fix(upstream): use parent context for shutdown detection in Docker recovery

[Dumbris]

Summary

Fixed a race condition where Docker recovery retry logging occurred after test completion despite having a context check from PR #124. The issue was using the child recoveryCtx instead of parent ctx for shutdown detection.

Root Cause

PR #124 added a context check before logging, but it checked the wrong context:

// PR #124 added this:
select {
case <-recoveryCtx.Done():  // ❌ Wrong context!
    return
default:
}
m.logger.Info("Docker recovery retry", ...)

The problem:

1. ctx = Parent context (cancelled when test/manager shuts down)
2. recoveryCtx = Child context created with context.WithCancel(ctx)
3. Child context cancellation may lag behind parent by nanoseconds
4. Race window: Check passes → Context cancelled → Log executes

On CI machines with scheduling delays, this tiny race window is hit more frequently.

Changes

File: internal/upstream/manager.go

Use parent context for shutdown detection (line 1905):

// Before (PR #124):
select {
case <-recoveryCtx.Done():  // ❌ Child context
    return
default:
}

// After (this PR):
select {
case <-ctx.Done():  // ✅ Parent context
    return
default:
}

Why This Works

• Parent ctx is cancelled immediately when shutdown starts
• No propagation delay to child context
• Eliminates race window between check and log execution
• Test shutdown is instantly visible to the check

Test Results

✅ Tests pass locally with the fix:

go test -v -race -run TestUpdateListenAddressValidation ./internal/runtime/...

Fixes

Main branch failure:

• Unit Tests (macos-latest, 1.23.10)
• https://github.com/smart-mcp-proxy/mcpproxy-go/actions/runs/19047644560

Error:

panic: Log in goroutine after TestUpdateListenAddressValidation has completed: 
2025-11-03T20:02:22.176Z	INFO	Docker recovery retry	
{"attempt": 1, "elapsed": "2.002340625s", "next_retry_in": "5s"}

mcpproxy-go/internal/upstream.(*Manager).handleDockerUnavailable
	internal/upstream/manager.go:1910 +0x940

Related PRs

• PR fix(upstream): prevent logging after context cancellation in Docker recovery retry #124: Added context check (used wrong context)
• PR fix(upstream): use parent context for shutdown detection in Docker recovery #126: Fixes context check to use parent context (this PR)

🤖 Generated with Claude Code

[github-actions]

📦 Build Artifacts for PR #126

Version: v0.9.24-dev-ea68fa7-pr126
Latest tag: v0.9.24

Available Artifacts

• frontend-dist-pr (177 KB)
• archive-linux-amd64 (9.7 MB)
• archive-linux-arm64 (8.8 MB)
• archive-darwin-arm64 (16 MB)
• installer-dmg-darwin-arm64 (18 MB)
• archive-windows-arm64 (16 MB)
• archive-windows-amd64 (18 MB)
• archive-darwin-amd64 (18 MB)
• installer-dmg-darwin-amd64 (20 MB)

How to Download

Option 1: GitHub Web UI (easiest)

1. Go to the workflow run page
2. Scroll to the bottom "Artifacts" section
3. Click on the artifact you want to download

Option 2: GitHub CLI

gh run download 19047803960 --repo smart-mcp-proxy/mcpproxy-go

# Or download a specific artifact:
gh run download 19047803960 --name dmg-darwin-arm64

Note: Artifacts expire in 14 days.