[Buildroot] WIP: glibc 2.16 bump

[Buildroot] WIP: glibc 2.16 bump

[Romain Naour]

Hi Thomas,

I started to look at glibc 2.26 bump.
All patches are uptream but the bump trigger some build issue with gcc 7 (older
gcc version are not tested yet).
For now I'm building only the toolchain on several arch, so no runtime test yet.
How the freeelectron farm can help to test new gcc/libc version ?

Best regards,
Romain

[Buildroot] WIP: glibc 2.16 bump

[Thomas Petazzoni]

Hello,

On Sat, 12 Aug 2017 12:24:50 +0200, Romain Naour wrote:

> I started to look at glibc 2.26 bump.

Good!

> All patches are uptream but the bump trigger some build issue with gcc 7 (older
> gcc version are not tested yet).
> For now I'm building only the toolchain on several arch, so no runtime test yet.
> How the freeelectron farm can help to test new gcc/libc version ?

If you submit the patch and/or push it to some branch somewhere, I can
apply it on a branch (not to the official Buildroot tree), and run a
test build of all the toolchains configurations, and give you the link
to the Gitlab CI build results.

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] WIP: glibc 2.16 bump

[Romain Naour]

Hi Thomas,

Le 12/08/2017 ? 13:46, Thomas Petazzoni a ?crit :
> Hello,
> 
> On Sat, 12 Aug 2017 12:24:50 +0200, Romain Naour wrote:
> 
>> I started to look at glibc 2.26 bump.
> 
> Good!
> 
>> All patches are uptream but the bump trigger some build issue with gcc 7 (older
>> gcc version are not tested yet).
>> For now I'm building only the toolchain on several arch, so no runtime test yet.
>> How the freeelectron farm can help to test new gcc/libc version ?
> 
> If you submit the patch and/or push it to some branch somewhere, I can
> apply it on a branch (not to the official Buildroot tree), and run a
> test build of all the toolchains configurations, and give you the link
> to the Gitlab CI build results.

Can you try with this branch ?
https://github.com/RomainNaour/buildroot/tree/glibc-2.26

Best regards,
Romain

> 
> Best regards,
> 
> Thomas
> 

[Buildroot] WIP: glibc 2.16 bump

[Romain Naour]

Hi Thomas,

Le 12/08/2017 ? 15:04, Romain Naour a ?crit :
> Hi Thomas,
> 
> Le 12/08/2017 ? 13:46, Thomas Petazzoni a ?crit :
>> Hello,
>>
>> On Sat, 12 Aug 2017 12:24:50 +0200, Romain Naour wrote:
>>
>>> I started to look at glibc 2.26 bump.
>>
>> Good!
>>
>>> All patches are uptream but the bump trigger some build issue with gcc 7 (older
>>> gcc version are not tested yet).
>>> For now I'm building only the toolchain on several arch, so no runtime test yet.
>>> How the freeelectron farm can help to test new gcc/libc version ?
>>
>> If you submit the patch and/or push it to some branch somewhere, I can
>> apply it on a branch (not to the official Buildroot tree), and run a
>> test build of all the toolchains configurations, and give you the link
>> to the Gitlab CI build results.
> 
> Can you try with this branch ?
> https://github.com/RomainNaour/buildroot/tree/glibc-2.26

I started a build on gcc farm and the only failing build is powerpc64le.

I added this hunk in gcc.mk

# Since Glibc 2.26, poerpc64le requires double/long double.
# See sysdeps/powerpc/powerpc64le/configure.ac
ifeq ($(BR2_TOOLCHAIN_USES_GLIBC)$(BR2_powerpc64le),yy)
HOST_GCC_COMMON_CONF_OPTS += \
	--with-long-double-128
endif

I updated glibc-2.26 branch on github.

Note: float128 require at least gcc 6.2 or gcc 7.1, so we need to disable older
gcc on powerpc64le with glibc.

Best regards,
Romain

> 
> Best regards,
> Romain
> 
>>
>> Best regards,
>>
>> Thomas
>>
> 

[Buildroot] WIP: glibc 2.16 bump

[Thomas Petazzoni]

Hello,

On Mon, 14 Aug 2017 15:04:10 +0200, Romain Naour wrote:

> > Can you try with this branch ?
> > https://github.com/RomainNaour/buildroot/tree/glibc-2.26  
> 
> I started a build on gcc farm and the only failing build is powerpc64le.
> 
> I added this hunk in gcc.mk
> 
> # Since Glibc 2.26, poerpc64le requires double/long double.
> # See sysdeps/powerpc/powerpc64le/configure.ac
> ifeq ($(BR2_TOOLCHAIN_USES_GLIBC)$(BR2_powerpc64le),yy)
> HOST_GCC_COMMON_CONF_OPTS += \
> 	--with-long-double-128
> endif
> 
> I updated glibc-2.26 branch on github.
> 
> Note: float128 require at least gcc 6.2 or gcc 7.1, so we need to disable older
> gcc on powerpc64le with glibc.

OK. I'm about to start a rebuild of all my toolchains + Qemu tests with
glibc 2.26 (and gcc 7.x, binutils 2.29, etc, i.e the bleeding edge
toolchains).

FYI: the two gcc patches that you needed to build glibc 2.26 are no
longer needed once you bump gcc to 7.2.0, which I have done for this
toolchain rebuild. I'll send the gcc 7.2.0 bump patch once all
toolchains have been rebuilt.

I'll let you know the result.

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] WIP: glibc 2.16 bump

[Romain Naour]

Hi Thomas,

Le 14/08/2017 ? 16:24, Thomas Petazzoni a ?crit :
> Hello,
> 
> On Mon, 14 Aug 2017 15:04:10 +0200, Romain Naour wrote:
> 
>>> Can you try with this branch ?
>>> https://github.com/RomainNaour/buildroot/tree/glibc-2.26  
>>
>> I started a build on gcc farm and the only failing build is powerpc64le.
>>
>> I added this hunk in gcc.mk
>>
>> # Since Glibc 2.26, poerpc64le requires double/long double.
>> # See sysdeps/powerpc/powerpc64le/configure.ac
>> ifeq ($(BR2_TOOLCHAIN_USES_GLIBC)$(BR2_powerpc64le),yy)
>> HOST_GCC_COMMON_CONF_OPTS += \
>> 	--with-long-double-128
>> endif
>>
>> I updated glibc-2.26 branch on github.
>>
>> Note: float128 require at least gcc 6.2 or gcc 7.1, so we need to disable older
>> gcc on powerpc64le with glibc.
> 
> OK. I'm about to start a rebuild of all my toolchains + Qemu tests with
> glibc 2.26 (and gcc 7.x, binutils 2.29, etc, i.e the bleeding edge
> toolchains).
> 
> FYI: the two gcc patches that you needed to build glibc 2.26 are no
> longer needed once you bump gcc to 7.2.0, which I have done for this
> toolchain rebuild. I'll send the gcc 7.2.0 bump patch once all
> toolchains have been rebuilt.
> 
> I'll let you know the result.

Ok perfect :)

Best regards,
Romain

> 
> Best regards,
> 
> Thomas
>