[SOLVED] Breakdown of automatic pointer capture in VBox

[tarazed]

From way back I have experienced an annoying problem when installing ISOs in VirtualBox. This affects the installation process only.

Mouse pointer and keyboard capture is automatic but while navigating the installation screens I find that the VBox manager loses the mouse pointer after a few clicks. The pointer freezes and the host mouse pointer appears outside the box. It is possible to reactivate the guest pointer by moving the host pointer back over the Vbox screen. The guest pointer at this stage is restricted to a small portion of the window so some options cannot be reached unless the tab key is pressed. Using the capture key (Right Control) reestablishes full capture of the mouse pointer but this lasts for only a few clicks. This is a royal pain in the fundament and really slows the whole process down. Experiments with click-to-focus and focus-follows-mouse in the host did not solve the problem. My preference is for sloppy focus.

Once the installation is effected there is no more trouble. Mouse pointer capture works perfectly after reboot.

Has anybody else suffered this problem and managed to solve it?

[doktor5000]

Does that happen for all distros you tested?

[tarazed]

Yes, as far as I remember. They have mostly been mga5 alphas and betas in the QA suite. I have been testing a subset of them. Non QA isos like mga4 and mga4.1 may have given trouble as well and I think it applied to both architectures. It looks like I shall have to run through a few test installs to make sure of the facts.

[tarazed]

Tried two things at once; switched from Mate to KDE and started to install mga4 x86_64. Started to lose the pointer very soon after starting and by the time custom disk partitioning was finished I was so fed up that I crashed out of the install. When pointer control is compromised you have to press the capture key twice - the little arrow goes from green to black and back to green. The error/bug/problem is that the virtualbox manager loses track of what the window size should be, as if it is miscounting the number of pixels the mouse traverses. The fault is exhibited on at least three machines with both wireless and USB mice. The final VM does not exhibit any of this behaviour, which might suggest there might be a fault in isolinux. Another question arises from that assumption; why then has this never been reported by anybody else? Time to check Bugzilla I think.

Another possibility has just occurred to me; maybe I have never configured/installed VirtualBox properly, missed out an addon or whatever.

[tarazed]

Checking the log for the aborted install I see several attempts to resize the screen over a period of 9 seconds.

Code: Select all

UIFrameBufferQImage::RequestResize .... w=800 h=600

Code: Select all

UIFrameBufferQImage::resizeEvent... 800x600
....Resizing to FALLBACK buffer due to format is invalid

These excerpts include just the pertinent bits.
The size requests are for 800x600, 720x400 and 640x480. All seem to report failure.

So, probably nothing to do with losing track of the mouse but a possible bug in the frame handling.

[tarazed]

This seems to have been resolved in the latest revision for Mageia Cauldron, virtualbox 4.3.26.

[doktor5000]

Please mark the thread accordingly by editing the topic of the first post and prefix it by [SOLVED], thanks