[oe] [PATCH][v3 3/4] uClibc: redo configuration

Bernhard Reutner-Fischer rep.dot.nop at gmail.com
Thu Jul 8 20:44:53 UTC 2010


On Thu, Jul 08, 2010 at 01:31:51PM -0700, Khem Raj wrote:
>On Thu, Jul 8, 2010 at 1:28 PM, Bernhard Reutner-Fischer
><rep.dot.nop at gmail.com> wrote:
>> On Thu, Jul 08, 2010 at 12:02:51PM -0700, Khem Raj wrote:
>>>On Tue, Jul 6, 2010 at 12:32 PM, Bernhard Reutner-Fischer
>>><rep.dot.nop at gmail.com> wrote:
>>
>>>> +       uclibc_cfg('bx',        machine_features,'USE_BX', cnf, rem)
>>>
>>>
>>>
>>>I would have a feature called 'nobx' because most of the machines
>>>we support actively in OE has BX. With current state this needs to
>>>be added to MACHINE features for almost all arm machines. We can instead
>>>have nobx and keep USE_BX enabled and only disable it if nobx appears
>>>in machine_features.
>>
>> I'd much prefer to stick bx into some generic arm include instead.
>> What do you think?
>
>I have added attached patch and turned bx into nobx and now all seems to be
>ok

This is conceptionally backwards but i won't complain if this helps in
finally getting the improved infrastructure in.

Separate issues, but what about the other parts below?
thanks,
>>
>>>                     Otherwise I have tested the patches myself
>>>and they seems to work well.
>>>
>>>I will push them once we have the bx turned into nobx
>>>
>>>I think you do not have a function to disable already enabled feature.
>>
>> the 'rem' below translates into sed /SOMETHING/d
>> Can you elaborate?
>> Except you mean DISTRO_FEATURES="ipv4"
>> and one of the gazillion distro-specific config snippets contain
>> CONFIG_IPV6=y
>> and that is not disabled.. About all those distro hunks have to go
>> anyway for they are redundant.
>>
>>>> +       return "\n".join(cnf), "\n".join(rem)
>>>> +# X, Y = ${@features_to_uclibc_settings(d)}
>>>> +# unfortunately doesn't seem to work with bitbake, workaround:
>>
>> you said you saw an ICE with current uclibc master. Which arch and which
>> versions of the toolchain did you use?
>> FYI, i tested qemuarm, qemumips, qemuppc, qemux86 and (my local)
>> qemux86_64 with
>> PREFERRED_BINUTILS = "2.20.1" ¹)
>> PREFERRED_GCC_VERSION = "4.5" ²) ³)
>>
>> and all of built fine for me.
>>
>> ¹) micro.conf incorrectly picks 2.19.51 per default, should be bumped
>> ²) micro.conf should also default to 4.5
>> ³) It would be great if we could settle on one scheme for
>>   binutils/gcc/libc version shorthands:
>>   PREFERRED_BINUTILS_VERSION _or_ PREFERRED_GCC, i'd go for
>>   PREFERRED_foo_VERSION, uniformly.
>>






More information about the Openembedded-devel mailing list