Mouse is off by x1.2 in OSX guests

[socratis]

Version 5.1.0_RC1 r108607
Host: OSX 10.9.5
Guest: Any OSX client.

Symptom: When the host mouse is over the center of the guest, host and guest cursor are in the same position. But if you move towards any of the four corners, the guest cursor seems to move faster, I would estimate by around 120 to 130%. So the guest cursor reaches the corner faster than the host. The real fun is in the middle areas where you can't tell which is which and you click at all the wrong places. Think "OK" instead of "Cancel".

[michaln]

But that's not new at all?

OS X does internal scaling for absolute pointing devices so that the pointer in the guest moves outwards from the center faster than the pointing device indicates, and there is corresponding "padding" around the edges where the pointing device moves but the cursor in OS X does not. It's always been that way.

[socratis]

But that's not new at all? ... It's always been that way.

No, no, no, no, no! It's never been that way! That's the first time I've seen anything like that! I can easily switch between several versions, from 4.1.x all the way to 5.0.24 and NONE has exhibited that behavior! That's a brand new thing. If I hadn't been bitten seriously by the <ExtraData> issue that was really preventing me from doing anything, I could have noticed earlier. To tell you the truth, I might have ( ), except data loss seemed more important at the time.

[michaln]

In that case, I have no idea what you're talking about. OS X scales the device coordinates and always has.

[socratis]

So, would a video demo would convince you? Same VM(s), different VBox. Just pick your VBox version, any VBox < 5.1.0.x. Any.

[socratis]

I'm even willing to create a backup of a VM (because I'm going to lose data), switch back to 5.1.0b1 and show you the effect. Just pick your OSX version (doesn't matter actually).

[michaln]

It's not about convincing. I simply don't understand what you're talking about. So if that can be explained with a video, sure, that would help.

On my host, the behavior I see in OS X guests is exactly the same it's always been, and the roughly 1.2x scaling is exactly what OS X has always been doing. That's why I don't know what you're talking about.

[michaln]

Actually... I just realized that my El Capitan VM does not do the scaling. Snow Leopard does it, Mavericks does it. Again, it's something the guest OS does, not VirtualBox.

[socratis]

Again, it's something the guest OS does, not VirtualBox.

Again, it cannot be something that the OS does, if it does it with this VBox version only. How can it be the OS since that variable hasn't changed? Change one variable (VBox>5.1b1) and you have a problem. I didn't make a movie, I made some screenshots. Faster, easier. BTW, as I said this happens with every OSX client, from 10.5 to 10.11 (no access to 10.12 yet).

510rc1-1.png (133.57 KiB) Viewed 14020 times

510rc1-2.png (132.87 KiB) Viewed 14020 times

510rc1-3.png (132.62 KiB) Viewed 14020 times

[socratis]

just realized that my El Capitan VM does not do the scaling. Snow Leopard does it, Mavericks does it.

every OSX client, from 10.5 to 10.11 10.9

OK, scratch mine. I could swear it was in all of them, but that's what happens when you test 7 VMs at the same time. After an all-nighter

10.10 and 10.11 don't do it. What you can notice however is that it seems that there are 2 cursors which move "in parallel". If you move the mouse fast enough you can see the guest's mouse trying to "catch up". Sort of like the mouse trails effect in some OSes.

So, somehow there was a separation(?) between the guest's and the host's mice? Are they drawing on their own? And in versions prior to 10.9 the scaling is off? It certainly didn't feel like that up to 5.0.24.

I remember (but can't find at the moment) an issue like that in recent VirtualBox versions, less than a month ago, with shadow cursors. Not on OSX guests but I believe it was Linux. I'll try to find it and I'll update this thread, in the rare case that there is some relevance.

[socratis]

it seems that there are 2 cursors which move "in parallel".

The same behavior appears on other VMs like OS/2 14.106_W4, Android 4.4.4. The effect is more pronounced if you have a display scaling factor > 1, something like the picture I attach with a scaling factor of 1.25. The larger the scaling factor, the more pronounced the effect.

Screenshot (64x64) has been magnified by 3 so that you can easier see what I'm talking about (192x192).

510rc1-OS2.png (54.29 KiB) Viewed 14014 times

[socratis]

@michaln
I found the thread that I remembered seeing similar behavior: viewtopic.php?f=6&t=77475&start=15#p361679
Thing is that it didn't do it for me (10.9.5) so far.

[michaln]

OK, now it makes sense. The bug is that both the host and guest cursors are shown. The ~1.2x scaling is really nothing new (and gone from latest OS X versions!) but if you only see the guest cursor, as it should be, it's not as noticeable. If the guest OS does not do such scaling then the two cursors are not so obvious.

[socratis]

The bug is that both the host and guest cursors are shown.

Yeap, that makes sense, because I've noticed it in other guests as well where it wasn't showing before. Case in point, booting into safe mode in Windows. You have double cursors. Are you telling me that prior to this behavior you were "hiding" the host cursor, that's why we weren't seeing this?

[socratis]

It's still happening after 5.1 was released. I would like to know if this is been worked on (an acknowledgment would be OK) or if I should open a bug report. I don't want to, because you must have a lot on your plate, but if it means it would be better as a bug report so it won't be forgotten, I will.