[Bug 294878] graphics/drm-61-kmod failed update (15-STABLE)

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=294878

--- Comment #3 from Chad Jacob Milios <[email protected]> ---
(In reply to Chad Jacob Milios from comment #1)

oh, it's because the inclusion of compiler_flags.mk now `CWARNFLAGS+=
something` before kern.mk has had its chance to `CWARNFLAGS?= -Wthis -Wthat`.
note that even CWARNEXTRA begins there with ?= so we may need to replace our
early manipulations of CWARNFLAGS with the setting of a new variable which then
appends to CWARNFLAGS only immediately after each instance of `.include
<bsd.kmod.mk>` (unless just reordering the includes themselves is trivial)

this affects all modules except dummygfx and amdkfd (because they didnt gain
the inclusion of compiler_flags.mk). on the other hand, ttm and drm seem to
already have had a similar issue (masking system CWARNFLAGS by setting it too
soon) in the earlier _7 distribution, but i dont believe i encountered any ill
effect because of it

if kern.mk is allowed to initialize CWARNFLAGS then my patch-i915_Makefile is
probably unnecessary. on the other hand maybe it is beneficial to be clearing
out the system provided list if we really now have a complete and more surgical
per-compiler, per-module, per-file lists, as it seems we might (or are darn
close). so if we go this route we should add an explicit CWARNFLAGS= atop
compiler_flags.mk with a comment that it precludes its population from kern.mk
(and also include it, either the empty variable assignment or the inclusion of
compiler_flags.mk, into amdkfd and dummygfx for good measure before seeing if
we need to add any back in there from the system list) because the use of += to
set it in compiler_flags.mk and some module Makefiles (drm, ttm) could be
misleading where it happens before <bsd.kmod.mk>

-- 
You are receiving this mail because:
You are the assignee for the bug.