v3/internal/assetserver/build_dev.go (2)

46-56: String-based error detection is fragile but pragmatic for dev tooling.

This approach works but could break if Go changes error message wording. A more robust alternative would use type assertions:

♻️ Optional: Type-based error detection

import (
	"errors"
	"net"
	"syscall"
)

func isConnectionError(err error) bool {
	if err == nil {
		return false
	}
	var netErr *net.OpError
	if errors.As(err, &netErr) {
		var syscallErr syscall.Errno
		if errors.As(netErr.Err, &syscallErr) {
			switch syscallErr {
			case syscall.ECONNREFUSED, syscall.ECONNRESET, syscall.EPIPE:
				return true
			}
		}
	}
	// Fallback to string matching for platform-specific errors like Windows connectex
	errStr := strings.ToLower(err.Error())
	return strings.Contains(errStr, "connectex")
}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@v3/internal/assetserver/build_dev.go` around lines 46 - 56, Replace fragile
string-only checks in isConnectionError with type-based detection: use errors.As
to check for *net.OpError and then extract a syscall.Errno to match
syscall.ECONNREFUSED, syscall.ECONNRESET, and syscall.EPIPE, and keep a fallback
string check only for platform-specific messages like "connectex"; update
isConnectionError to import and use errors, net, and syscall accordingly while
preserving the existing nil guard and fallback behavior.

---

74-92: Consider documenting the retry parameters and enhancing Transport configuration.

The retry parameters (50 retries, 50ms delay = 2.5s max) are effective but hardcoded. A brief comment explaining the rationale would help future maintainers.

The custom http.Transport only configures DialContext, leaving other settings at zero-values (vs. http.DefaultTransport which has tuned pooling/timeout settings). For a dev-only asset server this is acceptable, but consider copying relevant settings from http.DefaultTransport if connection pooling issues arise.

📝 Suggested documentation

 	proxy.Transport = &retryTransport{
 		base: &http.Transport{
+			// Inherit sensible defaults for connection pooling
+			MaxIdleConns:        100,
+			IdleConnTimeout:     90 * time.Second,
 			DialContext: func(ctx context.Context, network, addr string) (net.Conn, error) {
 				// Force IPv4 for localhost connections to avoid IPv6 issues on Windows
 				if parsedURL.Hostname() == "localhost" || parsedURL.Hostname() == "127.0.0.1" {
 					return dialer.DialContext(ctx, "tcp4", addr)
 				}
 				return dialer.DialContext(ctx, network, addr)
 			},
 		},
-		maxRetries: 50,
-		delay:      50 * time.Millisecond,
+		maxRetries: 50,               // High retry count to handle Vite's concurrent connection bursts
+		delay:      50 * time.Millisecond, // Total max retry window: ~2.5s
 	}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@v3/internal/assetserver/build_dev.go` around lines 74 - 92, Add a short
inline comment above the retryTransport instantiation explaining the choice of
maxRetries=50 and delay=50ms (total ~2.5s) so maintainers know the rationale and
can tune for dev use, and make those values easy to change (e.g., named
constants or configuration) instead of magic numbers; also improve the custom
http.Transport used as retryTransport.base by copying relevant production-tuned
fields from http.DefaultTransport (or cloning http.DefaultTransport and only
overriding DialContext) so connection pooling, IdleConnTimeout,
TLSHandshakeTimeout, MaxIdleConnsPerHost, etc., are preserved while still
forcing "tcp4" for localhost via DialContext; reference retryTransport,
proxy.Transport, DialContext, parsedURL and dialer when locating where to apply
these changes.

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes