VirtualBox 4.3 released | Hacker News Reader

VirtualBox 4.3 released

virtualbox.org

155 pointsconductor54 comments

A couple important notes about this:

* Vagrant is being updated to work with 4.3 now. A release should be out today or at the latest tomorrow. (UPDATE: Vagrant 1.3.5 is now out and supports VirtualBox 4.3)

* VirtualBox 4.3 doesn't run at all on Mavericks (10.9) because their kernel extensions aren't signed. OS X 10.9 requires signed kexts now. So the changelog where they said "limited" Mavericks support they actually should've said "no support". (UPDATE: Some people are reporting it is working for them on Mavericks. I can't get it to work. YMMV)

Based on these two bullet points, I would stick with VirtualBox 4.2 for the time being. The bullet point that says "rewritten VT-x and AMD-V code" is especially vague and could be "super unstable virtual machine manager" just as easily as it can mean "slight performance improvements." So be careful.

Other than that, it is good to see VirtualBox have some sort of big release. This is their first major ("4.x") release in over a year.

I also want to note that if you are on Mavericks, VMware Fusion works perfectly. As a disclaimer to this sentence though: I make money from Vagrant + VMware users. I'm not trying to advertise that, I just want to state that Vagrant _itself_ is fine. VirtualBox 4.3 doesn't work. VirtualBox 4.2 does. VMware Fusion 5 and 6 does.

> VirtualBox doesn't work. VMware does.

I haven't tried Virtualbox 4.3 but Virtualbox 4.2 sort of works on Mavericks. The kernel extensions don't load on boot, but can be loaded manually. I blogged about that here http://asquera.de/development/2013/06/20/vagrant-on-maverick...

Good catch. I updated that. Thanks!

It also depends on how your Gatekeeper settings are set; mine are set to "Mac App store and indentified developers". The OS alerted me that the kexts were unsigned, it said it would load them anyway.

Curious. I went ahead and updated, as my use of VirtualBox isn't critical to my workflow.

I received warning during install of the unsigned kexts, but the warning stated that they would be loaded anyway. I'm now running vbox 4.3 just fine on Mavericks GM.

Thanks a lot for this comment. After upgrading to Mavericks I had to ditch my hacked together and hand-compiled PHP setup, so I converted to Vagrant+Virtualbox. I'd have hated to break that with the upgrade to 4.3 haha.

I made something for you: https://puphpet.com

I tried googling for mac os x signed kernel extensions, but couldn't find anything useful.

Is this an easy issue to fix (get free key, slap a signature on), or could this be a more serious issue?

The change is still under NDA. If you sign into your Apple dev account and look up the WWDC talk on OS X kernel extension development, you'll find it within that slide deck.

Deep, I know.

I also reported the issue to the VirtualBox team many weeks ago and they confirmed the bug and know it exists. It just seems like it wasn't important enough for them to fix prior to release.

Honestly, the fact that the change is under NDA seems like reason enough for them to not bother yet. Who knows what NDA violation an open source project could make in discussing how they are going to implement it.

I know what NDA violation they could make. It is called "NDA violation". It is where you sign an NDA and then you disclose information. Release of source code could be considered a disclosure. Now Apple can go right ahead and do a whole lot of nothing against VirtualBox and Oracle because no one cares. Neither should you. Stop lionizing contract law. It is a civil matter and no big deal almost always and especially when no money has changed hands.

http://docs.vagrantup.com/v2/vmware/known-issues.html

So this is now the only issue left. If I remember correct there used to be multiple issues with VMWare + Vagrant. I just checked now and seems only one left. Correct?

I might actually switch to fusion + vagrant on my mac osx if the support is stable.

I can attest to the solid Vagrant+VMWare Fusion support. I've been using it for at least the last six months and have not seen a single problem.

I've completely stopped using VirtualBox now! Good work mitchellh :)

That is correct. The adapter is continuously improved and is at a very stable point now.

Awesome, thanks for the information - I've been using VirtualBox + Vagrant nightly for the last few weeks with a huge smile on my face, I really love that tool. What a pain in the ass it is to get a native RoR dev. environment working...

Unsigned extensions should run just fine under Mavericks, you'll get an error/warning message on first install but a reboot should load it fine

Does it perhaps depend on your Gatekeeper settings? which could explain the diverse experiences being reported.

It does, Gatekeeper has gotten a little more restrictive as the seeds have gone out (at least compared to where it started). If you don't have "allow me to install whatever I fucking want" checked Gatekeeper will refuse to load unsigned Kexts...which I suppose is ultimately not a bad thing. The majority of users have no idea what a kext is or recognize the security implications of them

actually, Vagrant works on Mavericks but on every updated version, from beta to beta, I had to remove Vagrant completely and reinstall vagrant fresh.. the kernel extensions don't seem to carry over, but it definitely working on Mavericks

VirtualBox is one of those projects that is a very important part of my daily use. But somehow I feel that it doesn't get much love from the open source community. Is it just me?

VirtualBox has a quality problem: https://lkml.org/lkml/2011/10/6/317

This has gotten a lot better, but for quite a while `vbox` was the thing that made my computer lock up most of the time, especially the USB drivers.

I've always found performance and stability to be quite lacking with vbox. Parallels and VMWare are both mature, performant, frequently updated, and well-supported, so I see little reason not to use one of those, especially given that I'm using it for work and don't want to wait on a VM for no good reason.

On the other hand, Parallels and VMWare frequently charge for their updates. I rarely find software that has such short license spans. I bought Fusion 5.0.3 in June and they already want to charge me for Fusion 6 (btw. the only version that officially supports Mavericks). Fusion 5 is now dead in the water.

This is okay for me on a single machine, but at scale of a development team, it gets expensive quick.

I installed 4.3 today on my iMac, the machine crashed twice within a couple of hours, it has never crashed otherwise in 2+ years I have owned the machine. Uninstalled, will upgrade Parallels if necessary.

When I was hired at my current company everyone used Virtualbox for their Linux test environments. Unfortunately, part of my job was building/rebuilding custom ISOs, which required long days of tweak->rebuild ISO->rebuild VM.

After I changed from job-installed Virtualbox to the Parallels license I already had, my workflow sped up dramatically. It was easily half the time to go through an install cycle as before. The performance increase was absolutely massive. Since then we've only had more and more issues with Virtualbox that makes it worth replacing at a relatively high cost.

I've tried contributing to it, but several factors makes it really hard.

They includes:

  - Poor documentation
  - Broken build scripts for some of the open source version
  - Lack of support on forum/IRC
  - Project still uses SVN
  - Procedure for submitting patch is very archaic

I personally would love to see them move to Github. It would make contributing much easier.

Isn't the choice of a version control systema a too low barrier to prevent anyone from contributing? I'd think that actual technical skills required in the VM domain area plus language skills would be a much greater barrier to entry, isn't so? Do we need everything to be so perfect?

I don't know the full story, but from what I've heard, it probably involves things like 1) Oracle owns it 2) It requires a non-free compiler to build the BIOS

1) It's GPL'd so nothing is stopping you or anyone else from forking it away from Oracle. 2) The Open Watcom compiler is not required to build VirtualBox at all. The actual compiler is free for anyone to download and use.[1]

Other people here raised some legitimate points with the build and patch submission process that I also would like to see improved, but yours are just narrow minded prejudice.

[1] https://www.virtualbox.org/ticket/12011

VirtualBox is only partially open source and free:

For example, USB2.0 and RDP support is non-free.

https://www.virtualbox.org/wiki/Licensing_FAQ

Also note that they do openly admit that stopping to release under GPL 2.0 is not planned, but possible and they reserve the right to do so in future.

Please tell this to the Debian team (that they're narrow minded and prejudice) because I'd like them to move VB off of contrib (non free) to their free repos.

The Linux community are focused on KVM, which is already in-tree and carried by most distributions.

I understand that this leaves Mac users behind. But OS X isn't open source, so there we are.

Vbox's "api" is kind of an abomination. Vagrant succeeds because it parses sometimes non-machine friendly command line executable output.

If you rely on VirtualBox, I would recommend to hold off on upgrading until you are sure that it will not break your workflow.

The big releases tend to have a way of introducing bugs, that although get fixed, can wreck your day if you depend on it.

I just upgraded to 4.3 (Win7 host, Sandy Bridge Core i7 processor), and VirtualBox is now complaining that I've got more virtual cores (4) assigned to my guest than I have physical cores (2) on my host.

While this is true, this wasn't an issue with 4.2. It still lets me run it, but it's not happy about it.

EDIT: Further investigation shows that VirtualBox docs have always said this was a bad thing to do, but didn't used to throw up a warning when you did it anyways.

> ... but didn't used to throw up a warning when you did it anyways.

This is mentioned in the changelog, FWIW.

“Networking improvements: A new Network Address Translation (NAT) option allows virtual machines to talk to each other on the same host, and communicate with the outside world. ”

This feature alone is worth the upgrade.

Nice, but a couple of months too late for me.

I've switched over to Parallels 9 on the Mac, and it is _much_ faster overall (not to mention better USB support).

Also, I highly recommend Vagrant users to look into the vagrant-LXC plugin - that, too, is one heck of a lot faster than using either VirtualBox or VMware providers.

I've documented my setup here: http://the.taoofmac.com/space/HOWTO/Vagrant

Congrats to the virtualbox team!

I still wish they'd consider adding retina support on OS X; that's the one thing keeping me from using it 24/7 :(.

I am using 4.2 branch on my retina pro from early 2013. What's the issue? Do you have issue with launch a desktop version of Ubuntu?

I normally launch server instance so I am fine. Just curious.

It doesn't have retina support, so everything looks pixelated.

got it. thanks!

I used virtualbox a lot for work and play, always with a debian guest. While my overall experience was good, there were two gotchas that really annoyed me:

1. When cloning VMs I would always have networking issues. The fix was known and simple (https://www.virtualbox.org/ticket/660) but not intuitive to a casual user.

2. Installing the guest additions (drag-and-drop file support, shared clipboard, basically stuff you really want) as a kernel module can be a huge pain in the ass depending on what kernel you run. I never had any issues with a "stable" 2.x kernel, but with 3.x I had a difficult time finding the correct kernel headers and putting them in the correct place.

some.distros have the guest additions in their repositories, that makes it easier.

I just installed 4.3 and tried running a directx9 program on XP with 3d acceleration and it caused the VM screen to flicker and not display any content.

I hope Vbox will catch up with VMware player on 3d acceleration soon.

Same issue here, just different Win version. In my case it was easy to solve since I didn't really need 3D acceleration, so I ended up disabling it.

It is worth noting that if you don't use the new features and running VirtualBox on old hardware there probably is no good reason to upgrade at the moment.

I have used virtualbox and vmware player (free for non commercial user ). vmware player gave me much better experience.

I have used virtualbox extensively and now have had to use VMware workstation 9 and 10 this last month and I much prefer virtual box. One more anecdote!

Out of curiosity what is generally your host/guest OSes? Things had gotten pretty buggy with VirtualBox running Ubuntu 13.04 inside a Windows 8 PC. (Wonder if it was just me or others too.)

2 out of 4 my boxes are Win 8 and I haven't noticed anything funny with Ubuntu 13.04. I use Linux guests fairly extensively.

I've used both for years, I consistently find VirtualBox to provide a much better experience.

Same here. I had been completely reliant on VirtualBox for ages. About a year or so ago it felt as though advancements slowed down, performance got incredibly slow, and things became buggy.

I started using VMWare Player which definitely is a lot more polished, reliable, and way faster in my opinion.

Maybe someone from the VirtualBox team can explain to me why VirtualBox for Ubuntu has a dependency on acroread-bin.

Reply on news.ycombinator.com