@debugmcp/mcp-debugger/docs/archive/refactoring/task-26-container-test-failures-fix-summary.md

Run-time step-through debugging for LLM agents.

# Task 26: Fix Test Failures Revealed by Skip Logic Fix - Summary

## Overview
Task 25 successfully fixed the inverted skip logic, causing tests to run by default. This revealed 3 failing tests that were previously hidden. All failures were related to container path handling in Docker environments.

## Test Status Before Fix
- **805 tests passing**
- **3 tests failing** (all container-related)
- **1 test still skipped**

## Test Status After Fix
- **808 tests passing** (+3)
- **0 tests failing** 
- **0 tests skipped** (removed the obsolete Python discovery test)

## Root Causes Identified

### 1. Container Working Directory Mismatch
The Dockerfile sets `WORKDIR /app`, but test volumes are mounted at `/workspace`. This caused path resolution failures.

### 2. Proxy Worker Path Validation
The proxy worker was checking if script paths exist from the container's working directory (`/app`) instead of the mounted volume location (`/workspace`).

### 3. DAP Launch Request Path Issues
The `sendLaunchRequest` was using relative paths and incorrect working directories in container mode.

### 4. Python Debug Target Working Directory
The debug target launcher wasn't using the correct working directory in container mode.

## Fixes Applied

### 1. Process Launcher Implementation (`src/implementations/process-launcher-impl.ts`)
```typescript
// Fixed working directory for container mode
const cwd = process.env.MCP_CONTAINER === 'true' 
  ? '/workspace'  // Use mounted volume directory
  : path.dirname(scriptPath);
```

### 2. Proxy Worker Path Validation (`src/proxy/dap-proxy-worker.ts`)
```typescript
// Check paths relative to /workspace in container mode
let scriptToCheck = payload.scriptPath;
if (process.env.MCP_CONTAINER === 'true' && !path.isAbsolute(payload.scriptPath)) {
  scriptToCheck = path.join('/workspace', payload.scriptPath);
}
```

### 3. DAP Connection Manager (`src/proxy/dap-proxy-connection-manager.ts`)
```typescript
// Convert relative paths to absolute and use correct working directory
if (process.env.MCP_CONTAINER === 'true') {
  if (!path.isAbsolute(scriptPath)) {
    programPath = path.join('/workspace', scriptPath);
  }
  workingDir = '/workspace';
}
```

## Container Path Resolution Strategy

The fix establishes a consistent path resolution strategy for container mode:

1. **Working Directory**: Always use `/workspace` as the working directory in containers
2. **Path Validation**: Check file existence relative to `/workspace` for relative paths
3. **Absolute Paths**: Convert relative paths to absolute paths anchored at `/workspace`
4. **Volume Mounts**: Ensure all volume mounts use `/workspace` as the container mount point

## Test Results

### Fixed Tests
1. **Container fibonacci.py test**: Now correctly resolves `examples/python/fibonacci.py` to `/workspace/examples/python/fibonacci.py`
2. **Container absolute path rejection test**: Now properly rejects absolute host paths while accepting relative paths
3. **SSE server temp directory test**: Now passes (was previously passing, counted in the 808)

### Remaining Work
- None - all test issues have been resolved
- Deleted the obsolete Python discovery test that was previously skipped

## Lessons Learned

1. **Container Path Consistency**: It's critical to maintain consistency between Dockerfile WORKDIR and volume mount points
2. **Path Resolution Logic**: Container mode requires special handling for path resolution at multiple layers
3. **Test Skip Logic**: Incorrect skip logic can hide failing tests, making it important to regularly audit skip conditions
4. **Relative vs Absolute Paths**: Container environments need clear rules about when to use relative vs absolute paths

## Impact

This fix ensures that the MCP debugger works correctly in containerized environments, which is essential for:
- Docker-based development workflows
- CI/CD pipelines using containers
- Cross-platform debugging scenarios
- Isolated debugging environments

The fixes make the container path handling more robust and predictable, reducing the likelihood of path-related issues in production container deployments.