis to include:
  • improve_required
  • insight
  • suggestions
• Added success property to AnalysisInfo.

6. Error Type Consolidation

• Grouped error types from SemanticCheckResult into an enumeration: FuzzTargetResult.
• Grouping logic:
  • NORM_ – Normal Situations (e.g., NORM_NO_SEMANTIC_ERR)
  • FP_ – False Positives (e.g., FP_TARGET_CRASH)
  • COV_ – Coverage Issues (e.g., COV_NO_INCREASE)
  • NON_SEC_CRASH_ – Not Security-Related Crashes (e.g., NON_SEC_CRASH_NULL_DEREF, NON_SEC_CRASH_SIGNAL)
  • GEN_ – General Problems (e.g., GEN_OVERWRITE_CONST)
• Moved error_type to AnalysisInfo for better alignment with analysis stages.

7. Backward Compatibility and New API Methods

• Introduced methods like is_build_successful() to preserve backward compatibility for current business logic.
• Ensured that each stage generates and propagates its own Result instance with relevant BuildInfo, RunInfo, or AnalysisInfo.

8. Code Quality and Refactor Considerations

• Addressed legacy issues such as inappropriate inheritance and template data classes.
• Improved modularity for result classes by ensuring that the Result class can be safely consumed across multiple stages without boundary errors.

Great thanks to @DonggeLiu for input on refining the UML Design and background context clariying and @oliverchang for reviewing the first version commits.

[DonggeLiu]

Thanks @zewei-wang!
The new design looks pretty good to me in general.

Would you prefer to merge this and start replacing the current results.py with the new one in another PR, or do both in one PR?

[zewei-wang]

Thanks @zewei-wang! The new design looks pretty good to me in general.

Would you prefer to merge this and start replacing the current results.py with the new one in another PR, or do both in one PR?

Thank you @DonggeLiu . Because the files are additional, I prefer to merge it first, then I will migrate the logic and create another PR for migration tasks.