feat(hierarchy): an element reports whether it masks what is typed into it

ios reads it off SecureTextField, which the companion already sent and
nothing read; web reads input[type=password]. Android cannot: the native
tree mapper drops the password attribute before the sidecar sees it, so
the fact is three-valued and null there rather than a false that would
read as "not secure".
This commit is contained in:
pj committed 2026-08-18 17:16:55 +05:30
1 parent 5f1f50c2fb
commit b1e95739ad
11 files changed
+186 -5

No files matched your search

+4
View File
@@ -746,6 +746,10 @@ func (d *Driver) Hierarchy(ctx context.Context) (string, error) {
// emitted is the flag the field declares and not the property's value.
checked: el.checked === true || null,
selected: el.selected === true || null,
// Emitted as a plain boolean, never null, on every editable field: a
// consumer deciding what a typed value may be recorded as has to tell
// "not a secure entry" apart from "nobody said", and android says nothing.
secure: isEditable ? el.type === 'password' : null,
// Emitted as a plain boolean, never null: internal/hierarchy falls back to
// the native heuristic when the field is absent, which reads any class
// name containing "EditText" as an Android text widget. On web that is a
@@ -191,6 +191,43 @@ func TestElementState_ReportsTheOtherDocumentedBooleans(t *testing.T) {
)
}
// Every editable field states `secure`, false included. internal/verifier
// redacts a typed value whenever the target does not positively report "not a
// secure entry", so a dump that omitted the key on an ordinary text field would
// redact the whole recent-action memory on web.
func TestElementState_SecureStatesEveryEditableField(t *testing.T) {
server := httptest.NewServer(http.FileServer(http.Dir("testdata")))
defer server.Close()
d := New()
defer d.Terminate(context.Background())
ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
defer cancel()
if err := d.Launch(ctx, server.URL+"/element-state.html", false, nil); err != nil {
t.Fatalf("Launch: %v", err)
}
for _, testCase := range []struct {
selector string
want bool
}{
{"id:secret", true},
{"id:editing", false},
} {
element := elementInHierarchyDump(ctx, t, d, testCase.selector)
if !element.SecureReported() {
t.Errorf("the dump states no secure fact for %s", testCase.selector)
}
if element.Secure != testCase.want {
t.Errorf("%s secure = %v, want %v", testCase.selector, element.Secure, testCase.want)
}
}
if button := elementInHierarchyDump(ctx, t, d, "id:save"); button.SecureReported() {
t.Error("a button is not a text entry and states nothing")
}
}
// Focus belongs to the node the user is typing into, not to the element the
// shadow tree is mounted on.
//
+1
View File
@@ -11,6 +11,7 @@
<input id="toggle-all" type="checkbox" />
<input id="toggle-done" type="checkbox" checked />
<input id="editing" type="text" value="buy milk" />
<input id="secret" type="password" />
<button id="save">save</button>
<button id="cancel" disabled>cancel</button>
<select id="filter">
+8 -1
View File
@@ -54,6 +54,7 @@ type treeNode struct {
Clickable *bool `json:"clickable,omitempty"`
Enabled *bool `json:"enabled,omitempty"`
Editable *bool `json:"editable,omitempty"`
Secure *bool `json:"secure,omitempty"`
}
// MapHierarchy converts a flat describe-all dump from the simulator companion
@@ -228,6 +229,11 @@ func mapElement(element *rawElement, scrollable bool) (treeNode, bool) {
if editable {
yes := true
node.Editable = &yes
// Stated on every editable field, false included: a consumer deciding
// what a typed value may be recorded as has to tell "not a secure
// entry" apart from "nobody said".
secure := element.Type == "SecureTextField"
node.Secure = &secure
}
if element.Type == "Button" {
yes := true
@@ -238,7 +244,8 @@ func mapElement(element *rawElement, scrollable bool) (treeNode, bool) {
}
func isEditable(elementType string) bool {
return elementType == "TextArea" || elementType == "TextField"
return elementType == "TextArea" || elementType == "TextField" ||
elementType == "SecureTextField"
}
// labelIsDisplayedText reports whether an element type's AXLabel is the string
@@ -388,3 +388,41 @@ func TestScrollableIgnoresAContainerOffTheScreen(t *testing.T) {
t.Fatalf("scrollable containers = %+v, want none off the screen", got)
}
}
// A secure text entry is what XCUITest reports a password field as
// (companion/Sources/ElementTypeName.swift). It has to reach the tree as a
// fact, because that is what lets a typed value be redacted from the record
// without redacting every other field's.
func TestSecureFactPerEditableType(t *testing.T) {
cases := []struct {
elementType string
wantReported bool
wantSecure bool
}{
{"SecureTextField", true, true},
{"TextField", true, false},
{"TextArea", true, false},
{"Button", false, false},
{"StaticText", false, false},
}
for _, testCase := range cases {
dump := `[{"type":"` + testCase.elementType + `","frame":{"x":0,"y":0,"width":10,"height":10},"enabled":true}]`
element := parseSingle(t, dump)
if element.SecureReported() != testCase.wantReported {
t.Errorf("%s reported secure = %v, want %v",
testCase.elementType, element.SecureReported(), testCase.wantReported)
}
if element.Secure != testCase.wantSecure {
t.Errorf("%s secure = %v, want %v", testCase.elementType, element.Secure, testCase.wantSecure)
}
}
}
// A secure text entry is a field a run must be able to type into, or the login
// screens every real app opens on are unreachable.
func TestSecureTextFieldIsEditable(t *testing.T) {
dump := `[{"type":"SecureTextField","frame":{"x":0,"y":0,"width":10,"height":10},"enabled":true}]`
if element := parseSingle(t, dump); !element.Editable {
t.Error("a secure text entry must be editable")
}
}
+19 -1
View File
@@ -72,10 +72,20 @@ type Element struct {
Focused bool `json:"focused,omitempty"`
Selected bool `json:"selected,omitempty"`
Editable bool `json:"editable,omitempty"`
Secure bool `json:"secure,omitempty"`
Bounds Bounds `json:"bounds"`
Attributes map[string]string `json:"attrs,omitempty"`
}
// SecureReported reports whether the producer stated this element's secure
// fact at all. Android never does, so an element without it is unknown rather
// than known not to be a secure entry, and a caller deciding what may be
// written down has to tell those two apart.
func (e *Element) SecureReported() bool {
_, reported := e.Attributes["secure"]
return reported
}
// Node is one node in the hierarchy tree.
type Node struct {
Element
@@ -194,6 +204,7 @@ type treeNodeJSON struct {
Checked flagJSON `json:"checked"`
Selected flagJSON `json:"selected"`
Editable flagJSON `json:"editable"`
Secure flagJSON `json:"secure"`
}
// flagJSON is one boolean field of a node. A value that is not a boolean
@@ -223,7 +234,7 @@ func (f *flagJSON) UnmarshalJSON(data []byte) error {
func (n *treeNodeJSON) unreadableFlags() int {
count := 0
for _, flag := range []flagJSON{n.Clickable, n.Enabled, n.Focused, n.Checked, n.Selected, n.Editable} {
for _, flag := range []flagJSON{n.Clickable, n.Enabled, n.Focused, n.Checked, n.Selected, n.Editable, n.Secure} {
if flag.unreadable {
count++
}
@@ -305,6 +316,7 @@ var selectorKeys = []string{
"placeholderValue",
"resource-id",
"scrollable",
"secure",
"selected",
"tag",
"testID",
@@ -583,6 +595,9 @@ func elementFromNode(node *treeNodeJSON) *Element {
if node.Selected.set {
element.Selected = node.Selected.value
}
if node.Secure.set {
element.Secure = node.Secure.value
}
if node.Editable.set {
element.Editable = node.Editable.value
} else {
@@ -613,6 +628,9 @@ func elementFromNode(node *treeNodeJSON) *Element {
if node.Selected.set {
element.Attributes["selected"] = strconv.FormatBool(node.Selected.value)
}
if node.Secure.set {
element.Attributes["secure"] = strconv.FormatBool(node.Secure.value)
}
element.Attributes["editable"] = strconv.FormatBool(element.Editable)
return element
+47
View File
@@ -235,6 +235,53 @@ func TestBoolFieldsFromNode(t *testing.T) {
}
}
// secure is the one state flag with three answers: a producer that reports
// nothing leaves the element unknown rather than known-not-secure, and a
// consumer deciding what a typed value may be written into a record reads the
// difference.
func TestSecureIsUnknownUntilAProducerReportsIt(t *testing.T) {
cases := []struct {
name string
node string
wantReported bool
wantSecure bool
}{
{"reported secure", `{"attributes": {"bounds": "[0,0,10,10]"}, "secure": true}`, true, true},
{"reported not secure", `{"attributes": {"bounds": "[0,0,10,10]"}, "secure": false}`, true, false},
{"never reported", `{"attributes": {"bounds": "[0,0,10,10]"}}`, false, false},
}
for _, testCase := range cases {
t.Run(testCase.name, func(t *testing.T) {
tree, err := Parse(testCase.node)
if err != nil {
t.Fatalf("Parse: %v", err)
}
element := tree.Elements[0]
if element.SecureReported() != testCase.wantReported {
t.Errorf("SecureReported = %v, want %v", element.SecureReported(), testCase.wantReported)
}
if element.Secure != testCase.wantSecure {
t.Errorf("Secure = %v, want %v", element.Secure, testCase.wantSecure)
}
})
}
}
// A selector reaches the fact by the same route every other boolean state does.
func TestSecureIsSelectable(t *testing.T) {
tree, err := Parse(`{"attributes": {"resource-id": "root", "bounds": "[0,0,10,10]"}, "children": [
{"attributes": {"resource-id": "pwd", "bounds": "[0,0,10,5]"}, "secure": true, "children": []},
{"attributes": {"resource-id": "email", "bounds": "[0,5,10,10]"}, "secure": false, "children": []}
]}`)
if err != nil {
t.Fatalf("Parse: %v", err)
}
element := tree.Find("secure:true")
if element == nil || element.ResourceID != "pwd" {
t.Errorf("secure:true resolved to %+v, want the pwd element", element)
}
}
func TestEditableDerivation(t *testing.T) {
cases := []struct {
name string
+3
View File
@@ -42,6 +42,7 @@ export interface KnownAttrSelectors {
checked?: boolean;
selected?: boolean;
editable?: boolean;
secure?: boolean;
}
/**
@@ -88,6 +89,8 @@ export interface AccessibilityElement {
focused?: boolean;
selected?: boolean;
editable?: boolean;
/** Field masks what is typed into it; null where the platform does not report it. */
secure?: boolean | null;
bounds?: { left: number; top: number; right: number; bottom: number };
x?: number;
y?: number;
+10 -1
View File
@@ -176,6 +176,7 @@ const SELECTOR_KEYS: readonly string[] = [
"placeholderValue",
"resource-id",
"scrollable",
"secure",
"selected",
"tag",
"testID",
@@ -552,6 +553,8 @@ function elementHandle(
const y = Math.round(rect.top + rect.height / 2);
const ariaLabel = element.getAttribute("aria-label") ?? "";
const text = (element.textContent ?? "").trim().slice(0, 200);
const editable =
element.matches(EDITABLE_SELECTOR) && isEditableElement(element as HTMLElement);
const datasetCopy: Record<string, string> = {};
const dataset = (element as HTMLElement).dataset ?? {};
for (const key of Object.keys(dataset)) {
@@ -579,7 +582,7 @@ function elementHandle(
// a contenteditable container typeable here while collectTargets and the
// hierarchy dump, which both require the element ITSELF to match
// EDITABLE_SELECTOR, called the same span inert.
editable: element.matches(EDITABLE_SELECTOR) && isEditableElement(element as HTMLElement),
editable,
focused: focusedElement === element,
// Checkbox and option state lives in the DOM PROPERTY: the markup attribute
// records only what the page started with, so a handle reading it reports a
@@ -588,6 +591,12 @@ function elementHandle(
// dump the goja host gets.
checked: state.checked === true,
selected: state.selected === true,
// Three-valued, unlike the other state flags: null on anything that is not
// a field, matching the hierarchy dump in internal/driver/chrome/driver.go.
// A consumer deciding what a typed value may be written into a record has
// to tell "not a password field" apart from "nobody said", and Android says
// nothing.
secure: editable ? state.type === "password" : null,
x,
y,
bounds: {
+2 -1
View File
@@ -29,6 +29,7 @@
"placeholderValue",
"resource-id",
"scrollable",
"secure",
"selected",
"tag",
"testID",
@@ -40,5 +41,5 @@
"unknownKeyExample": [
"descripton"
],
"unknownKeyMessage": "selector key \"descripton\" cannot match: no element carries that attribute, and it is not one of the accepted keys: accessibilityIdentifier, accessibilityLabel, accessibilityText, aria-label, ariaLabel, bounds, checked, class, className, clickable, content-desc, contentDescription, data-testid, desc, descPrefix, editable, elementType, enabled, focused, hintText, id, idPrefix, identifier, label, package, placeholder, placeholderValue, resource-id, scrollable, selected, tag, testID, testTag, text, title, value"
"unknownKeyMessage": "selector key \"descripton\" cannot match: no element carries that attribute, and it is not one of the accepted keys: accessibilityIdentifier, accessibilityLabel, accessibilityText, aria-label, ariaLabel, bounds, checked, class, className, clickable, content-desc, contentDescription, data-testid, desc, descPrefix, editable, elementType, enabled, focused, hintText, id, idPrefix, identifier, label, package, placeholder, placeholderValue, resource-id, scrollable, secure, selected, tag, testID, testTag, text, title, value"
}
+17 -1
View File
@@ -628,11 +628,12 @@ function domElement(spec: {
text?: string;
checked?: boolean;
contentEditable?: boolean;
type?: string;
}): unknown {
const attributes = spec.attributes ?? {};
return {
tagName: spec.tag.toUpperCase(),
type: spec.tag === "input" ? "text" : "",
type: spec.type ?? (spec.tag === "input" ? "text" : ""),
isContentEditable: spec.contentEditable ?? false,
checked: spec.checked,
id: attributes.id ?? "",
@@ -719,6 +720,21 @@ test("checked reads the live property, not the markup attribute", () => {
assert.equal(handleOf(cleared).checked, false);
});
// `secure` answers three ways. A consumer deciding what a typed value may be
// written into a record has to tell "not a password field" apart from "no
// platform said", and android says nothing: a field answering false only when
// asked about a password would make every web field look like an android one.
test("secure states the field type either way, and nothing off a field", () => {
const password = domElement({ tag: "input", attributes: { id: "pwd" }, type: "password" });
assert.equal(handleOf(password).secure, true);
const email = domElement({ tag: "input", attributes: { id: "email" }, type: "email" });
assert.equal(handleOf(email).secure, false);
const heading = domElement({ tag: "h1", attributes: { id: "title" } });
assert.equal(handleOf(heading).secure, null);
});
test("attrs carries every other attribute alongside tag and aria-label", () => {
const attrs = attrsOf(
domElement({ tag: "input", attributes: { id: "txn-note", placeholder: "What's this for?" } }),