Space key and Selenium testing framework

[jongund]

In the updated version of the Javascript for menubar-editor I switched from using event.keyCode to event.key property. The testing script menbar_menubar-navigator.js (line 500) has a test for the SPACE key and it was failing even though it worked with manual testing on the example. So I added a couple lines of code in menubar-editor.js to detect event.keyCode === 32 and if this is true change event.key to equal (see line 524). The test script now passed for all tests including the SPACE key test. Seems to me a bug with Selenium, but maybe it is something I am doing.

Test Files with "failing" tests due to the SPACE key issue:

• /test/tests/combobox_datepicker.js
• /test/tests/dialog-modal-datepicker.js
• /test/tests/menubar_menubar-navigation.js
• /test/tests/menubar_menubar-editor.js

[css-meeting-bot]

The ARIA Authoring Practices (APG) Task Force just discussed Space key and Selenium testing framework.

The full IRC log of that discussion

<MarkMccarthy> TOPIC: Space key and Selenium testing framework
<MarkMccarthy> jongund: everything's fine in the browser, but the code is funky. it's affecting everything i've done in the last couple weeks
<MarkMccarthy> mck: 1350 is kind of the root of this, right?
<MarkMccarthy> github: https://github.com//issues/1358
<Jemma> KeyboardEvent.keyCode Deprecated #1350
<MarkMccarthy> jongund: that's not a problem, we can move to <key>. the problem seems to be selenium
<MarkMccarthy> Jemma: maybe valerie could help us, since she helped with the framework
<MarkMccarthy> mck: i put 1358 in simon's queue for infrastructure
<MarkMccarthy> mck: whats unclear is what the solution is.
<MarkMccarthy> mck: was also wondering - some things in 1350...would it make sense to have a utility class or utils script to define some of this? so it's only in one place, then reuse from there?
<Jemma> s/the framework/the regression testing framework
<MarkMccarthy> jongund: i don't think we really need a utility. the main issue with <key> is that edge used different constants
<MarkMccarthy> jongund: and now that they're moving to chrome, things might be different. i don't think a utility would help. all the new event handlers are a lot more streamlines
<MarkMccarthy> s/streamlines/streamlined
<MarkMccarthy> jongund: and everything is a bit more literal and eaiser to read
<MarkMccarthy> jongund: the previous examples just used numerical constants, so this is easier. but the space issue is frustrating, especially becuase it's -just- the regression tests that are having an issue.
<MarkMccarthy> jongund: maybe we just let those tests fail for now, and rewrite the code correctly.
<ZoeBijl> regrets+
<MarkMccarthy> mck: so right now, this affects 4 PRs, right?
<MarkMccarthy> mck: so if we let them fail and allow these things to be merge-able, we can leave 1358 open and for each example marked as failing, add it as a checklist in 1358
<MarkMccarthy> jongund: that might work
<MarkMccarthy> mck: i'd be alright with that, to help us keep track
<MarkMccarthy> jongund: and to keep track of new issues as they arise
<MarkMccarthy> mck: yeah. but then we can keep track and whoever takes up 1358 will have a list to work off of
<MarkMccarthy> mck: or maybe a new version of Selenium will fix it who knows
<MarkMccarthy> jongund: right, maybe it's just an obscure bug
<MarkMccarthy> mck: okay, so let's use that plan for now.

[smhigley]

It appears there are known issues with GeckoDriver + Selenium + sendKeys(Key.SPACE) on an element that is not a text input. This works if you use sendKeys(' ') instead, or the Actions API.

Refs:
mozilla/geckodriver#777
mozilla/geckodriver#846
https://bugzilla.mozilla.org/show_bug.cgi?id=1068733

I'm going to close this since there are multiple workarounds, though anyone can reopen if they feel there's more discussion needed.

[jongund]

@smhigley

Sara thank you for figuring it out!

Jon