Feedback: "Request information about an unchecked radio button" (Rating Radio Group, Test 14, Command "'Insert Up Arrow' (browse mode on)", V25.04.22)

[jha11y]

NVDA bot out put was "Rating" which was incorrect. I think it did Up arrow instead of Insert+Up arrow.

Test Setup

• Test File: test-40-reqInfoAboutUncheckedRadio-nvda.collected.html
• AT: NVDA (Version 2024.4.2)
• Browser: Chrome (Version 135)
• Command: 'Insert+Up Arrow' (browse mode on)

[ccanash]

Thank you for reporting @jha11y . We will check it out. Moving to the aria at app repo for triaging.

[gnarf]

There is definitely something going on here. Insert + Up seems to be incorrectly detected on the machines when running in a VM. I tried reproducing this locally on my own computer, as well as a clean VM I keep around for testing the bot setups.

On my local host OS, I am able to get Insert+Up to read:

radio button  not checked    one star  radio button  not checked    two stars  radio button  not checked    three stars  radio button  not checked    four stars  radio button  not checked    five stars

However, when running in the VM ( either via the automation script, or manually pressing Insert + Up), 100% of the time it only reads Rating. Attached are some screen capture recordings, the one with the big terminal window on it is running via the bot (so wait for the second test) and the other is manually pressing run test setup -> insert space -> insert up

https://github.com/user-attachments/assets/4231b518-be7d-480c-ad52-4d67f3d58c86
https://github.com/user-attachments/assets/4a7fd86b-1f81-4085-9b6c-96dce19419cb

both of these only read "Rating"

I was looking around for other uses of Insert+Up in the tests, and ran across this one https://github.com/bocoup/aria-at-gh-actions-helper/wiki/20250501.Consistency-Report#testscommand-button-nvda-chrome (command-button - test #9 (by the count of the filename, #3 by the website), response #2 - it also seems to be reading the wrong thing.

The automation harness seems to be pressing the correct keys (as my manual testing in the same environment results in the same speech) but I'm unsure why the VM example only ever says "Rating" and my main host OS never says that. Both environments are running the same portable version of NVDA (2024.4.1 from https://github.com/bocoup/aria-at-automation-nvda-builds/releases). I'm curious if anyone else could test this in a VM environment (I'm using VMWare Workstation, the VM only has chrome, NVDA, and a few dev tools (go and node)) to see if we can figure out what it is about that environment resulting in the incorrect output.

[gnarf]

I think this problem was linked to the Insert + Up problems with "laptop" keyboard layout. The bot's NVDA had this setting instead of "desktop" - hoping this has fixed the problem, can someone please confirm and close?

[ccanash]

Yes, will coordinate to test this week @gnarf