Version
7.2.6
Host OS Type
Windows
Host OS name + version
Windows 11 Home 25H2 (Build 26200.8246)
Host Architecture
x86
Guest OS Type
Linux
Guest Architecture
x86
Guest OS name + version
Ubuntu 24.04.4 LTS
Component
Other
What happened?
When switching the input method using Super+Space, pressing CapsLock, and then switching the input source via cursor click, the CapsLock key behavior becomes permanently inverted until manually corrected. In the inverted state, typing without CapsLock produces uppercase, and CapsLock cannot be turned off normally.
Expected Behavior
CapsLock state should not be affected by input method switching. The modifier key state should remain consistent regardless of how the input source is changed.
Actual Behavior
CapsLock behavior becomes inverted:
- CapsLock OFF → uppercase input
- CapsLock ON → lowercase input
Additionally, CapsLock may become stuck and unable to be turned off through normal operation while in this inverted state.
Workarounds
Workaround A — Kana key reset (fixes CapsLock stuck issue only, does NOT fix inversion):
Press the カナ/かな key → CapsLock can be toggled again normally
Note: Pressing the カナ/かな key can restore CapsLock toggle functionality when CapsLock becomes stuck and cannot be turned off. However, the underlying inverted CapsLock state remains — CapsLock OFF still produces uppercase and CapsLock ON still produces lowercase.
Workaround B — Input source cycle (fixes CapsLock inversion):
- Switch to another input source (via Super+Space or cursor)
- Press CapsLock ON
- Switch back to the original input source → CapsLock behavior is fully restored
Additional Notes
This is likely caused by a desynchronization between the host and guest modifier key state when VirtualBox processes the Super+Space shortcut. The Super key may not be fully released in the guest's key event queue before the input source switch occurs, leaving CapsLock in an inconsistent state. The カナ/かな key workaround suggests the issue is resolvable by triggering a different modifier path. Related to VirtualBox bug #5758, but the specific trigger sequence and workarounds appear to be new findings.
How can we reproduce this?
- Open a VirtualBox Ubuntu guest with multiple input methods configured (e.g. Japanese + Japanese (Mozc) via IBus)
- Press Super+Space to switch the input method
- Press CapsLock
- Click the input source indicator with the cursor to switch the input source
- Observe that CapsLock is now inverted — typing without CapsLock produces uppercase characters
Did you upload all of your necessary log files, screenshots, etc.?
Version
7.2.6
Host OS Type
Windows
Host OS name + version
Windows 11 Home 25H2 (Build 26200.8246)
Host Architecture
x86
Guest OS Type
Linux
Guest Architecture
x86
Guest OS name + version
Ubuntu 24.04.4 LTS
Component
Other
What happened?
When switching the input method using Super+Space, pressing CapsLock, and then switching the input source via cursor click, the CapsLock key behavior becomes permanently inverted until manually corrected. In the inverted state, typing without CapsLock produces uppercase, and CapsLock cannot be turned off normally.
Expected Behavior
CapsLock state should not be affected by input method switching. The modifier key state should remain consistent regardless of how the input source is changed.
Actual Behavior
CapsLock behavior becomes inverted:
Additionally, CapsLock may become stuck and unable to be turned off through normal operation while in this inverted state.
Workarounds
Workaround A — Kana key reset (fixes CapsLock stuck issue only, does NOT fix inversion):
Press the カナ/かな key → CapsLock can be toggled again normally
Workaround B — Input source cycle (fixes CapsLock inversion):
Additional Notes
This is likely caused by a desynchronization between the host and guest modifier key state when VirtualBox processes the Super+Space shortcut. The Super key may not be fully released in the guest's key event queue before the input source switch occurs, leaving CapsLock in an inconsistent state. The カナ/かな key workaround suggests the issue is resolvable by triggering a different modifier path. Related to VirtualBox bug #5758, but the specific trigger sequence and workarounds appear to be new findings.
How can we reproduce this?
Did you upload all of your necessary log files, screenshots, etc.?