Bug 253409 – maxlength test for <textarea> failing when ran in a third-party iframe

• Component: Forms
• Status: NEW

[Ahmad Saleem]

Hi Team,

I came across following two tests from Blink repo / Chromium Source and noted that WebKit ToT (261243@main) are failing some of them:

Test (max length): https://jsfiddle.net/pLahgb8q/show (Firefox Nightly 112 only fails two while WebKit / Safari fails a lot more)

Test (min length): https://jsfiddle.net/ts7ad8zp/show. (Firefox Nightly 112 fails both same test as of Safari / WebKit)

^ Chrome Canary 113 passes all of above tests.

Just wanted to raise so we can track them.

Thanks!

[Chris Dumez]

https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2 failures are not real failures, the exception name matches, only the exception string doesn't. We can choose whatever string/description we want for exceptions.

https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures though. Firefox and Chrome pass all the tests, we don't.

[Chris Dumez]

(In reply to Chris Dumez from comment #1)
> https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> failures are not real failures, the exception name matches, only the
> exception string doesn't. We can choose whatever string/description we want
> for exceptions.
> 
> https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> though. Firefox and Chrome pass all the tests, we don't.

The weird thing though is I thought I had imported https://jsfiddle.net/pLahgb8q/show as layout test as LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all the checks for this test.

[Ahmad Saleem]

(In reply to Chris Dumez from comment #2)
> (In reply to Chris Dumez from comment #1)
> > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > failures are not real failures, the exception name matches, only the
> > exception string doesn't. We can choose whatever string/description we want
> > for exceptions.
> > 
> > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > though. Firefox and Chrome pass all the tests, we don't.
> 
> The weird thing though is I thought I had imported
> https://jsfiddle.net/pLahgb8q/show as layout test as
> LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> the checks for this test.

I took updated copy of test from https://source.chromium.org/.

[Chris Dumez]

(In reply to Ahmad Saleem from comment #3)
> (In reply to Chris Dumez from comment #2)
> > (In reply to Chris Dumez from comment #1)
> > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > failures are not real failures, the exception name matches, only the
> > > exception string doesn't. We can choose whatever string/description we want
> > > for exceptions.
> > > 
> > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > though. Firefox and Chrome pass all the tests, we don't.
> > 
> > The weird thing though is I thought I had imported
> > https://jsfiddle.net/pLahgb8q/show as layout test as
> > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > the checks for this test.
> 
> I took updated copy of test from https://source.chromium.org/.

LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.

[Ahmad Saleem]

(In reply to Chris Dumez from comment #4)
> (In reply to Ahmad Saleem from comment #3)
> > (In reply to Chris Dumez from comment #2)
> > > (In reply to Chris Dumez from comment #1)
> > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > failures are not real failures, the exception name matches, only the
> > > > exception string doesn't. We can choose whatever string/description we want
> > > > for exceptions.
> > > > 
> > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > though. Firefox and Chrome pass all the tests, we don't.
> > > 
> > > The weird thing though is I thought I had imported
> > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > the checks for this test.
> > 
> > I took updated copy of test from https://source.chromium.org/.
> 
> LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.

You are right. I just did 'diffcheck' on both test and except 'Error String' and one comment, there is no difference in both.

Weird - if these tests are failing in JSFiddle but passing via Test Runner.

[Chris Dumez]

(In reply to Ahmad Saleem from comment #5)
> (In reply to Chris Dumez from comment #4)
> > (In reply to Ahmad Saleem from comment #3)
> > > (In reply to Chris Dumez from comment #2)
> > > > (In reply to Chris Dumez from comment #1)
> > > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > > failures are not real failures, the exception name matches, only the
> > > > > exception string doesn't. We can choose whatever string/description we want
> > > > > for exceptions.
> > > > > 
> > > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > > though. Firefox and Chrome pass all the tests, we don't.
> > > > 
> > > > The weird thing though is I thought I had imported
> > > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > > the checks for this test.
> > > 
> > > I took updated copy of test from https://source.chromium.org/.
> > 
> > LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> > web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.
> 
> You are right. I just did 'diffcheck' on both test and except 'Error String'
> and one comment, there is no difference in both.
> 
> Weird - if these tests are failing in JSFiddle but passing via Test Runner.

Indeed! Might be a behavior difference when running in a third-party iframe (since that's how jsFiddle runs the test case).

[Ahmad Saleem]

(In reply to Chris Dumez from comment #6)
> (In reply to Ahmad Saleem from comment #5)
> > (In reply to Chris Dumez from comment #4)
> > > (In reply to Ahmad Saleem from comment #3)
> > > > (In reply to Chris Dumez from comment #2)
> > > > > (In reply to Chris Dumez from comment #1)
> > > > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > > > failures are not real failures, the exception name matches, only the
> > > > > > exception string doesn't. We can choose whatever string/description we want
> > > > > > for exceptions.
> > > > > > 
> > > > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > > > though. Firefox and Chrome pass all the tests, we don't.
> > > > > 
> > > > > The weird thing though is I thought I had imported
> > > > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > > > the checks for this test.
> > > > 
> > > > I took updated copy of test from https://source.chromium.org/.
> > > 
> > > LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> > > web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.
> > 
> > You are right. I just did 'diffcheck' on both test and except 'Error String'
> > and one comment, there is no difference in both.
> > 
> > Weird - if these tests are failing in JSFiddle but passing via Test Runner.
> 
> Indeed! Might be a behavior difference when running in a third-party iframe
> (since that's how jsFiddle runs the test case).

Let me make local HTML and then run them on WebKit ToT local build.

[Ahmad Saleem]

(In reply to Ahmad Saleem from comment #7)
> (In reply to Chris Dumez from comment #6)
> > (In reply to Ahmad Saleem from comment #5)
> > > (In reply to Chris Dumez from comment #4)
> > > > (In reply to Ahmad Saleem from comment #3)
> > > > > (In reply to Chris Dumez from comment #2)
> > > > > > (In reply to Chris Dumez from comment #1)
> > > > > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > > > > failures are not real failures, the exception name matches, only the
> > > > > > > exception string doesn't. We can choose whatever string/description we want
> > > > > > > for exceptions.
> > > > > > > 
> > > > > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > > > > though. Firefox and Chrome pass all the tests, we don't.
> > > > > > 
> > > > > > The weird thing though is I thought I had imported
> > > > > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > > > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > > > > the checks for this test.
> > > > > 
> > > > > I took updated copy of test from https://source.chromium.org/.
> > > > 
> > > > LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> > > > web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.
> > > 
> > > You are right. I just did 'diffcheck' on both test and except 'Error String'
> > > and one comment, there is no difference in both.
> > > 
> > > Weird - if these tests are failing in JSFiddle but passing via Test Runner.
> > 
> > Indeed! Might be a behavior difference when running in a third-party iframe
> > (since that's how jsFiddle runs the test case).
> 
> Let me make local HTML and then run them on WebKit ToT local build.

Yes - it is JSFiddle playing around. Using local copy on WebKit ToT, we pass all tests.

[Chris Dumez]

(In reply to Ahmad Saleem from comment #8)
> (In reply to Ahmad Saleem from comment #7)
> > (In reply to Chris Dumez from comment #6)
> > > (In reply to Ahmad Saleem from comment #5)
> > > > (In reply to Chris Dumez from comment #4)
> > > > > (In reply to Ahmad Saleem from comment #3)
> > > > > > (In reply to Chris Dumez from comment #2)
> > > > > > > (In reply to Chris Dumez from comment #1)
> > > > > > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > > > > > failures are not real failures, the exception name matches, only the
> > > > > > > > exception string doesn't. We can choose whatever string/description we want
> > > > > > > > for exceptions.
> > > > > > > > 
> > > > > > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > > > > > though. Firefox and Chrome pass all the tests, we don't.
> > > > > > > 
> > > > > > > The weird thing though is I thought I had imported
> > > > > > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > > > > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > > > > > the checks for this test.
> > > > > > 
> > > > > > I took updated copy of test from https://source.chromium.org/.
> > > > > 
> > > > > LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> > > > > web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.
> > > > 
> > > > You are right. I just did 'diffcheck' on both test and except 'Error String'
> > > > and one comment, there is no difference in both.
> > > > 
> > > > Weird - if these tests are failing in JSFiddle but passing via Test Runner.
> > > 
> > > Indeed! Might be a behavior difference when running in a third-party iframe
> > > (since that's how jsFiddle runs the test case).
> > 
> > Let me make local HTML and then run them on WebKit ToT local build.
> 
> Yes - it is JSFiddle playing around. Using local copy on WebKit ToT, we pass
> all tests.

Looking at the test case, I bet this is a focus behavior difference. I suspect the JS is not able to focus the fields in the third-party iframe and thus we end up with empty fields.

[Ahmad Saleem]

(In reply to Chris Dumez from comment #9)
> (In reply to Ahmad Saleem from comment #8)
> > (In reply to Ahmad Saleem from comment #7)
> > > (In reply to Chris Dumez from comment #6)
> > > > (In reply to Ahmad Saleem from comment #5)
> > > > > (In reply to Chris Dumez from comment #4)
> > > > > > (In reply to Ahmad Saleem from comment #3)
> > > > > > > (In reply to Chris Dumez from comment #2)
> > > > > > > > (In reply to Chris Dumez from comment #1)
> > > > > > > > > https://jsfiddle.net/ts7ad8zp/show is already fully passing on trunk. The 2
> > > > > > > > > failures are not real failures, the exception name matches, only the
> > > > > > > > > exception string doesn't. We can choose whatever string/description we want
> > > > > > > > > for exceptions.
> > > > > > > > > 
> > > > > > > > > https://jsfiddle.net/pLahgb8q/show does have a few legit-looking failures
> > > > > > > > > though. Firefox and Chrome pass all the tests, we don't.
> > > > > > > > 
> > > > > > > > The weird thing though is I thought I had imported
> > > > > > > > https://jsfiddle.net/pLahgb8q/show as layout test as
> > > > > > > > LayoutTests/fast/forms/textarea-maxlength.html. However, we are passing all
> > > > > > > > the checks for this test.
> > > > > > > 
> > > > > > > I took updated copy of test from https://source.chromium.org/.
> > > > > > 
> > > > > > LayoutTests/fast/forms/textarea-maxlength.html is a straight copy of Blink's
> > > > > > web_tests/fast/forms/textarea/textarea-maxlength.html as far as I can tell.
> > > > > 
> > > > > You are right. I just did 'diffcheck' on both test and except 'Error String'
> > > > > and one comment, there is no difference in both.
> > > > > 
> > > > > Weird - if these tests are failing in JSFiddle but passing via Test Runner.
> > > > 
> > > > Indeed! Might be a behavior difference when running in a third-party iframe
> > > > (since that's how jsFiddle runs the test case).
> > > 
> > > Let me make local HTML and then run them on WebKit ToT local build.
> > 
> > Yes - it is JSFiddle playing around. Using local copy on WebKit ToT, we pass
> > all tests.
> 
> Looking at the test case, I bet this is a focus behavior difference. I
> suspect the JS is not able to focus the fields in the third-party iframe and
> thus we end up with empty fields.

Should we mark this as "RESOLVED INVALID" or rename title to something along the lines of 'focus' issue in third party iframes?

[Chris Dumez]

In Element::focus(), we have:
```
        if (!frame.hasHadUserInteraction() && !frame.isMainFrame() && !document->topOrigin().isSameOriginDomain(document->securityOrigin()))
            return;
```

Which means that we don't allow elements in third-party frames to get focus in JS without a user interaction. This behavior appears to be stricter than both Chrome and Firefox so it may be a compatibility risk

I am adding a few people in CC would know more about focus than I do to see if we should update our behavior.

[Radar WebKit Bug Importer]

<rdar://problem/106615378>

[Ahmad Saleem]

*** Bug 249907 has been marked as a duplicate of this bug. ***

[Ahmad Saleem]

*** Bug 250650 has been marked as a duplicate of this bug. ***

[Ahmad Saleem]

*** Bug 265282 has been marked as a duplicate of this bug. ***