If you, like me, track the RELEASE versions of DragonFly, you’ll be moving from version 1.2 to 1.4 in one jump. There’s a few extra upgrade steps to accomodate the drastic underlying changes between 1.2 and 1.4, and I’ve documented them in the src/UPDATING file.
Csaba Henk has contributed code to make nullfs. It’s not complete yet, but it’s close.
Matthew Dillon added a note that describes the release creation process.
Better hope I’m talking about dump(8), eh? YONETANI Tomokazu found a problem in the way dump files were created after some changes were made in 1.3 development; the problem’s been fixed, but be warned: dumps from that time period won’t be compatible.
1.5 is now going to be the Experimental version of DragonFly BSD (or HEAD for people used to FreeBSD); see the commit.
If you’re running the experimental branch of DragonFly code and you (re)built your system in the last 12 hours, Simon ‘corecode’ Schubert has a quick fix for a typo that was present during that time.
Also, Simon has set up bridging between network cards, and posted instructions on how to set it up.
Simon ‘corecode’ Schubert has added example pf configs — just in case you need them.
Sascha Wildner will soon be (mostly?) completing the last of the changes from K&R to ANSI style C in the DragonFly source tree – not glamourous, but certainly good.
Simon ‘corecode’ Schubert (who has been very active with the commits lately) added CVS version 1.12.13. He noticed some troubles, though, but this version of CVS is staying for now. Matthew Dillon counseled some changes, and also noted that the upcoming DragonFly 1.4 release is next week. Next week!
Matthew Dillon commited changes to make the next release require pkgsrc.
The default “Waiting for SCSI devices to settle” delay has been reduced to 5 seconds.
Just to follow up on earlier threads: the first part of the multiprocessor-safe network interrupt code has gone in.
There’s been two updates on the DragonFly website: the 1.2 page now lists the changes in version 1.2.6, and the download page now lists the pkgsrc binary mirrors.
Matthew Dillon’s making some changes get get us just a little closer to removing the Big Giant Lock for multiprocessor systems; it’s now possible to treat certain interrupts, traps, and syscalls as mpsafe. mpsafe, for the acronymically challenged, is “multi-processor safe”.
Matthew Dillon’s tsleep patches are in, making another chunk of code multiprocessor safe. The commit message comes with a good bit of explanation, too.
If you have a multi-processor computer, Matthew Dillon would appreciate testing of his latest , more efficient changes to the scheduling system.
Matthew Dillon has created a way to run a SMP kernel without using the APIC_IO option. This lets his Shuttle XPCs boot. He also solved a BIOS problem on AMDX2 dual-core machines.
Matthew Dillon is planning some major work on the routing of interrupts; if you are running the latest code, report any new breakages to bugs@, please.