Date: Wed, 29 Oct 2003 22:28:28 -0700 (MST) From: "M. Warner Losh" <imp@bsdimp.com> To: nate@root.org Cc: cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/conf options src/sys/i386/acpica Makefile acpi_wakecode.S src/sys/i386/conf NOTES Message-ID: <20031029.222828.133432192.imp@bsdimp.com> In-Reply-To: <20031029210945.A86732@root.org> References: <20031029100945.F84859@root.org> <20031029.220629.102576808.imp@bsdimp.com> <20031029210945.A86732@root.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In message: <20031029210945.A86732@root.org>
Nate Lawson <nate@root.org> writes:
: On Wed, 29 Oct 2003, M. Warner Losh wrote:
: > In message: <20031029100945.F84859@root.org>
: > Nate Lawson <nate@root.org> writes:
: > : On Wed, 29 Oct 2003, M. Warner Losh wrote:
: > : > In message: <20031028235425.N82902@root.org>
: > : > Nate Lawson <nate@root.org> writes:
: > : > : Is this needed? options alone should be sufficient and acpi_wakeup.S
: > : > : should include opt_acpi.h.
: > : >
: > : > Not for kernel modules...
: > :
: > : Muagh.
: > :
: > : Ok, I'm starting a pot for whoever will finally implement the combined
: > : kernel modules and kernel config build tree. $10 US via PayPal. Care to
: > : contribute?
: >
: > Not for this. Options for kernel compilation should only really be
: > used for ABI issues. better to make this a tunable/sysctl.
:
: Non-sequitur.
Not a non-sequitur. We're moving away from kernel options for things
that are tunable. I'm pointing out that this is the case here and we
should recognize it. Your desire for opt_*.h stuff is the
non-sequitur :-).
: The issue still stands that module builds don't pull in the
: kernel config file, forcing duplication in make.conf.
Not entirely true. We tend to build modules that have all the options
turned on. The world view is that opt_foo is used for big options to
leave in/out, not change minor runtime behavor.
: When building a
: kernel+modules, options in opt_*.h should be included in the module build.
Agreed, it basically should. I'm just saying that this isn't a good
reason for it, and we should be fixing the right problem.
: This is even under the 5.2-TODO as "Revised kld build infrastructure". I
: was just indicating that the number of times this has bothered me was
: worth $10 to the author of the fix. Maybe it's bothered enough others
: that it's worth a few hundred $ aggregate. :)
:
: I'll up it to $20.
Patches have been posted for about 90% of the problem.
However, the last 10% is hard. How do you deal with people that build
the modules outside of the kernel tree, so there is no kernel opt_*.h
files. This happens for third parties with drivers, as well as
developers making changes and loading modules. That's why things
haven't been committed: there's not been a good way proposed with how
to deal with this part of the problem. You'd still have the problem
with acpi if I cd to sys/modules/acpi and type make. And that's why I
harped on this needing to be a tunable/sysctl in my last message.
Warner
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031029.222828.133432192.imp>