[Buildroot] [autobuild.buildroot.net] Build results for 2016-05-23

[Gustavo Zacarias <gustavo@zacarias.com.ar>]

On 24/05/16 10:36, Vicente Olivert Riera wrote:

> Hello everyone,
>
> On 24/05/16 07:30, Thomas Petazzoni wrote:
>> ===================
>>
>>           arm |                 collectd-5.5.1 | NOK | http://autobuild.buildroot.net/results/24a35b63921cb9d747052f3bbfc3549ed20e18d7/
>>      mips64el |                 collectd-5.5.1 | NOK | http://autobuild.buildroot.net/results/af1eec00e472349c7dca76dcc8dd5cd37d82f5c3/
>>         nios2 |                   connman-1.32 | NOK | http://autobuild.buildroot.net/results/af3743bc0840cecf13664c081d6d205e5a8c9eca/
>>      mips64el |                   connman-1.32 | NOK | http://autobuild.buildroot.net/results/6593e1ca94f98e3c2a5fd90401f709cf2f7b1743/
>>           arm |                   connman-1.32 | NOK | http://autobuild.buildroot.net/results/5937a416c1f320a581b54a7bc61cbb69075e9e43/
>>        mipsel |                   connman-1.32 | NOK | http://autobuild.buildroot.net/results/5ff215f7c91411fef335890d8a86a3482dfdac3b/
>
> I've been looking at these ones and they fail due to some changes made
> in the Linux kernel that causes conflicts between Linux's linux/if.h and
> glibc's net/if.h.
>
> The failure looks like this in collectd's config.log:
>
> linux/if.h:71:2: error: redeclaration of enumerator 'IFF_UP'
>        IFF_UP    = 1<<0,  /* sysfs */
>
> net/if.h:44:5: note: previous definition of 'IFF_UP' was here
>        IFF_UP = 0x1,  /* Interface is up.  */
>
> And like this on conman:
>
> linux/if.h:86:2: error: redeclaration of enumerator 'IFF_DYNAMIC'
>    IFF_DYNAMIC   = 1<<15, /* sysfs */
>
> net/if.h:79:5: note: previous definition of 'IFF_DYNAMIC' was here
>       IFF_DYNAMIC = 0x8000 /* Dialup device with changing addresses.  */
>
> This problem has been already fixed upstream:
>
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=4a91cb61bb995e5571098188092e296192309c77
>
> Anyone has an idea how to workaround this downstream?

Hi Vicente.
Yes, don't use 4.5.0 - 4.5.4 headers, it's already fixed by dropping the 
connman workaround patch which isn't compatible with the proper fix, 
Thomas rebuilding the test toolchains and bumping the headers to 4.5.5.
As for covering the broken headers scenario any patch that's created for 
that is unlikely to be accepted upstream, and any toolchain that ships 
with those broken headers could be called "broken".
Regards.