Hi!
As a CI project, I'm doing quite a lot of automated NetBSD builds,
basically calling build.sh for
tools/libs/release and {iso,install,live}-image . That's done on a
quite recent NetBSD -current VM (auto-installed from the install ISO
from the previous round) as well as in a Linux-based Docker container
running `sid` Debian userland.
I noticed that some in machine/architecture combinations (e.g.
evbarm with earmv4eb) the install-image doesn't build as it tries to
fetch kern-GENERIC (which isn't built.)
Haven't yet tried, but would I need to also supply `kernel=<foo>` in
that case? Or something else (`releasekernel=<foo>` or `kernels` or
something like that?) Or should these ports/archs set a proper kernel
on their own?
(On the larger scale, that is to test reproducibility, and finally
reach a state where a Linux-based cross-build consistently and
continuously monitored produces the very same release artifacts as
does a NetBSD native build. The last breakage in that field was
non-matching INSTALL.ps files where a Linux-based build produces A4
while NetBSD produces US Letter output.)
Thanks,
Jan-Benedict
--
Attachment:
signature.asc
Description: PGP signature