Re: [PATCHv2] Kbuild: kconfig: Verbose version of --listnewconfig
Hi,
On Sat, Dec 4, 2010 at 4:53 PM, Ben Hutchings <ben@decadent.org.uk> wrote:
> On Sat, 2010-12-04 at 16:14 -0500, Arnaud Lacombe wrote:
>> Hi,
>>
>> On Sat, Dec 4, 2010 at 4:07 PM, Ben Hutchings <ben@decadent.org.uk> wrote:
>> [...]
>> >> >> I would not expect to see any header if there is no new symbol(s).
>> >> >> However, that might complicate the code too much. Btw, I find
>> >> >> "Changes" to be misleading, is that header necessary ?
>> >> >
>> >> > We use this feature (or an earlier version of it) in automated kernel
>> >> > builds in Debian, so we expect the output to appear in build logs and
>> >> > the header makes it easier to pick out.
>> >> >
>> >> That's easily doable outside kconfig.
>> >
>> > No it isn't, as the Kconfig code must be built as part of the
>> > listnewconfig target. (Building just the code first will provoke
>> > warnings about the invalid config.)
>> >
>> echo "#"
>> echo "# Changes:"
>> echo "#"
>> make [...] listnewconfig
>>
>> does the job, without introducing extra header others do not seem to
>> want to see.
>
> Now try that with a clean tree and an outdated config file.
>
make -s listnewconfig
is perfectly silent[0]. Ok, I admit, I had to cheat, ie. I built
kconfig with HOST_EXTRACFLAGS="-Wno-char-subscripts". You may also
want to use `silentoldconfig' beforehand to tell you if the config
needs adjustment or not, in which case, kconfig no longer has to be
built.
Now, if by "warnings about the invalid config" you mean the illegal
use of the "select" keyword, I'd say that these warnings should not
happen in a release kernel, and time has better to be spent fixing
them than putting some marker to skip them. Note that this task is up
to the subsystem maintainer, in most case.
- Arnaud
[0]: on top of Linus' tree as of Nov. 5.
Reply to: