Just a small question: are the VGA versions of the nanobsd supposed to be automatically built, or is that something you build manually for the stable releases, or is vga presence now automatically detected and is part of the regular nanobsd build anyway?
We have to set a flag in the builder config, it's an either-or build for nanobsd or nanobsd-vga, so we do one run of each for releases. There is no automatic detection.
The current builder doesn't support doing both in the same run. It could, but that would practically double the already loooooong amount of time it takes to do a snapshot run.
How about on odd dates a serial build, and on even days a VGA build?
That way the build time doesn't get longer, and we who require the VGA feature aren't left out of the fun, and at most someone has to wait one day to get updated build?
Just an idea ;)
Just not feasible. When we test NanoBSD, the gear we test with is primarily ALIX, so having to way for an alternate day or snapshot run to see if a test build works is out of the question, especially when it already can take 6-8 hours for a snapshot run to finish as-is.
Bummer. I guess someone needs to donate some SSDs to speed up the build times, presuming disk speed is the bottleneck…
...and/or come up with a smart idea like an autodetecting kernel that makes two separate builds superfluous....
...and/or find a way to do it in a single build in parallel, i.e. build everything at once, and then just package up the different components, which should differ only minimally, and output two packages. That would at least mean all the source is only compiled once, the linker would have to run twice on some parts, and the disk image would have to be packaged up twice
Kind of bad when a difference that small requires an entire 6-8h build run :(
Anyway, thanks for the answer.
[Split this off into its own topic]
The servers are plenty fast, it's just that all of the IO operations required to create and duplicate nanobsd images are very high. It would probably kill an SSD even over time.
There are too many things that differ to do them both in one run effectively, namely it would have to build two completely different sets of kernels and other various adjustments that need to happen along the way. There is no way a single image could ever be "automatic" to do this, some hardware will bomb if it tries to load a kernel with VGA in it, there's no working around that.
Even if we altered/injected the kernels into NanoBSD images it's still quite a bit of duplication of files and such. That's what we do for the serial memstick image.
If someone wants to tinker with the builder code and find a way to do it, more power to them, have at it. The current NanoBSD+VGA code was contributed or it may still not exist.
Is it possible to upgrade from a 2.0.1 VGA nanobsd to a 2.1 serial nanobsd?
Is it possible to later upgrade from a pre-release 2.1 serial nanobsd to a release 2.1 VGA kernel?
If so, I could still test 2.1, because I really only need the console for the initial setup, after that, I have ssh and the web interface, but I would have to have a way to revert back to a VGA
I assume the config files are not downward compatible, such that going from a 2.1 prerelease back to a 2.0.1 release wouldn't work, right?
Yes/Correct on all counts.
If i create wrap_vga kernel and replace it in latest snapshot img and change bootloader from boot0sio to boot0
will vga work?