Re: [oe-commits] Robert Yang : autogen-native: upgrade to 5.18.2

[Robert Yang <liezhi.yang@windriver.com>]

On 01/21/2014 07:56 AM, Richard Purdie wrote:
> On Mon, 2014-01-20 at 10:29 -0800, Darren Hart wrote:
>> Reverting the patch allows me to move forward, but yes, I'm surprised we
>> aren't seeing more build failures.... I wonder what makes us special?
>
> For some reason I still don't have the email I can see you sent from the
> list archives. What is interesting is that:
>
> a) you have gcc 4.6, I have gcc 4.8
> b) the error appears to come from error.h in SCM_NORETURN definitions
> c) the config log for autogen shows different results for
>     "checking for working stdnoreturn.h", yes in my case, no in yours.
>
> The definition of SCM_NORETURN is from __scm.h:
>
> #define SCM_NORETURN __attribute__ ((noreturn))
>
> and the dummy stdnoreturn.h that autogen would generate does:
>
> #define noreturn _Noreturn
>
> and then _Noreturn.h defines:
>
> #  define _Noreturn __attribute__ ((__noreturn__))
>
> so I think what is happening is there is double macro overlap with
>
> #define SCM_NORETURN __attribute__ ((noreturn))
>
> becoming
>
> #define SCM_NORETURN __attribute__ ((__attribute__ ((__noreturn__))))
>

Yes, indeeded, I removed the stdnoreturn.h from my host, then I can reproduce
the error, and the upstream has a patch to fix it:

http://git.savannah.gnu.org/gitweb/?p=guile.git;a=patch;h=36c40440078c005cd5e239cca487d29f6f60007d

I will send the PR sooner.

// Robert


> How we fix it I have no idea. We should check and see if someone fixed
> this upstream to start with. Robert, now we know what to look for, can
> you help?
>
> Cheers,
>
> Richard
>
>
>