It's mostly the same and the indirection through custom
is annoying to work with. Since it is just a helper for
the VM image type don't document it further.
Only VM images really handle ZFS. While here move the handling
to the core repository so rc.loader knows what to do.
We only need a 10 second boot relay on installer vga/serial/dvd/nano.
Also go back to setting the pool name using PRODUCT_ZFS which is
now safeguarded through ALLOWED/REQUESTED.
Also renames PRODUCT_SERVER to shorter SERVER because it's not really
part of the product. ;)
Allow the REMOTEDIR to be empty so we can automatically select a matching
directory in download/upload cases.
Suppressing errors is a bit risky. Now we require print to
only print existing variables and can check of the evaluation
into the composite script worked ok before continuing. Also
a bit easier to read.
Other composite scripts will follow as they are being used.
This simply scans ${PORTSDIR}/MOVED file for the port at hand.
Our skim doesn't delete used but removed ports but thus we also
fail to see if FreeBSD ports (re)moved it until it breaks for good.
So core and plugins are different from ports for historic reasons
and scope but acutally merge the code of both to make this more
obvious. Use run_stage exclusively even for the single-step runs
to make the whole thing consistent.
So the following now works:
# make hotfix-dhcp6c,os-ddclient-devel,opnsense
And rebuilds all packages using their matching steps and runs the
packages step at the end to provide a final packages set without
the aux packages.
I have been unintentionally scrubbing packages sets in the past which
is not a lot of fun so make the default hotfix mode a little less coarse.
While here the biggest issue of the hotfixing is not being able to see
package messages (not errors, these are shown as the build aborts) so
collect them in a variable to show at the end if all was building ok.
This is only relevant for multi-step hotfixing between core/plugins.
We do this to avoid /usr/obj longering causing it to be caught in
ZFS snapshots allegedly done by freebsd-update reducing our available
space ever so slightly until nothing works anymore.
We don't intend to pass down arguments to individual steps passed
down but if someone makes that mistake (like me) at least we don't
run the ports build step now which produces a bogus result.
