diff options
Diffstat (limited to 'testing/web-platform/tests/fenced-frame/resources/navigate-by-name-reporting-helper.html')
| -rw-r--r-- | testing/web-platform/tests/fenced-frame/resources/navigate-by-name-reporting-helper.html | 34 |
1 files changed, 34 insertions, 0 deletions
diff --git a/testing/web-platform/tests/fenced-frame/resources/navigate-by-name-reporting-helper.html b/testing/web-platform/tests/fenced-frame/resources/navigate-by-name-reporting-helper.html new file mode 100644 index 0000000000..d3bd955697 --- /dev/null +++ b/testing/web-platform/tests/fenced-frame/resources/navigate-by-name-reporting-helper.html @@ -0,0 +1,34 @@ +<!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> |