Add test coverage for main.rs pipeline orchestration

[codesoda]

Summary

main.rs contains the bulk pipeline orchestration — run_single_test_file, run_all_test_files, run_discovered_file — with critical error handling, multi-file summary, and teardown logic, but these functions have zero test coverage. The existing integration test re-implements the pipeline rather than calling these functions.

Severity: S2 (Medium) | Confidence: 0.90 | Blast radius: Medium — regressions in the most user-facing code path go undetected.

Technical Details

Untested orchestration functions in src/main.rs:

Function	Lines	What it does
run_single_test_file	158-580	Full single-file pipeline: config load → commands → execution → report
run_all_test_files	587-670	Multi-file orchestration with summary aggregation
run_discovered_file	679-780	Discovery-mode wrapper with glob matching

What is currently tested:

• src/main.rs lines 810-946: 10 tests, but they only cover CLI argument parsing via Cli::parse_from
• tests/pipeline_integration.rs: Tests individual phases (command execution, step execution) using a mock provider, but does not call run_single_test_file or run_all_test_files

What's at risk without coverage:

• Error handling and graceful degradation when commands fail mid-pipeline
• Multi-file summary aggregation logic (exit code selection, report merging)
• Best-effort teardown ordering (processes killed in correct order on failure)
• Config validation + merge logic in the orchestration path
• Edge cases: empty test files, all steps skipped, provider init failure

Proposed Fix

1. Refactor for testability: Extract an AgentSessionFactory trait or closure parameter so run_single_test_file can accept a mock provider:

   pub fn run_single_test_file(
       config: &Config,
       test_file: &TestFile,
       session_factory: impl Fn(&Config) -> Box<dyn AgentSession>,
       // ...
   ) -> Result<RunResult, PipelineError>

2. Write integration tests that call run_single_test_file directly:

   • Happy path: all steps pass
   • Command failure: short-lived command exits non-zero
   • Provider failure: session init returns error
   • Partial success: some steps pass, some fail
   • Teardown verification: long-lived processes are cleaned up

3. Write a multi-file test for run_all_test_files:

   • Two test files, one passes, one fails → verify correct exit code and summary

Estimated effort: 1-2 days

Related

• Related to F9: shared test infrastructure needed first
• Part of refactor Bundle 8: Orchestration Test Coverage

---

🔍 Found by vibe-code-audit — automated codebase audit skill for Claude Code.