Route trailing readReadyForQuery error through handleError on extended-protocol cleanup

conn.go

@@ -1760,7 +1760,7 @@ func (cn *conn) readParseResponse() error {
 		return nil
 	case proto.ErrorResponse:
 		err := parseError(r, "")
-		_ = cn.readReadyForQuery()
+		_ = cn.handleError(cn.readReadyForQuery())
 		return err
 	default:
 		cn.err.set(driver.ErrBadConn)
@@ -1788,7 +1788,7 @@ func (cn *conn) readStatementDescribeResponse() (paramTyps []oid.Oid, colNames [
 			return paramTyps, colNames, colTyps, nil
 		case proto.ErrorResponse:
 			err := parseError(r, "")
-			_ = cn.readReadyForQuery()
+			_ = cn.handleError(cn.readReadyForQuery())
 			return nil, nil, nil, err
 		default:
 			cn.err.set(driver.ErrBadConn)
@@ -1809,7 +1809,7 @@ func (cn *conn) readPortalDescribeResponse() (rowsHeader, error) {
 		return rowsHeader{}, nil
 	case proto.ErrorResponse:
 		err := parseError(r, "")
-		_ = cn.readReadyForQuery()
+		_ = cn.handleError(cn.readReadyForQuery())
 		return rowsHeader{}, err
 	default:
 		cn.err.set(driver.ErrBadConn)
@@ -1827,7 +1827,7 @@ func (cn *conn) readBindResponse() error {
 		return nil
 	case proto.ErrorResponse:
 		err := parseError(r, "")
-		_ = cn.readReadyForQuery()
+		_ = cn.handleError(cn.readReadyForQuery())
 		return err
 	default:
 		cn.err.set(driver.ErrBadConn)
@@ -1853,7 +1853,7 @@ func (cn *conn) postExecuteWorkaround() error {
 		switch t {
 		case proto.ErrorResponse:
 			err := parseError(r, "")
-			_ = cn.readReadyForQuery()
+			_ = cn.handleError(cn.readReadyForQuery())
 			return err
 		case proto.CommandComplete, proto.DataRow, proto.EmptyQueryResponse:
 			// the query didn't fail, but we can't process this message

conn_test.go

@@ -7,6 +7,7 @@ import (
 	"database/sql"
 	"database/sql/driver"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"maps"
@@ -578,6 +579,62 @@ func TestUnexpectedEOF(t *testing.T) {
 	pqtest.QueryRow[int](t, db, `select okay`)
 }
 
+// #1320
+func TestUnexpectedClose(t *testing.T) {
+	t.Parallel()
+
+	// Emit non-fatal ErrorResponse and close the connection on any Parse.
+	// Previously this would cause conn.inProgress to remain "stuck" on true
+	// because cn.err was never set and no ReadyForQuery was received.
+	var triggered atomic.Bool
+	f := pqtest.NewFake(t, func(f pqtest.Fake, cn net.Conn) {
+		f.Startup(cn, nil)
+		for {
+			code, _, ok := f.ReadMsg(cn)
+			if !ok {
+				return
+			}
+			switch code {
+			case proto.Terminate:
+				cn.Close()
+				return
+			case proto.Query: // Ping(): empty simple query.
+				f.WriteMsg(cn, proto.EmptyQueryResponse, "")
+				f.WriteMsg(cn, proto.ReadyForQuery, "I")
+			case proto.Parse:
+				triggered.Store(true)
+				f.WriteMsg(cn, proto.ErrorResponse, "SERROR\x00C08P01\x00Mserver conn crashed?\x00\x00")
+				cn.Close()
+				return
+			}
+		}
+	})
+	defer f.Close()
+	db := pqtest.MustDB(t, f.DSN())
+	db.SetMaxOpenConns(1)
+	db.SetMaxIdleConns(1)
+
+	_, err := db.Exec("select $1::int", 1)
+	if err == nil {
+		t.Fatal("first Exec: want non-nil error from pooler fault, got nil")
+	}
+	if !triggered.Load() {
+		t.Fatal("server-side fault was never triggered")
+	}
+
+	// Must not short-circuit to errQueryInProgress and have the poisoned conn
+	// stays in the pool with inProgress=true. database/sql should open a new
+	// connection we get a proper *pq.Error from the new fault, not the "there
+	// is already a query being processed" guard.
+	_, err = db.Exec("select $1::int", 1)
+	if err == nil {
+		t.Fatal("second Exec: want non-nil error, got nil")
+	}
+	if errors.Is(err, errQueryInProgress) {
+		t.Fatalf("second Exec: poisoned conn was reused from pool: %v", err)
+	}
+}
+
 func TestConnClose(t *testing.T) {
 	// Ensure the underlying connection can be closed with Close after an error.
 	t.Run("CloseBadConn", func(t *testing.T) {