[OE-core] [RFT] eglibc 2.19

[Khem Raj]

Hi Richard, All

On Feb 12, 2014, at 10:20 AM, Khem Raj <raj.khem at gmail.com> wrote:

> 
> On Feb 12, 2014, at 7:41 AM, Richard Purdie <richard.purdie at linuxfoundation.org> wrote:
> 
>> On Sat, 2014-02-08 at 08:08 -0800, Khem Raj wrote:
>>> Hi all
>>> 
>>> I have staged a branch with eglibc 2.19 upgraded recipes here
>>> 
>>> http://git.openembedded.org/openembedded-core-contrib/log/?h=kraj/eglibc-2.19
>>> 
>>> These are not final recipes there may be tweaks needed like adjusting SRC_URI
>>> later on.
>>> 
>>> please help testing it out in your environments and machines
>>> I would be happy to help out with issues we find
>> 
>> The autobuilder is still going but it found:
>> 
>> http://autobuilder.yoctoproject.org/main/builders/nightly-fsl-ppc/builds/25/steps/BuildImages/logs/stdio
> 
> I could reproduce it.
> 
> This seems to be surfaced by a commit in glibc where the  math_ldbl.h implementations for ppc were unified. I have for now reverted it via patch and pushed to pull branch. however thats not only the one issue :(. I am now seeing
> 
> | ../sysdeps/powerpc/fpu/fenv_libc.h:48:6: error: impossible constraint in 'asm'
> |       asm volatile ("mtfsf 0xff,%0" : : "f" (d)); \
> |       ^
> 

Alright, the real problem was that in glibc 2.19 the SPE patch was upsteamed from eglibc and now the SPE support lives under nofpu/ which is correct unlike
eglibc where it was living under fpu/ class. We were not passing —without-fp
to eglibc configure which caused this issue. I have now fixed it in metadata
and pushed a new set of patches ( total 2) to the pull branch, which should take of this issue as well as valgrind issue. 

Thanks
-Khem