Translation(s): English - Français
Debian > Debian GNU > Debian GNU/kFreeBSD > kFreeBSD Roadmap
Debian GNU/kFreeBSD is still actively being developed.
FTBFS in sid
We need to ensure up-to-date packages keep building in sid. All of these FTBFS would be RC bugs if kfreebsd were part of stretch.
I began a Ports Dashboard here that makes it easier from my point of view: https://kfreebsd.eu/dashboards/ports/
*buntuBSD
This derivative of kfreebsd has many users already, and they are reporting actionable bugs back to us. We should continue to support them, merge patches and provide features they ask for.
Upstreaming
I've set up CI of many upstream projects, checking the latest development version still builds on kfreebsd, so that we can notify upstream and fix issues before the package even enters Debian: http://jenkins.kfreebsd.eu/jenkins/view/upstream/
Debian carries a large kfreebsd patchset for openjdk-8, that need to go upstream, but needs significant cleanup first. Also some for GCC.
Updating to FreeBSD 10.3 kernel and userland
We must stop using FreeBSD 10.1 kernel because it will be end-of-life before stretch is even released. 10.3 is a long-term supported kernel upstream.
Most of the new 10.3 packages are ready in experimental.
The 10.3 kernel is working nicely on a few amd64 machines I've tested. I still need to test on i386.
I'd like to mass-rebuild many reverse dependencies of the 10.3 userland. In particular, things that use kernel interfaces (net-snmp? gtop?). I've already tested some reverse-deps of libusb.
We also wait for the upload of glibc/2.22-8 to sid before we can begin the transition.
Updating to FreeBSD 11.x kernel and userland
This may not happen in time for stretch.
If FreeBSD 11.0 is not a long-term supported kernel upstream, it would be unsuitable for stretch. We maybe should wait for 11.1, but stretch may be frozen by then or it would be too late for us to transition.
Hardening flags
The kernel can't build with stack-smashing protection yet.
The kernel of FreeBSD 10.x doesn't implement ASLR yet.
I'm trialling PIE+relro in the 10.3 userland packages now, starting with freebsd-libs, and then hopefully in all of kfreebsd's core packages in time for stretch. This leads to someday having ASLR.
Housekeeping
There's a lot of duplication in the debian/rules of kfreebsd's core packages. We should keep these in sync, remove redundant parts, and try to refactor.
Reproducible Builds
kfreebsd's core packages pass the reproducible-builds.org tests at the moment. But testing is only done on linux machines, we need to set up a kfreebsd machine to test the others.
Furthermore, I'd like the arch:all packages to be reproducible from linux and from kfreebsd. We may need new chmod or tar features for that.
Rebootstrapping
Rebootstrapping kfreebsd from linux: https://jenkins.debian.net/view/rebootstrap/
Rebootstrapping linux from kfreebsd: http://jenkins.kfreebsd.eu/jenkins/view/rebootstrap/
Currently we wait for glibc/2.22-8 to fix a regression, and then for 10.3 userland to enter sid.
We may be able to rebootstrap kfreebsd-i386_gcc5_nobiarch rather soon, but more work is needed for other host archictectures.
kfreebsd-armhf
Maybe coming soon...