linux-yocto and gcc 5.x

linux-yocto and gcc 5.x

[Richard Purdie]

Hi Bruce,

One of the things we're looking at in 1.9 is gcc 5.x support by default.
One of the areas this has significant impact is on the kernels,
linux-yocto in particular. I'm starting to see a potential problem here
since the kernels don't appear to work well with gcc 5.x:

http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE

(this is a test build I ran just to see how bad things were).

Are we going to be in a position to get 5.x support backported into 3.14
(and maybe 3.10)? Or was we going to have to wait for later versions?

Cheers,

Richard

Re: linux-yocto and gcc 5.x

[Bruce Ashfield]

On 2015-05-11 09:20 AM, Richard Purdie wrote:
> Hi Bruce,
>
> One of the things we're looking at in 1.9 is gcc 5.x support by default.
> One of the areas this has significant impact is on the kernels,
> linux-yocto in particular. I'm starting to see a potential problem here
> since the kernels don't appear to work well with gcc 5.x:
>
> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
>
> (this is a test build I ran just to see how bad things were).
>
> Are we going to be in a position to get 5.x support backported into 3.14
> (and maybe 3.10)? Or was we going to have to wait for later versions?

I should be able to bring gcc5 support back to the previous LTSI
kernel (3.14), but need to poke at the breakage myself to know
for sure.

What's the incantation to configure gcc5 into a test build ?

Bruce

>
> Cheers,
>
> Richard
>
>
>

Re: linux-yocto and gcc 5.x

[Richard Purdie]

On Mon, 2015-05-11 at 09:51 -0400, Bruce Ashfield wrote:
> On 2015-05-11 09:20 AM, Richard Purdie wrote:
> > Hi Bruce,
> >
> > One of the things we're looking at in 1.9 is gcc 5.x support by default.
> > One of the areas this has significant impact is on the kernels,
> > linux-yocto in particular. I'm starting to see a potential problem here
> > since the kernels don't appear to work well with gcc 5.x:
> >
> > http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
> >
> > (this is a test build I ran just to see how bad things were).
> >
> > Are we going to be in a position to get 5.x support backported into 3.14
> > (and maybe 3.10)? Or was we going to have to wait for later versions?
> 
> I should be able to bring gcc5 support back to the previous LTSI
> kernel (3.14), but need to poke at the breakage myself to know
> for sure.

Sorry, I was confused with versions, I of course mean 3.17 and 3.14
above, we dropped 3.10! I think 3.17 is ok so its 3.14 that we're having
the problems with by the looks of it.

> What's the incantation to configure gcc5 into a test build ?

http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=d72ca3dca225a4801e595c95e98aacbe738d8303

or just set

GCCVERSION = "5.1%"

Cheers,

Richard

Re: linux-yocto and gcc 5.x

[Khem Raj]

On Mon, May 11, 2015 at 6:20 AM, Richard Purdie
<richard.purdie@linuxfoundation.org> wrote:
>
> One of the things we're looking at in 1.9 is gcc 5.x support by default.
> One of the areas this has significant impact is on the kernels,
> linux-yocto in particular. I'm starting to see a potential problem here
> since the kernels don't appear to work well with gcc 5.x:
>
> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
>

as mentioned in another thread. linux-yocto needs to backport
https://patchwork.ozlabs.org/patch/403787/
to beging with. I have 3.8 running on beaglebone white but thats not linux-yocto

https://github.com/koenkooi/meta-beagleboard/commit/07e27c22e707e26f3f321f51999c4f6a294b5b18
https://github.com/koenkooi/meta-beagleboard/commit/765e7f35af7e4358573868389d4ea144eacd6b7a

there is a patch or two in there that needs to go upstream into kernel
as well but I havent bothered yet.

I was not able to boot linux yocto 3.19 few weeks ago on any of
emulators when build with gcc5
rootfs did not have any issues.

I also have 3.10 compiling and booting on another Arm platform


> (this is a test build I ran just to see how bad things were).
>
> Are we going to be in a position to get 5.x support backported into 3.14
> (and maybe 3.10)? Or was we going to have to wait for later versions?
>
> Cheers,

Re: linux-yocto and gcc 5.x

[Robert Yang]

I think that I met another problem when MACHINE = qemuarm64 with gcc5:
linux-yocto_3.19.bb:

/tmp/ccWk2lM2.s: Assembler messages:
/tmp/ccWk2lM2.s:42: Error: .err encountered
/tmp/ccWk2lM2.s:43: Error: .err encountered
/tmp/ccWk2lM2.s:44: Error: .err encountered
/tmp/ccWk2lM2.s:103: Error: .err encountered
/tmp/ccWk2lM2.s:104: Error: .err encountered
/tmp/ccWk2lM2.s:105: Error: .err encountered
make[3]: *** [arch/arm64/kernel/psci.o] Error 1
make[3]: *** Waiting for unfinished jobs....
   CC      mm/mmzone.o
   CC      mm/vmstat.o
   CC      mm/backing-dev.o
   CC      net/socket.o
make[2]: *** [arch/arm64/kernel] Error 2
make[2]: *** Waiting for unfinished jobs....

// Robert

On 05/11/2015 09:20 PM, Richard Purdie wrote:
> Hi Bruce,
>
> One of the things we're looking at in 1.9 is gcc 5.x support by default.
> One of the areas this has significant impact is on the kernels,
> linux-yocto in particular. I'm starting to see a potential problem here
> since the kernels don't appear to work well with gcc 5.x:
>
> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
>
> (this is a test build I ran just to see how bad things were).
>
> Are we going to be in a position to get 5.x support backported into 3.14
> (and maybe 3.10)? Or was we going to have to wait for later versions?
>
> Cheers,
>
> Richard
>
>
>

Re: linux-yocto and gcc 5.x

[Bruce Ashfield]

I'll do an explicit round of gcc5 support in master for the
kernels we have there in the not to distant future. To throw
the switch for gcc5, we'll obviously want all the architectures
to build and boot.

To keep things practical, I'll start from the -dev kernel, then
back to the LTSI kernels we still have in master.

In the meantime, if anyone has patches, sending them to the linux-yocto
mailing list is appreciated. Since everyone has eyes and ears
on different lists it makes the effort that much easier if we collect
changes in advance on a single list.

Bruce

On 2015-05-13 03:39 AM, Robert Yang wrote:
>
> I think that I met another problem when MACHINE = qemuarm64 with gcc5:
> linux-yocto_3.19.bb:
>
> /tmp/ccWk2lM2.s: Assembler messages:
> /tmp/ccWk2lM2.s:42: Error: .err encountered
> /tmp/ccWk2lM2.s:43: Error: .err encountered
> /tmp/ccWk2lM2.s:44: Error: .err encountered
> /tmp/ccWk2lM2.s:103: Error: .err encountered
> /tmp/ccWk2lM2.s:104: Error: .err encountered
> /tmp/ccWk2lM2.s:105: Error: .err encountered
> make[3]: *** [arch/arm64/kernel/psci.o] Error 1
> make[3]: *** Waiting for unfinished jobs....
>    CC      mm/mmzone.o
>    CC      mm/vmstat.o
>    CC      mm/backing-dev.o
>    CC      net/socket.o
> make[2]: *** [arch/arm64/kernel] Error 2
> make[2]: *** Waiting for unfinished jobs....
>
> // Robert
>
> On 05/11/2015 09:20 PM, Richard Purdie wrote:
>> Hi Bruce,
>>
>> One of the things we're looking at in 1.9 is gcc 5.x support by default.
>> One of the areas this has significant impact is on the kernels,
>> linux-yocto in particular. I'm starting to see a potential problem here
>> since the kernels don't appear to work well with gcc 5.x:
>>
>> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
>>
>>
>> (this is a test build I ran just to see how bad things were).
>>
>> Are we going to be in a position to get 5.x support backported into 3.14
>> (and maybe 3.10)? Or was we going to have to wait for later versions?
>>
>> Cheers,
>>
>> Richard
>>
>>
>>

Re: linux-yocto and gcc 5.x

[akuster808]

Robert,

this was the error I hit last week. There may be a fix I have not tried.

https://lkml.org/lkml/2015/2/18/422

- armin

On 05/13/2015 12:39 AM, Robert Yang wrote:
>
> I think that I met another problem when MACHINE = qemuarm64 with gcc5:
> linux-yocto_3.19.bb:
>
> /tmp/ccWk2lM2.s: Assembler messages:
> /tmp/ccWk2lM2.s:42: Error: .err encountered
> /tmp/ccWk2lM2.s:43: Error: .err encountered
> /tmp/ccWk2lM2.s:44: Error: .err encountered
> /tmp/ccWk2lM2.s:103: Error: .err encountered
> /tmp/ccWk2lM2.s:104: Error: .err encountered
> /tmp/ccWk2lM2.s:105: Error: .err encountered
> make[3]: *** [arch/arm64/kernel/psci.o] Error 1
> make[3]: *** Waiting for unfinished jobs....
>    CC      mm/mmzone.o
>    CC      mm/vmstat.o
>    CC      mm/backing-dev.o
>    CC      net/socket.o
> make[2]: *** [arch/arm64/kernel] Error 2
> make[2]: *** Waiting for unfinished jobs....
>
> // Robert
>
> On 05/11/2015 09:20 PM, Richard Purdie wrote:
>> Hi Bruce,
>>
>> One of the things we're looking at in 1.9 is gcc 5.x support by default.
>> One of the areas this has significant impact is on the kernels,
>> linux-yocto in particular. I'm starting to see a potential problem here
>> since the kernels don't appear to work well with gcc 5.x:
>>
>> http://errors.yoctoproject.org/Errors/Search/?items=10&query=1426e31f0bcd19f066931d2ecbdec3752b14e88e&limit=150&order_by=RECIPE
>>
>>
>> (this is a test build I ran just to see how bad things were).
>>
>> Are we going to be in a position to get 5.x support backported into 3.14
>> (and maybe 3.10)? Or was we going to have to wait for later versions?
>>
>> Cheers,
>>
>> Richard
>>
>>
>>