gcc-4.2.0 breakage on powerpc?

gcc-4.2.0 breakage on powerpc?

[Mikael Pettersson]

On Sat, 19 May 2007 01:39:43 +0200, Segher Boessenkool wrote:
>> and we'll end up
>> having major kernel releases which don't build on i386 with major gcc
>> releases, which isn't altogether desirable.
>
>Yeah, like 4.2.0 with powerpc.  Seems like no one tested it :-(

Details please. What exactly are the gcc-4.2.0 problems on powerpc?

/Mikael

Re: gcc-4.2.0 breakage on powerpc?

[Segher Boessenkool]

>>> and we'll end up
>>> having major kernel releases which don't build on i386 with major gcc
>>> releases, which isn't altogether desirable.
>>
>> Yeah, like 4.2.0 with powerpc.  Seems like no one tested it :-(
>
> Details please. What exactly are the gcc-4.2.0 problems on powerpc?

http://gcc.gnu.org/PR31490 .  Most stuff using named
sections won't build; that includes CONFIG_MODULES.
There is no viable workaround.  The problem is being
handled.  I'll post a patch for the kernel refusing
to build with this combination soon.


Segher

Re: gcc-4.2.0 breakage on powerpc?

[Clemens Koller]

Mikael Pettersson schrieb:
> On Sat, 19 May 2007 01:39:43 +0200, Segher Boessenkool wrote:
>>> and we'll end up
>>> having major kernel releases which don't build on i386 with major gcc
>>> releases, which isn't altogether desirable.
>> Yeah, like 4.2.0 with powerpc.  Seems like no one tested it :-(
> 
> Details please. What exactly are the gcc-4.2.0 problems on powerpc?

[embedded platform!]
Over here, the 2.6.21.1 Kernel _builds_ fine with gcc-4.2.0 on an mpc8540 e500 powerpc
(mpc8540_ads_defconfig) but it just crashes silently on boot. But that's IMHO a
DeviceTree/U-Boot issue I haven't solved yet.

$ uname -a
Linux ecam.anagramm.de 2.6.21-rc5-g9a5ee4cc #4 Mon Apr 2 21:31:53 CEST 2007 ppc e500 GNU/Linux
ARCH=ppc compiled with GCC-4.1.2 works fine.

Greets,

Clemens Koller
_______________________________
R&D Imaging Devices
Anagramm GmbH
Rupert-Mayer-Str. 45/1
81379 Muenchen
Germany

http://www.anagramm-technology.com
Phone: +49-89-741518-50
Fax: +49-89-741518-19

Re: gcc-4.2.0 breakage on powerpc?

[Segher Boessenkool]

>>>> and we'll end up
>>>> having major kernel releases which don't build on i386 with major 
>>>> gcc
>>>> releases, which isn't altogether desirable.
>>>
>>> Yeah, like 4.2.0 with powerpc.  Seems like no one tested it :-(
>>
>> Details please. What exactly are the gcc-4.2.0 problems on powerpc?
>
> http://gcc.gnu.org/PR31490 .  Most stuff using named
> sections won't build; that includes CONFIG_MODULES.
> There is no viable workaround.  The problem is being
> handled.  I'll post a patch for the kernel refusing
> to build with this combination soon.

[Cc:ing the PowerPC list]

Oh btw, most likely this is a problem on 64-bit only.


Segher

Re: gcc-4.2.0 breakage on powerpc?

[Segher Boessenkool]

> Over here, the 2.6.21.1 Kernel _builds_ fine with gcc-4.2.0 on an 
> mpc8540 e500 powerpc
> (mpc8540_ads_defconfig)

Yeah, this problem likely doesn't show up on 32-bit platforms.

> but it just crashes silently on boot. But that's IMHO a
> DeviceTree/U-Boot issue I haven't solved yet.

Quite possible.  If you have any further problems, take
it to the linuxppc-dev list please.


Segher

Re: gcc-4.2.0 breakage on powerpc?

[Mikael Pettersson]

On Sat, 19 May 2007 18:11:20 +0200, Segher Boessenkool wrote:
> >>>> and we'll end up
> >>>> having major kernel releases which don't build on i386 with major 
> >>>> gcc
> >>>> releases, which isn't altogether desirable.
> >>>
> >>> Yeah, like 4.2.0 with powerpc.  Seems like no one tested it :-(
> >>
> >> Details please. What exactly are the gcc-4.2.0 problems on powerpc?
> >
> > http://gcc.gnu.org/PR31490 .  Most stuff using named
> > sections won't build; that includes CONFIG_MODULES.
> > There is no viable workaround.  The problem is being
> > handled.  I'll post a patch for the kernel refusing
> > to build with this combination soon.
> 
> [Cc:ing the PowerPC list]
> 
> Oh btw, most likely this is a problem on 64-bit only.

I believe so too. PR31490 mostly talks about ppc64,
and I've built both ppc32 kernels and complex ppc32
user-space programs with gcc-4.2.0 without noticing
any regressions.

/Mikael