[Buildroot] [PATCH v4 3/3] core/pkg-kconfig: Use oldnoconfig when possible

[Romain Izard]

2016-04-19 19:58 GMT+02:00 Yann E. MORIN <yann.morin.1998 at free.fr>:
> Romain, All,
>
> On 2016-04-19 19:40 +0200, Yann E. MORIN spake thusly:
>> On 2016-04-19 11:11 +0200, Romain Izard spake thusly:
>> > 2016-04-18 23:08 GMT+02:00 Yann E. MORIN <yann.morin.1998 at free.fr>:
> [--SNIP--]
>> > > What I would have loved however, would be a way to detect whether
>> > > oldnoconfig (better, olddefconfig) was supported before calling it.
>> > > Maybe something along the lines of:
>> > >
>> > >     if [ -n "`$$($(2)_KCONFIG_MAKE) -pqn .DEFAULT 2>/dev/null |grep -E '^olddefconfig'`" ]; then \
>> > >         $$($(2)_KCONFIG_MAKE) olddefconfig || exit 1; \
>> > >     elif [ -n "`$$($(2)_KCONFIG_MAKE) -pqn .DEFAULT |grep -E '^oldnoconfig'`" ]; then \
>> > >         $$($(2)_KCONFIG_MAKE) oldnoconfig || exit 1; \
>> > >     else
>> > >         (yes "" | $$($(2)_KCONFIG_MAKE) oldconfig) || exit 1; \
>> > >     fi
>> > >
>> > > Yes, this is more code, less compact, but it is more systematic. At
>> > > the very least, it would not fallback to the yes|oldconfig trick in
>> > > case old{def,no}config do exist but fail, which we do want to catch.
>> > >
>> > > BTW, the make trick above is highly (as in: almost identical) from the
>> > > bash-autocompletion package.
>> >
>> > I see the idea, but the specific implementation is not working: the
>> > kernel's Makefile targets are dynamic, and the targets we want to check
>> > are only present when the target we try to build matches %config.
>>
>> I just checked, and it works on the current master on the kernel tree:
>>
>>     $ make -pqn .DEFAULT 2>/dev/null |grep -E '^olddefconfig:'
>>     olddefconfig: scripts/kconfig/conf
>>
>>     $ make -pqn .DEFAULT 2>/dev/null |grep -E '^oldnoconfig:'
>>     oldnoconfig: olddefconfig
>
> I also checked on v3.6 (which is missing olddefconfig) and v3.7 (which
> introduced olddefconfig), and the tests above still work: oldnconfig is
> found, olddefconfig is not.
>
When I tried it yesterday it did not work, but it works today. I must have done
something wrong when testing it.

Best regards,
-- 
Romain Izard