This is a step in the right direction. I’ve been using Python 3 for all my projects in the last few months and I have to say, outside of whatever performance differences there might be (haven’t noticed them yet) the transition has been smooth, and things like MyPy are great.
“The swap is now a file, not a partition that will scale to what your system needs, making it easier to install Ubuntu on any machine.”
Huh. I vaguely remember something about systems that supported swap files tearing support out. As I recall, it was because it was deadlock prone. The file system may need to allocate in the write path, which triggers swapping, which needs to write, which triggers allocations, which triggers swapping, which needs to write… you get the idea.
“The swap is now a file, not a partition that will scale to what your system needs, making it easier to install Ubuntu on any machine.”
Would there be any issues with doing this as opposed to keeping your swap on a separate partition? Is the only upside ease of installation for those unfamiliar with disk partitioning?
There’s potentially a very slight performance difference, but it would be tiny. It’s far outweighed by the benefits at installation time. Depending on how they’re doing it (I haven’t looked in depth), it would also simplify the handling of encrypted swap (that is, they wouldn’t have to do it anymore; if the root filesystem is on an encrypted partition, encrypted swap comes for free).
The only downside of a swap file for me is that it makes hibernation a bit more complicated because you must specify the offset of the file from the partition start as a kernel parameter. But this is only a minor issue and I haven’t used a dedicated swap partition in years.
I’d be surprised; keeping swap in a file has been supported for well over a decade (could be two or even three, I’m not a historian).
I’m very familiar with disk partitioning, but I’m also very happy not to have to do it! And swap is a pretty safe thing to put on even your root partition, as it’s fixed-size and won’t surprise you by filling your disk.
I actually had an issue with my swap on the last 17.10 beta. There existed a /swapfile, and it was listed in /etc/fstab, but for some reason it was not being used (swapon -s and top didn’t show it). A couple of times when I opened a few Chrome tabs in a row, the machine went south, and as far as I could tell kswapd was going nuts. Both times the machine hung.
I blasted the old /swapfile, made a new one in the usual way, activated it, and now the machine is running fine.
I’ve tried to dig into this in the past, but what is the difference between the .04 and the .10 ubuntu releases? is the .10 incremental updates to the .04 but the .04 is the long term support?
I’m interested in how the transition to Wayland goes. I’ve heard that there are still some problems with some mainstream graphics cards and older games that rely on Xorg…
Wayland is the default on Fedora, and has noticeable performance benefits (it feels like 60FPS). The only problems I’ve noticed are Chromecast not working and color picker picking black all the time. This and memory bloat, but that might be a Fedora issue, I have to compare with regular X gnome.
Improved security (X was designed for another era)
Simpler pipeline from calling an API to draw something to it actually displaying
Fractional HiDPI support: I hate having a high resolution laptop and connecting it to a “normal resolution” 30in monitor. Everything is out of whack
Last but not least: it might weed out some of the ancient UI libraries that people are still using for new software. I hate to see old software go to waste, but a more consistent desktop experience would be nice.
Xwayland will be supported for a long time. The only software that will truly go to waste is stuff that relies on security holes access to everything all the time: utilities for global keyboard shortcuts, screenshots, etc.
Also from release notes: “Python 2 is no longer installed by default. Python 3 has been updated to 3.6.”
This is a step in the right direction. I’ve been using Python 3 for all my projects in the last few months and I have to say, outside of whatever performance differences there might be (haven’t noticed them yet) the transition has been smooth, and things like MyPy are great.
On that note, I got 10x speedup with PyPy which now supports Python 3.5. I’m super impressed, but it only works well on server-type workloads.
Some interesting changes, from: https://www.ubuntu.com/desktop/1710
“The biggest change is the desktop environment. Ubuntu 17.10 is retiring Unity in favour of GNOME, version 3.26.1.”
“Ubuntu 17.10 features the Linux kernel 4.13.”
“The swap is now a file, not a partition that will scale to what your system needs, making it easier to install Ubuntu on any machine.”
“Ubuntu 17.10 will be supported for 9 months until July 2018. If you need Long Term Support, it is recommended you use Ubuntu 16.04 LTS instead.”
Huh. I vaguely remember something about systems that supported swap files tearing support out. As I recall, it was because it was deadlock prone. The file system may need to allocate in the write path, which triggers swapping, which needs to write, which triggers allocations, which triggers swapping, which needs to write… you get the idea.
Windows swaps to a file on NTFS just fine…
It must’ve been about ZFS. Swapping to a file on ZFS or a zvol WILL deadlock :)
Would there be any issues with doing this as opposed to keeping your swap on a separate partition? Is the only upside ease of installation for those unfamiliar with disk partitioning?
There’s potentially a very slight performance difference, but it would be tiny. It’s far outweighed by the benefits at installation time. Depending on how they’re doing it (I haven’t looked in depth), it would also simplify the handling of encrypted swap (that is, they wouldn’t have to do it anymore; if the root filesystem is on an encrypted partition, encrypted swap comes for free).
The only downside of a swap file for me is that it makes hibernation a bit more complicated because you must specify the offset of the file from the partition start as a kernel parameter. But this is only a minor issue and I haven’t used a dedicated swap partition in years.
I hadn’t thought about hibernation…but then again, I gave up trying to get it working on any OS ten years ago and haven’t thought about it since. :)
I’d be surprised; keeping swap in a file has been supported for well over a decade (could be two or even three, I’m not a historian).
I’m very familiar with disk partitioning, but I’m also very happy not to have to do it! And swap is a pretty safe thing to put on even your root partition, as it’s fixed-size and won’t surprise you by filling your disk.
I actually had an issue with my swap on the last 17.10 beta. There existed a
/swapfile, and it was listed in/etc/fstab, but for some reason it was not being used (swapon -sandtopdidn’t show it). A couple of times when I opened a few Chrome tabs in a row, the machine went south, and as far as I could tellkswapdwas going nuts. Both times the machine hung.I blasted the old
/swapfile, made a new one in the usual way, activated it, and now the machine is running fine.I’ve tried to dig into this in the past, but what is the difference between the .04 and the .10 ubuntu releases? is the .10 incremental updates to the .04 but the .04 is the long term support?
04 and 10 are the April and Oct annual releases. LTS is long term support == five years. “Even”.04 is LTS, so 16.04 is LTS and 18.04 will be LTS.
If you are running an Ubuntu server in production, you should be using an LTS release.
You’ll notice that the .04 are released in April and the .10 in October ;)
I’m interested in how the transition to Wayland goes. I’ve heard that there are still some problems with some mainstream graphics cards and older games that rely on Xorg…
Wayland is the default on Fedora, and has noticeable performance benefits (it feels like 60FPS). The only problems I’ve noticed are Chromecast not working and color picker picking black all the time. This and memory bloat, but that might be a Fedora issue, I have to compare with regular X gnome.
games/anything else. Non GTK+ software ends up running under xwayland… I just don’t understand all this wayland-fever
A drowning victim will reach out for anything. X11 is not a nice stack to develop on.
Like, directly? Maybe not. But even xlib is not bad and most people use much more abstract things than that :)
Just curious, what would you consider a worse UI dev experience than xlib?
Trying to build a rich application UI in HTML5+CSS3 ;) (which is a big part of $DAYJOB)
Seriously, though, it depends on the kind of UI you’re building and other such things.
There’s a few reasons why I’m bullish:
Last but not least: it might weed out some of the ancient UI libraries that people are still using for new software. I hate to see old software go to waste, but a more consistent desktop experience would be nice.
Xwayland will be supported for a long time. The only software that will truly go to waste is stuff that relies on
security holesaccess to everything all the time: utilities for global keyboard shortcuts, screenshots, etc.