Treat io.ErrUnexpectedEOF as driver.ErrBadConn in handleError

henhouse · arp242 · commit 6d77ced41719 · 2026-04-01T13:04:09.000+01:00
When recvMessage does an io.ReadFull on a partially-received message body and the connection drops mid-read, the result is io.ErrUnexpectedEOF. handleError classifies io.EOF as driver.ErrBadConn but not io.ErrUnexpectedEOF, so cn.err is never set, IsValid() returns true, and database/sql keeps recycling the broken connection. The inProgress flag stays stuck at true (ReadyForQuery never arrived), and the CAS guard rejects every subsequent query with "there is already a query being processed on this connection" — permanently poisoning the pool. So treat io.ErrUnexpectedEOF the same as io.EOF in handleError: both indicate a dead connection. Fixes #1298
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,3 +1,12 @@
+unreleased
+----------
+
+- Treat io.ErrUnexpectedEOF as driver.ErrBadConn so database/sql discards the
+  connection. Since v1.12.0 this could result in permanently broken connections,
+  especially with CockroachDB which frequently sends partial messages ([#1299]).
+
+[#1299]: https://github.com/lib/pq/pull/1299
+
 v1.12.1 (2026-03-30)
 --------------------
 
diff --git a/conn_test.go b/conn_test.go
@@ -16,6 +16,7 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -507,7 +508,7 @@ func TestErrorDuringStartupClosesConn(t *testing.T) {
 
 func TestBadConn(t *testing.T) {
 	t.Parallel()
-	for _, tt := range []error{io.EOF, &Error{Severity: pqerror.SeverityFatal}} {
+	for _, tt := range []error{io.EOF, &Error{Severity: pqerror.SeverityFatal}, io.ErrUnexpectedEOF} {
 		t.Run(fmt.Sprintf("%s", tt), func(t *testing.T) {
 			var cn conn
 			err := cn.handleError(tt)
@@ -521,6 +522,63 @@ func TestBadConn(t *testing.T) {
 	}
 }
 
+func TestUnexpectedEOF(t *testing.T) {
+	t.Parallel()
+
+	// On the first "select truncate" it sends a correct RowDescription followed
+	// by a truncated DataRow (header declares 96 body bytes, only 5 are sent)
+	// and then close the connection. database/sql should discard the connection
+	// and retry, and subsequent queries succeed.
+	var failed atomic.Bool
+	f := pqtest.NewFake(t, func(f pqtest.Fake, cn net.Conn) {
+		f.Startup(cn, nil)
+		for {
+			code, q, ok := f.ReadMsg(cn)
+			if !ok {
+				return
+			}
+			switch code {
+			case proto.Terminate:
+				cn.Close()
+				return
+			case proto.Query:
+				switch q := string(q[:bytes.IndexByte(q, 0)]); {
+				case q == ";": // Ping()
+					f.WriteMsg(cn, proto.EmptyQueryResponse, "")
+					f.WriteMsg(cn, proto.ReadyForQuery, "I")
+				case q == "select truncate" && !failed.Swap(true):
+					f.WriteMsg(cn, proto.RowDescription, "\x00\x01truncate\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x19\xff\xff\xff\xff\xff\xff\x00\x00")
+					cn.Write([]byte("D\x00\x00\x00\x64short"))
+					cn.Close()
+					return
+				case q == "select truncate":
+					f.SimpleQuery(cn, "SELECT", "truncate", "1")
+					f.WriteMsg(cn, proto.ReadyForQuery, "I")
+				case q == "select okay":
+					f.SimpleQuery(cn, "SELECT", "okay", "1")
+					f.WriteMsg(cn, proto.ReadyForQuery, "I")
+				default:
+					panic(fmt.Sprintf("unexpected query: %q", q))
+				}
+			}
+		}
+	})
+	defer f.Close()
+
+	db := pqtest.MustDB(t, f.DSN())
+	db.SetMaxOpenConns(1)
+	db.SetMaxIdleConns(1)
+
+	// This should work as database/sql retries for us.
+	pqtest.QueryRow[int](t, db, `select truncate`)
+	if !failed.Load() {
+		t.Fatal("select truncate never failed")
+	}
+
+	// Make sure it doesn't break the connection.
+	pqtest.QueryRow[int](t, db, `select okay`)
+}
+
 func TestConnClose(t *testing.T) {
 	// Ensure the underlying connection can be closed with Close after an error.
 	t.Run("CloseBadConn", func(t *testing.T) {
diff --git a/error.go b/error.go
@@ -307,7 +307,7 @@ func (cn *conn) handleError(reported error, query ...string) error {
 			reported = driver.ErrBadConn
 		}
 	case error:
-		if err == io.EOF || err.Error() == "remote error: handshake failure" {
+		if err == io.EOF || err == io.ErrUnexpectedEOF || err.Error() == "remote error: handshake failure" {
 			reported = driver.ErrBadConn
 		}
 	default:

Original file line number	Diff line number	Diff line change
`@@ -307,7 +307,7 @@ func (cn *conn) handleError(reported error, query ...string) error {`
`307`	`307`	`reported = driver.ErrBadConn`
`308`	`308`	`}`
`309`	`309`	`case error:`
`310`		`- if err == io.EOF \|\| err.Error() == "remote error: handshake failure" {`
	`310`	`+ if err == io.EOF \|\| err == io.ErrUnexpectedEOF \|\| err.Error() == "remote error: handshake failure" {`
`311`	`311`	`reported = driver.ErrBadConn`
`312`	`312`	`}`
`313`	`313`	`default:`