Robert P. J. Day wrote:
> On Tue, 30 Jan 2007, David Chinner wrote:
>
>> On Mon, Jan 29, 2007 at 06:24:29AM -0500, Robert P. J. Day wrote:
>
>>> user-configurable kernel setting, it should not have a
>>> configuration macro name starting with "CONFIG_" -- that should be
>>> reserved for configurable settings defined in Kconfig files.
>> Yup, and that's exactly what we use them for.
>>
>> However, we don't anyone to use them in mainline because they are
>> for debugging or features that aren't in mainline, but they are not
>> easily of ussefully separable from the core XFS code. Hence we use
>> CONFIG_XFS_??? options that you can't turn on in mainline to
>> enforce this - making them CONFIG_XFS_??? options means we don't
>> have to maintain different code in the dev tree and mainline and IMO
>> that is a Good Thing.
One thing that may not be crystal clear from the thread (?) is that if
you download xfs cvs, you will in fact find these in the Kconfig files:
config XFS_DEBUG
bool "XFS Debugging support (EXPERIMENTAL)"
depends on XFS_FS && EXPERIMENTAL
...etc
but they aren't pushed out to kernel.org...
David, is CONFIG_XFS_DEBUG really so bad to expose? CONFIG_XFS_TRACE
-is- pretty useless w/o KDB, though, IIRC...
I suppose it would be a real headache to remove the unused CONFIG_FOO
stuff from kernel.org and leave them in cvs...
In the end it's not the most pressing problem, although it adds to the
perception of swaths of dead code in xfs.
-Eric
|