Fix LOCAL_MODULES and improve the make output.

[jhb]

Summary

The exists() check guarding the invocation of ls was not working
correctly as it was expanding '$L' to determine the path of the
local modules directory. Fix by using {} around the variable
name.

Inline some of the logic from bsd.subdir.mk when invoking local
module builds. This gives output in 'make buildkernel' the same
as if there was a Makefile in /usr/local/sys/modules with
SUBDIR = ${LOCAL_MODULES}.

Test Plan

• tested with drm-current-kmod patched to install sources

[jhb]

Sample output from the tail of a NO_KERNELCLEAN=yes build with nothing to do:

===> x86bios (all)
===> xl (all)
===> xz (all)
===> zfs (all)
===> zlib (all)
===> drm-current-kmod (all)
===> drm-current-kmod/linuxkpi (all)
===> drm-current-kmod/ttm (all)
===> drm-current-kmod/drm (all)
===> drm-current-kmod/vmwgfx (all)
===> drm-current-kmod/i915 (all)
===> drm-current-kmod/amd (all)
===> drm-current-kmod/amd/amdgpu (all)
===> drm-current-kmod/amd/amdkfd (all)
===> drm-current-kmod/radeon (all)
--------------------------------------------------------------
>>> Kernel build for GENERIC completed on Thu Jul 18 15:06:31 PDT 2019
--------------------------------------------------------------

[jhb]

@imp any thoughts?

[imp]

sys/conf/kern.post.mk

80 ↗

(On Diff #59902)

Just a question: do you know why we do this substitution? I see you followed the same convention as other places, so it's not wrong wrt this commit, but was wondering if you knew the history.

[jhb]

sys/conf/kern.post.mk

80 ↗

(On Diff #59902)

Hmm, clobber => cleandir is from https://svnweb.freebsd.org/base?view=revision&revision=111802. I suspect we can just kill clobber entirely. It has no replacement in the buildkernel/installkernel wrappers. Peter added it way back in r73133. The reinstall => install thing dates back to when modules were added to kernel builds and exists because reinstall isn't a valid target for bsd.kmod.mk. Only kernels need reinstall to avoid renaming the kernel to kernel.old, etc.