Matthew Dillon’s made some scheduler changes, which blogbench tests are showing give the default scheduler better performance under heavy load. It’s a pretty technical writeup, so I’ll just point you at it rather than attempt to summarize.
So, I felt lucky recently. I updated shiningsilence.com from DragonFly 2.6 to DragonFly 2.8, and wanted to upgrade my pkgsrc packages from pkgsrc-2010Q3 to pkgsrc-2010Q4.
You can do this with pkg_rolling-replace, or various other tools, but I wanted to see if I could do it completely with binary packages. I used pkg_radd -uv <pkgname> for each of the major packages I had installed.
Surprisingly, it worked, for every package. I had to force-install some Perl modules because I was moving from 5.10 to 5.12, but I think I may have been able to use an additional -u switch to get by that problem. I did use pkg_leaves to identify packages I didn’t need, and removed them to reduce the number of items to upload.
It was exactly what I wanted. Previous pkgsrc upgrades had taken most of a day, as I had to build from source and figure out what went where. We’ve had a better success rate in bulk builds recently, and this paid off in an upgrade process that only took perhaps an hour.
If you’re interested in mentoring for DragonFly and Google Summer of Code for 2011, please speak up. You don’t have to mentor if you don’t see any projects you like – I just need an initial count for the application. If you don’t want to mentor at all, but you’ve got ideas: there’s a place to tell people about it.
Peter Avalos updated OpenSSL to 1.0.0d, due to a recently discovered bug. He’s also brought the fix back to DragonFly 2.6 and 2.8, so it’s available for most anyone.
BIND version 9.5 has reached End of Life status. In fact, it did it some time ago. However, net/bind95 in pkgsrc has just been deleted. Update to 9.6/9.7, if you still had9.5 in place.
This is one of those scenarios that I’m noting because it might bite someone, some day: if your root partition is encrypted, you can’t fit in a different keymap. However, kernel options to build in a different keymap will fix this issue.
If you’ve ever installed pkgsrc packages from source, you probably typed ‘bmake install’. There’s a ‘bin-install’ target that will use binary packages if they are available, but you have to set the appropriate environment variables to do so.
It’s now much easier, on DragonFly. If you have pkgsrc-current as of yesterday or later, or pkgsrc-2011Q1 when it arrives, you can type ‘bmake bin-install’ for a pkgsrc application and it will download the binary package automatically, if it’s available, and build from source if it’s not.
This is a setup I’ve wanted for a while – the speed of a binary install, plus a fallback if the binary isn’t available.
If you’re using the binary pkgsrc package installer pkgin, version 0.4 is available for testing.
They’re finally uploaded! See my rather lengthy post about it on users@ for all the details.
Chris Turner went off into some extra detail on how the rc system works, with extra links for anyone interested in some history.
Matthew Dillon’s made some changes that will speed up the booting process for people with a ridiculous amount of memory, like 64G. This is x86_64 only, but that should not be a surprise if you think about it.
Due to a crash yesterday on git.dragonflybsd.org, the Git repo was not up to date, briefly. It’s been fixed. This will only really matter if you’re running bleeding-edge DragonFly and rebuilt your system in the last 24 hours or so.
I’ve had the bulk builds of pkgsrc-2010Q4 finish on 2.9 systems, for i386 and for x86_64. The uploads for 2.9/x86_64 seem to have completed…
‘file’ has been updated to version 5.05 by Peter Avalos. file(1) is one of those utilities that I forget is a contributed, external piece of software, even though it’s been in Unix since 1973.
(file is one year older than me!)
Sourceforge had/has a security problem, so they’ve turned off some services until it’s fixed. However, anything planning to download from Sourceforge will be affected, so some packages in pkgsrc may not be able to build for … a day or two?
Tim Bisson posted new network tests contrasting the virtio driver against emulated re(4) in virtual environments. Previously, the virtio driver performed worse, but a more developed test suite seems to deliver more positive results.
Samuel J. Greear has written a summary of DragonFly’s experience with Google Code-In 2011, noting that the students tacked harder projects than expected, and relatively easy documentation projects were less popular than expected. He has hard numbers on tasks done, too.
I think this article holds the “number of hyphens in a title” record for this blog.