I had mentioned the new malloc in DragonFly before, but Matthew Dillon has a nice explanation of its performance improvements and the relationship to the existing slab allocator.
Month: April 2009
Hasso Tepper, who has been working very hard on pkgsrc on DragonFly, has a few strange pkgsrc issues he’d like help on. Anyone have ideas? (Follow the thread to see what’s been done so far.)
Make Hasso Tepper’s life a bit easier and take heed of this list: Maintaining stuff in pkgsrc.
Edit: Meant to publish this a bit ago; missed it. Sorry!
Bring in money and then take it out again. It’s conceivable that Impi Linux would have fizzled on its own, being one of a zillion Linux distributions out there, but becoming a commercial product seems to put different, and tougher, contstrictions on any open source project.
The 2.2.1 release of DragonFly, rounding up changes since the release (I don’t have a list), should be tomorrow.
Mashing together to make one post:
These PC-BSD 7.1 vs. Kubuntu 9.04 Benchmarks are interesting but not that conclusive – different versions of gcc were used. (thanks J. Kanowitz) Here’s a different comparison of performance inside a VM from Ivan Voras.
Naoya Sugioka has kqemu working with the intended performance improvements; please test if you use qemu. (Johannes Hofmann has done some initial tests.)
Hasso Tepper has lang/gcc3 working in DragonFly 2.3, which means that we could potentially drop the version of gcc3 in contrib/ after the 2.4 release. He asks for other testers, please.
This would have the nice side effect of speeding up buildworld tremendously, as gcc3 and gcc4 get rebuilt each time.
- Thanks to Stathis Kamperis, it is now possible to build DragonFly 2.2 on a DragonFly 2.3 system, if for some reason you need to move to a system from before the recent libc changes.
- Matthew Dillon has replaced the existing BSD malloc with a port of the slab allocator, which makes malloc() faster, with minor benefits for a buildworld.
- Matthew Dillon also has a patch for people wanting to look for the elusive ‘file-missing-in-directory-listing’ Hammer bug. Caveat Emptor.
There’s a new @Play column focusing on more of the entries at the Seven Day RogueLike competition. I mention this because roguelikes have been around on Unix-ish systems forever, some of these may work on DragonFly, and because they are much more complex and interesting than I would have thought possible.
Of particular interest is Robert Luciani’s talk about M:N threading in DragonFly. Yes, that’s the same Robert Luciani who is participating in Summer of Code with DragonFly to profile kernel contention on multiprocessor systems.
There’s 5 slots for DragonFly in Summer of Code for 2009, and the students in those slots are listed below. We had some very good applications; more than we had room for and higher quality than last year. If you did not get in, please consider working independently.
Student: Alexander Hornung
Project: DevFS for DragonFly BSD
Mentor: joe talbott
Student: Dan Chis
Project: Support debugging of multi-threaded applications
Mentor: schubert simon
Student: robert luciani
Project: Profile kernel contention on MP systems
Mentor: Samuel Greear
Student: Jordan Gordeev
Project: Finish amd64 port of DragonFly
Mentor: Matthew Dillon
Student: efstathios kamperis
Project: C99/POSIX Conformance Audit
Mentor: hasso tepper
Bad: having the system used for mirroring DragonFly crash a lot.
I’ll note here that I have a Hauppauge TV card (848 chipset, I think) that’s possibly the oldest continuously functioning computer equipment I own; I’ve been using it for close to a decade without a single problem. I have nothing else that has reached the benchmark.
Student projects for Google Summer of Code will be announced this Monday, for DragonFly, and for all other participating organizations. DragonFly has 5 slots, and more than 5 excellent proposals, which is a good kind of problem to have. We’ll see what we can fit.
There’s a new tool being put together for pkgsrc installation and management, called pkg_dry. There’s an initial version for download with instructions from its creator, Emile “iMil” Heitor. It looks to duplicate the functionality of apt-get or yum, by handling binary-only remote package management.
Someone please test this on DragonFly, though not on a production machine… If it does end up matching apt-get (the only thing I like in Debian) in terms of functionality, that will be fantastic! I have wanted something like this for a while.
Matthew Dillon is trying to track down a Hammer bug where directory entries (files, usually) are missed, whether it’s with ls or find or similar. Has this happened to you? It’s apparently very hard to duplicate, so please speak up if it has.
Can someone who uses git more heavily than I do look at Tig, a git viewer, and mention if it is useful? It looks interesting, as one of the features that git ‘lacks’ is a visual client other than at the command line.
Two recent pkgsrc changes that won’t directly affect how your packages are installed, but are still worth knowing about:
- Thomas Klausner made some adjustments to license handling for pkgsrc. It’s possible to prevent packages with or without certain licenses from installing, even as a dependency. His post contains details of the by-default accepted licenses.
- Joerg Sonnenberger has removed most instances of NO_CHECKSUM. This will help people with poor network connections, and (I would imagine) will be better for security, too.
My apologies, folks – the site is going up and down, along with my network connection. It’s probably because of Time Warner (my network provider) trying and failing to correctly implement bandwidth measuring tools for their upcoming volume cap. (I keep getting passed a private IP.)
If you’re a student with a Summer of Code application, make sure to subscribe to it. Doing this will ensure you are automatically notified of any mentor requests for more information.
There’s also some recent stats published by Google on the applications so far; DragonFly is one of the surveyed orgs it mentions, and the results are the same – less applications, better quality.
My plans for building binary packages for pkgsrc are detailed in this message to the mailing lists. No timeline yet…
The due date for the Summer of Code proposals is past, and DragonFly has 18 proposals. The consensus from other SoC organizations is the same: less applicants everywhere this year, but the proposal quality is up.
Potential mentors can now discuss the proposals and ask for more detail from the students, until April 15th.
19:00 UTC today is the deadline for all student applications for Google’s Summer of Code program. You can revise applications up to April 15th based on feedback, but the initial proposal has to be in the system as of tonight. That’s 5 hours from now, if I have my time calculations correct.
DragonFly has 15 applications at this point, and general application quality looks to be better this year than last.
I’m going to mesh together two unrelated items in this post:
The April OSBR is out, with this issue being a focus on Open APIs.
The newest @Play column covers winners of the 7DRL, or “7 Day Roguelike” contest, where contestants build a new roguelike game in a week. There are some real oddities, like Decimation. I’m not sure how many of these will build on DragonFly, darnit.
An interesting tidbit turned up by Google searches: the invention of ping, from the man who wrote it. The ping -> vocoder story near the end is entertaining.