Originally filed as: https://github.com/mozilla/geckodriver/issues/2139

Note that this problem only occurs when running Nightly and early beta builds of Firefox. Using mozregression bug 1840519 has been revealed.

Note that this happens only in these builds of Firefox because there the preference dom.event.keypress.dispatch_once_per_surrogate_pair is set to false. If it's true the following Marionette testcase will pass:

    def test(self):
        self.marionette.set_pref(
            "dom.event.keypress.dispatch_once_per_surrogate_pair", False)

        self.marionette.navigate(
            "https://www.selenium.dev/selenium/web/cn-test.html")
        elem = self.marionette.find_element(By.NAME, "i18n")
      
        text = "𠀀𠜎𠀋𪆐𪚲"
        elem.send_keys(text)
        self.assertEqual(elem.get_property("value"), text)

Masayuki, can you please take a look? For now I filed it as a Marionette bug but I assume it's somewhere in DOMWindowUtils or platform code?

This testcase can easily be run after applying with the following command:

./mach marionette-test --headless -vv --gecko-log - --setpref="remote.log.truncate=true" testing/marionette/harness/marionette_harness/tests/unit/test_typing.py

I guess that the Marionette synthesizes different style of keypress events.

It seems that here splits surrogate pairs to surrogate:
https://searchfox.org/mozilla-central/rev/edb2612db13e89f1c44ab95b1e4d4366c16eb9fb/testing/marionette/client/marionette_driver/marionette.py#188-190

However, I'm not familiar with Python and its string so that I'm not 100% sure whether here is buggy, and I'm not sure whether Python can check whether a surrogate or not for each "character" smoothly.

Another possible fix point (if here splits surrogate pairs) is, at the receiver side. If it gets a high-surrogate keydown and/or keyup event, make it wait for the following keydown and/or keyup for a low-surrogate. Then, they can synthesize a set of keys with checking the pref. I guess that this is better solution, but I've not found the receiver side. Could you let me know where is it?

(I.e., a surrogate pair will synthesize a set of keydown and keyup with one or two keypress events once the new behavior will be shipped and that's what same behavior as user input at least on Windows.)

And what do you think for this comment from your point of view?

Discussed in todays triage meeting and James will have a look. If it's a problem with Marionette we might turn off this new feature for now until we got it correctly implemented in Marionette.

Just two additional maybe related issues that I noticed:

• Broken WPT tests for multiple Unicode scalar values: https://github.com/web-platform-tests/wpt/pull/40785
• IME support in actions: https://github.com/w3c/webdriver/issues/1683

This is mostly reusing the implementation from the marionette client.

Previously, given a string in element send keys, we iterated over each
UCS2 "character" in the string, so that non-BMP characters ended up as
separate key events, one for the high surrogate and one for the low
surrogate.

With this change we instead use the built in string iterator which
provides the surrogate pair in the case of non-BMP characters, so we
emit a single event for each unicode codepoint.

The second patch also needs to be pushed. As such lets mark as leave-open for now.

https://hg.mozilla.org/mozilla-central/rev/b2886b896d0d

https://hg.mozilla.org/mozilla-central/rev/f19c3570bc57

The patch landed in nightly and beta is affected.
:jgraham, is this bug important enough to require an uplift?

• If yes, please nominate the patch for beta approval.
• If no, please set status-firefox122 to wontfix.

For more information, please visit BugBot documentation.

The fix here on this bug is actually for a feature that is only enabled in Firefox Nightly. So beta is not affected.