Build-Target failed in linux-header (stage 0). Fixed by modifying base/linux24/linux-conf.in file

[Mike N]

Hello,

I have used SVN t2-trunk version for a while already, but today it broke when I updated latest changes from svn repo.

After the update every build failed in "base/linux-header" stage-0 step. After digging in more it turn out that compilation environment tried to call Linux Kernel makefiles with "make ARCH=x86 CROSSCOMPILE= KCC=kcc oldconfig" parameters.

Well, as we know Linux 2.6.x doesn't have arch/x86/ directory but instead arch/i386/. The error turn out to be in "base/linux24/linux-conf.in" file. The latest SVN changelog (r27693 by aldas) gives a hint about the change but looks like it breaks the x86 (32bit) compilation.

The lastest T2 svn package/base/linux24/linux-conf.in file has following line 20:
lx_cpu=echo $arch | arch2uname | sed s/.... ppc and mips and bfin stuff here... -e  s/i386/x86

The last "-e s/i386/x86" part is new addition and it wrongly changes arch kernel compilation parameter from i386 to x86 value.

I removed this i386->x86 sed conversion part from linux-conf.in file and I was once again able to compile something.

Is this "workaround fix" valid or have I missed something? Or maybe the latest SVN change was meant to deal with only x86_64 arch but it accidentally changes the 32-bit x86 Kernel arch parameter?

Best regards,
Mike

[Rene Rebe]

Hi,

On 28.01.2008, at 20:07, Mike N wrote:

> Hello,
>
> I have used SVN t2-trunk version for a while already, but today it  
> broke
> when I updated latest changes from svn repo.
>
> After the update every build failed in "base/linux-header" stage-0  
> step.
> After digging in more it turn out that compilation environment tried  
> to call
> Linux Kernel makefiles with "make ARCH=x86 CROSSCOMPILE= KCC=kcc  
> oldconfig"
> parameters.
>
> Well, as we know Linux 2.6.x doesn't have arch/x86/ directory but  
> instead
> arch/i386/. The error turn out to be in "base/linux24/linux-conf.in"  
> file.
> The latest SVN changelog (r27693 by aldas) gives a hint about the  
> change but
> looks like it breaks the x86 (32bit) compilation.

This was changed in 2.6.24. i386 and x86_64 where unified to just x86  
with
a Config option whether to build 32 or 64 bit.

However, this changes apparently broken the slightly older linux-header
package which still has i386 and x86_64.

> The lastest T2 svn package/base/linux24/linux-conf.in file has  
> following
> line 20:
> lx_cpu=echo $arch | arch2uname | sed s/.... ppc and mips and bfin  
> stuff
> here... -e  s/i386/x86
>
> The last "-e s/i386/x86" part is new addition and it wrongly changes  
> arch
> kernel compilation parameter from i386 to x86 value.
>
> I removed this i386->x86 sed conversion part from linux-conf.in file  
> and I
> was once again able to compile something.
>
> Is this "workaround fix" valid or have I missed something? Or maybe  
> the
> latest SVN change was meant to deal with only x86_64 arch but it
> accidentally changes the 32-bit x86 Kernel arch parameter?

This needs to be addressed, maybe by simply updating the older linux-
header,
or changing this line to use different substitutes depending on the  
version
to build.

Yours,

--
   René Rebe - ExactCODE GmbH - Europe, Germany, Berlin
   Geschäftsführer: Susanne Klaus, René Rebe
   Sitz: Berlin, Amtsgericht Charlottenburg HRB 105 123 B
   USt-IdNr.: DE251602478
   http://exactcode.de | http://t2-project.org | http://rene.rebe.name


-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
lists@... with a subject of: unsubscribe t2

[Aldas Nabazas-2]

Hi,

Sorry Mike for the breakage. As Rene pointed out I did that changes
due to x86-32/64 arch reunification in 2.6.24. I guess it will take
some time to adapt everything to this kind of changes.

Rene, maybe it's time to update linux-header or it won't help us a lot?

Aldas.

On Jan 29, 2008 10:20 AM, René Rebe <rene@...> wrote:

> Hi,
>
> On 28.01.2008, at 20:07, Mike N wrote:
>
> > Hello,
> >
> > I have used SVN t2-trunk version for a while already, but today it
> > broke
> > when I updated latest changes from svn repo.
> >
> > After the update every build failed in "base/linux-header" stage-0
> > step.
> > After digging in more it turn out that compilation environment tried
> > to call
> > Linux Kernel makefiles with "make ARCH=x86 CROSSCOMPILE= KCC=kcc
> > oldconfig"
> > parameters.
> >
> > Well, as we know Linux 2.6.x doesn't have arch/x86/ directory but
> > instead
> > arch/i386/. The error turn out to be in "base/linux24/linux-conf.in"
> > file.
> > The latest SVN changelog (r27693 by aldas) gives a hint about the
> > change but
> > looks like it breaks the x86 (32bit) compilation.
>
> This was changed in 2.6.24. i386 and x86_64 where unified to just x86
> with
> a Config option whether to build 32 or 64 bit.
>
> However, this changes apparently broken the slightly older linux-header
> package which still has i386 and x86_64.
>
> > The lastest T2 svn package/base/linux24/linux-conf.in file has
> > following
> > line 20:
> > lx_cpu=echo $arch | arch2uname | sed s/.... ppc and mips and bfin
> > stuff
> > here... -e  s/i386/x86
> >
> > The last "-e s/i386/x86" part is new addition and it wrongly changes
> > arch
> > kernel compilation parameter from i386 to x86 value.
> >
> > I removed this i386->x86 sed conversion part from linux-conf.in file
> > and I
> > was once again able to compile something.
> >
> > Is this "workaround fix" valid or have I missed something? Or maybe
> > the
> > latest SVN change was meant to deal with only x86_64 arch but it
> > accidentally changes the 32-bit x86 Kernel arch parameter?
>
>
> This needs to be addressed, maybe by simply updating the older linux-
> header,
> or changing this line to use different substitutes depending on the
> version
> to build.
>

-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
lists@... with a subject of: unsubscribe t2

[Mike N]

Hi,

Thanks for the tip. And you are correct that the "fix" I mentioned above was only half way through. After the fix linux-header package indee compiled but then it failed in 2.6.24 version of Linux26 package.

As a temporary fix I added following lines in linux-conf.in file:

if [ "$pkg" = "linux-header"]; then
  lx_cpu='echo ....sed mentioned above as a fix without the i386->x86 replacing'
else
  lx_cpu='echo ..... the original sed line here with i386->x86 replacing'
fi

However, this is just a temporary fix for myself because naturally this breaks again when/if linux-header package is updated to 2.6.24 version.

Additional question. I tried to find env variable holding the version number of the package but could not find one so I did the "fix" based on package name. $pkg variable gives the package name but how can I find out the package version shown in some status lines and package tar balls? Just in case I sometimes need it in the future.

[Rene Rebe]

Hi,

On Wednesday 30 January 2008 09:24:05 Mike N wrote:

> Hi,
>
> Thanks for the tip. And you are correct that the "fix" I mentioned above was
> only half way through. After the fix linux-header package indee compiled but
> then it failed in 2.6.24 version of Linux26 package.
>
> As a temporary fix I added following lines in linux-conf.in file:
>
> if [ "$pkg" = "linux-header"]; then
>   lx_cpu='echo ....sed mentioned above as a fix without the i386->x86
> replacing'
> else
>   lx_cpu='echo ..... the original sed line here with i386->x86 replacing'
> fi
>
> However, this is just a temporary fix for myself because naturally this
> breaks again when/if linux-header package is updated to 2.6.24 version.
>
> Additional question. I tried to find env variable holding the version number
> of the package but could not find one so I did the "fix" based on package
> name. $pkg variable gives the package name but how can I find out the
> package version shown in some status lines and package tar balls? Just in
> case I sometimes need it in the future.

The version of the currently building package is stored in the variable
$desc_V, the other tags are stored in simillar named variables like:
$desc_I, $desc_T, ...

They are briefly mentioned here:

  http://www.t2-project.org/handbook/html/t2.wrapper.html#t2.variable.environment

Guess this deserves some real pages :-)

Btw. I noticed the recent kernel change does not build for x86_64. I'll fix both
problems over the day.

Yours,

--
  René Rebe - ExactCODE GmbH - Europe, Germany, Berlin
  Geschäftsführer: Susanne Klaus, René Rebe
  Sitz: Berlin, Amtsgericht Charlottenburg HRB 105 123 B
  USt-IdNr.: DE251602478
  http://exactcode.de | http://t2-project.org | http://rene.rebe.name


-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
lists@... with a subject of: unsubscribe t2

[Michael Tross-2]

Hi,

On 29.01.2008 21:25, Aldas Nabazas wrote:
> Hi,
>
> Sorry Mike for the breakage. As Rene pointed out I did that changes
> due to x86-32/64 arch reunification in 2.6.24. I guess it will take
> some time to adapt everything to this kind of changes.
>
> Rene, maybe it's time to update linux-header or it won't help us a lot

as some quick test builds did not show serious regressions, we agreed to
update the linux-header package to 2.6.24, too (committed as r27745).

On the way I removed all outdated patches from the package. That may
cause some applications to fail which directly include kernel headers...

Michael



-----------------------------------------------------------
If you wish to unsubscribe from this mailing, send mail to
lists@... with a subject of: unsubscribe t2