blob: d3bd955697ab8548df6b49f470371b6eff4c4cce (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
|
<!DOCTYPE html>
<script src="utils.js"></script>
<title>Navigate reporting helper</title>
<!-- This document is used as a helper by `../navigate-by-name.html`. That test
attempts to navigate various frames all named `target_frame`, to this
document. All of these navigations should fail, due to the frames being
unreachable to the initiator (because of the "fence" of the fenced frame).
As a result, this document should always load in a new top-level
"outermost" pop-up window.
-->
<script>
const [navigation_success_key] = parseKeylist();
// We're currently using `window.opener` as a proxy for "did this load in a new
// outermost popup window?". Note that if we try and test navigations initiated
// from inside a fenced frame and they open up in a new outermost popup, there
// will be no opener by default (crbug.com/1250694) so using `window.opener` as
// a signal will be insufficient. In order to test anchor navigations to this
// document from within a fenced frame, we'll need a better signal for
// outermost-ness. We should consider adding a value to `document.loadingMode`
// for fenced frames:
// https://wicg.github.io/nav-speculation/prerendering.htmlprerendering.html#browsing-context-loading-mode.
//
// Alternatively if we really want to detect if this loaded inside a fenced
// frame, we could just remove the opt-in headers and then implementations that
// support fenced frame opt-ins would timeout if they somehow don't honor the
// fence on named frame navigations, but that's not a very good outcome.
if (window.opener) {
writeValueToServer(navigation_success_key, "PASS");
} else {
writeValueToServer(navigation_success_key, "FAIL");
}
</script>
|
