Test some more edge cases for X-Frame-Options #24618
Conversation
wpt-pr-bot
requested review from
annevk,
foolip,
jdm,
jgraham,
mikewest,
zcorpan and
zqzhang
|
See also #21730. |
There was a problem hiding this comment.
This looks really great. One thing that is somewhat hard to review is that
Header: Value,Value2
and
Header: Value
Header: Value2
are expected to have identical results.
Could we perhaps generate the former from the data for the latter? And if the problem is that this would result in too many tests we could use variants I suppose.
Edit: Mike's remark about embed (which also goes for object) is worth noting in the README at least as a TODO.
| } | ||
| case "no message": { | ||
| waitForMessageFrom(i, t).then(t.unreached_func("Frame should not have sent a message.")); | ||
| i.onload = t.step_func_done(() => { |
There was a problem hiding this comment.
I suspect a problem here is that Firefox does not dispatch the load event for (all) network errors. See also whatwg/html#125 (and references to that in other tests). If initially i.contentDocument was available as I suspect it would be given initial about:blank, you could wait for it to not be available using step_wait and friends.
|
I've made it generate comma variants from multi-header variants, and also removed the need to specify both orders. (If headerValue and headerValue2 are different, then it will also generate tests for what happens when you swap them.) The end result is that multiple.html contains 42 tests; that doesn't seem so bad. I'll leave embed/object and any Firefox workarounds for a future PR. |
|
Should we merge this ahead of the spec PR, since it increases coverage? Or should we wait until we settle on what to do for conflicts? |
|
I think ideally we merge this when we merge the specification text and file impl bugs. |
Follows whatwg/html#5737.