[Buildroot] AVR32 is broken (toolchain)

[Buildroot] AVR32 is broken (toolchain)

[Thiago A. Corrêa]

Hi,

     I was able to track it down to this commit:
8e8103bfb55a814d68eabb51afebfb35c3e7cec8
     The toolchain builds right before it, but doesn't after.
     I'm trying to figure it out, but I'm not very experienced with
the toolchain build or git, so it's taking me quite a while. Perhaps
someone with more experience could take a look at the issue?

     It's reproductible with make atngw100_defconfig

Kind Regards,
     Thiago A. Correa

[Buildroot] AVR32 is broken (toolchain)

[Peter Hüwe]

Am Donnerstag 22 Juli 2010 00:02:31 schrieb Thiago A. Corr?a:
> Hi,
> 
>      I was able to track it down to this commit:
> 8e8103bfb55a814d68eabb51afebfb35c3e7cec8
>      The toolchain builds right before it, but doesn't after.
>      I'm trying to figure it out, but I'm not very experienced with
> the toolchain build or git, so it's taking me quite a while. Perhaps
> someone with more experience could take a look at the issue?
> 
>      It's reproductible with make atngw100_defconfig
> 
> Kind Regards,
>      Thiago A. Correa


Hi,

thanks for reporting this issue.
I'll have a look into it - currently the code still compiles :)

Meanwhile, maybe you could post the results of the build, i.e. the build 
failure / error messages here, or use a pasting service like 
http://dpaste.com/ and then post the link here.

Thanks,
Peter

[Buildroot] AVR32 is broken (toolchain)

[Thiago A. Corrêa]

Hi Peter,

2010/7/21 Peter H?we <PeterHuewe@gmx.de>:
>
> thanks for reporting this issue.
> I'll have a look into it - currently the code still compiles :)

What do you mean it still compiles?
I get an error building the toolchain, something along the lines of:

make[3]: *** Waiting for unfinished jobs....
/root/buildroot-head/output/toolchain/gcc-4.2.2-avr32-2.1.5-final/./gcc/as:
line 2: exec: --: invalid option
exec: usage: exec [-cl] [-a name] [command [arguments ...]] [redirection ...]
make[3]: *** [crtbeginS.o] Error 1


> Meanwhile, maybe you could post the results of the build, i.e. the build
> failure / error messages here, or use a pasting service like
> http://dpaste.com/ and then post the link here.
>

This is what my ssh window had in it's buffer: http://dpaste.com/220727/

I can get a full log if needed. I get this with HEAD and
8e8103bfb55a814d68eabb51afebfb35c3e7cec8
But not with 4fdecac9d692b8d6f071ba6ad938b6ad68b675fd

Thanks,
    Thiago A. Correa

[Buildroot] AVR32 is broken (toolchain)

[Peter Hüwe]

Peter 	> I'll have a look into it - currently the code still compiles :)
Thiago 	> What do you mean it still compiles?
This just means that my machine is a little bit slow and it takes awhile to 
compile buildroot.


Thiago > I get an error building the toolchain, something along the lines of:
Thiago > make[3]: *** Waiting for unfinished jobs....
Thiago > /root/buildroot-head/output/toolchain/gcc-4.2.2-avr32-2.1.5-
final/./gcc/as:
Thiago > line 2: exec: --: invalid option
Thiago> exec: usage: exec [-cl] [-a name] [command [arguments ...]] 
[redirection
Thiago >  ...] make[3]: *** [crtbeginS.o] Error 1

Yeah - I can reproduce this issue.

However I got another error first (something like avr32-.....-ar: command not 
found). But I got the same error after adding the 
        AR_FOR_BUILD="$(HOSTAR)" \
        AR_FOR_TARGET="$(TARGET_AR)" \
 lines again.

Currently I'm trying compiling a version of BR with and without the changes in 
Thomas' commit - tomorrow I'll compare the output :)

Thanks,
Peter

[Buildroot] AVR32 is broken (toolchain)

[Thomas Petazzoni]

Hello,

On Thu, 22 Jul 2010 04:57:23 +0200
Peter H?we <PeterHuewe@gmx.de> wrote:

> Yeah - I can reproduce this issue.
> 
> However I got another error first (something like avr32-.....-ar:
> command not found). But I got the same error after adding the 
>         AR_FOR_BUILD="$(HOSTAR)" \
>         AR_FOR_TARGET="$(TARGET_AR)" \
>  lines again.
> 
> Currently I'm trying compiling a version of BR with and without the
> changes in Thomas' commit - tomorrow I'll compare the output :)

FWIW, I can also reproduce the issue. It seems to be due to the fact
that the --prefix location is not writable (since it's /usr). See
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32101#c4.

So, I've cooked the attached patch, which allows the toolchain to
build, but the toolchain fails when compiling Busybox with messages
such as :

dc.c:(.text+0x10c): relocation truncated to fit: R_AVR32_22H_PCREL
against symbol `__avr32_f64_to_u32' defined in .text section
in /home/test/issue/old-br/buildroot/output/staging/usr/lib/gcc/avr32-unknown-linux-uclibc/4.2.2/libgcc.a(_avr32_f64_to_s32.o)

I don't yet know what is happening.

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: toolchain-fix.patch
Type: text/x-patch
Size: 2623 bytes
Desc: not available
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20100722/10832c77/attachment.bin>

[Buildroot] AVR32 is broken (toolchain)

[Thomas Petazzoni]

Hello,

On Thu, 22 Jul 2010 14:17:15 +0200
Thomas Petazzoni <thomas.petazzoni@free-electrons.com> wrote:

> FWIW, I can also reproduce the issue. It seems to be due to the fact
> that the --prefix location is not writable (since it's /usr). See
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32101#c4.
> 
> So, I've cooked the attached patch, which allows the toolchain to
> build, but the toolchain fails when compiling Busybox with messages
> such as :
> 
> dc.c:(.text+0x10c): relocation truncated to fit: R_AVR32_22H_PCREL
> against symbol `__avr32_f64_to_u32' defined in .text section
> in /home/test/issue/old-br/buildroot/output/staging/usr/lib/gcc/avr32-unknown-linux-uclibc/4.2.2/libgcc.a(_avr32_f64_to_s32.o)
> 
> I don't yet know what is happening.

Ok, the patch that Gustavo has posted on Busybox at 3:50 PM today fixes
the problem. So with my patch on the toolchain + Gustavo's patch on
Busybox, I've been able to build an AVR32 toolchain, BusyBox, DirectFB
and a sample DirectFB application. However, I don't have any hardware
to test.

BTW, if there are people working at Atmel on this list, it'd be great
if Atmel could put online updated version of gcc, binutils and gdb for
the AVR32 architecture. Currently, we are stuck at gcc 4.2.x and
binutils 2.18.

On the Atmel website, it is possible to download a single zip file that
contains tarballs for gcc, gdb, binutils and newlib. Having that in a
single zip file is very unpractical for build systems like Buildroot.
Moreover, having to register on the website (name, address, phone
number and all the stuff) is even more inconvenient.

Would it be possible to just drop the individual tarballs at some
public location, with no registration required ? Or even better, get
the AVR32 support merged in the respective upstream projects :-)

Thanks!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] AVR32 is broken (toolchain)

[Thiago A. Corrêa]

Hi,

On Thu, Jul 22, 2010 at 12:10 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:

> Ok, the patch that Gustavo has posted on Busybox at 3:50 PM today fixes
> the problem. So with my patch on the toolchain + Gustavo's patch on
> Busybox, I've been able to build an AVR32 toolchain, BusyBox, DirectFB
> and a sample DirectFB application. However, I don't have any hardware
> to test.

Thanks for the fast reply. I don't have any standard kit here, but I
can test on my own hardware.

> BTW, if there are people working at Atmel on this list, it'd be great
> if Atmel could put online updated version of gcc, binutils and gdb for
> the AVR32 architecture. Currently, we are stuck at gcc 4.2.x and
> binutils 2.18.
>

I don't work for Atmel, but would it be too bad to get the patches
back into buildroot? That's what OpenWRT is doing, and they have
updated to gcc 4.2.4 and binutils 2.19

I could make the changes and post a pull request.

Kind Regards,
    Thiago A. Correa

[Buildroot] AVR32 is broken (toolchain)

[Thiago A. Corrêa]

Hi Thomas,

On Thu, Jul 22, 2010 at 12:10 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
>
> Ok, the patch that Gustavo has posted on Busybox at 3:50 PM today fixes
> the problem. So with my patch on the toolchain + Gustavo's patch on
> Busybox, I've been able to build an AVR32 toolchain, BusyBox, DirectFB
> and a sample DirectFB application. However, I don't have any hardware
> to test.
>

   I can't find his patch in the archives. Could you forward his patch
or provide a link?
   Your patch worked fine for me and now I'm able to build the toolchain.

Kind Regards,
    Thiago A. Correa

[Buildroot] AVR32 is broken (toolchain)

[Thomas Petazzoni]

On Fri, 23 Jul 2010 16:12:56 -0300
Thiago A. Corr?a <thiago.correa@gmail.com> wrote:

>    I can't find his patch in the archives. Could you forward his patch
> or provide a link?

It's already in Git (committed yesterday)

 http://git.buildroot.net/buildroot/commit/?id=d8a41bb5364461dd62a96fbc384649edd259c485

>    Your patch worked fine for me and now I'm able to build the
> toolchain.

Great!

I will make some more testing with my patch and I'll push it soon.

Thanks,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] AVR32 is broken (toolchain)

[Peter Hüwe]

Am Freitag 23 Juli 2010 23:18:50 schrieb Thomas Petazzoni:
> Thiago A. Corr?a <thiago.correa@gmail.com> wrote:
> >    Your patch worked fine for me and now I'm able to build the
> > toolchain.
> 
> Great!
> 
> I will make some more testing with my patch and I'll push it soon.

Hi Thomas,

I just tried to compile the atngw100_defconfig using your patch and the latest 
git repo but unfortunately it fails while building host-lzo

>>> host-lzo 2.03 Configuring
...
configure: loading cache /avr/buildroot/output/build/host-config.cache
configure: Configuring LZO 2.03
checking build system type... (cached) x86_64-unknown-linux-gnu
checking host system type... (cached) x86_64-unknown-linux-gnu
checking target system type... config.sub: missing argument
Try `config.sub --help' for more information.
configure: error: /bin/sh autoconf/config.sub  failed
make: *** [/avr/buildroot/output/build/host-lzo-2.03/.stamp_configured] Error 
1

You can find the whole log at http://dpaste.com/hold/222214/

Do you think this is related to the original problem?
I'm running buildroot on a 64-bit gentoo installation - 
gcc: x86_64-pc-linux-gnu-4.1.2
Es werden eingebaute Spezifikationen verwendet.
Ziel: x86_64-pc-linux-gnu
Konfiguriert mit: /var/tmp/portage/sys-
devel/gcc-4.1.2/work/gcc-4.1.2/configure --prefix=/usr --bindir=/usr/x86_64-
pc-linux-gnu/gcc-bin/4.1.2 --includedir=/usr/lib/gcc/x86_64-pc-linux-
gnu/4.1.2/include --datadir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.1.2 --
mandir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.1.2/man --
infodir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.1.2/info --with-gxx-include-
dir=/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.2/include/g++-v4 --host=x86_64-pc-
linux-gnu --build=x86_64-pc-linux-gnu --disable-altivec --enable-nls --
without-included-gettext --with-system-zlib --disable-checking --disable-
werror --enable-secureplt --enable-multilib --enable-libmudflap --disable-
libssp --enable-java-awt=gtk --enable-languages=c,c++,java,treelang,fortran --
enable-shared --enable-threads=posix --enable-__cxa_atexit --enable-
clocale=gnu
Thread-Modell: posix
gcc-Version 4.1.2 (Gentoo 4.1.2 p1.3)


Thanks,
Peter

[Buildroot] AVR32 is broken (toolchain)

[Thomas Petazzoni]

Hello Peter,

On Tue, 27 Jul 2010 03:14:14 +0200
Peter H?we <PeterHuewe@gmx.de> wrote:

> I just tried to compile the atngw100_defconfig using your patch and
> the latest git repo but unfortunately it fails while building host-lzo
> 
> >>> host-lzo 2.03 Configuring

host-lzo is built for the host, so it's quite unlikely that any AVR32
change can have an impact on it.

> configure: loading cache /avr/buildroot/output/build/host-config.cache
> configure: Configuring LZO 2.03
> checking build system type... (cached) x86_64-unknown-linux-gnu
> checking host system type... (cached) x86_64-unknown-linux-gnu
> checking target system type... config.sub: missing argument
> Try `config.sub --help' for more information.

It sounds like the autoconf host-config.cache has been filled with an
empty value for --target.

Can you show the contents
of /avr/buildroot/output/build/host-config.cache after the failure, and
can you post the *full* build log somewhere ?

Thanks,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] AVR32 is broken (toolchain)

[Peter Hüwe]

Am Donnerstag 22 Juli 2010 17:10:35 schrieb Thomas Petazzoni:
> BTW, if there are people working at Atmel on this list, it'd be great...

Maybe CC or write to avr32 at atmel.com ?

> On the Atmel website, it is possible to download a single zip file that
> contains tarballs for gcc, gdb, binutils and newlib. Having that in a
> single zip file is very unpractical for build systems like Buildroot.
Yeah. But since I guess it is all gpl licenced software, wouldn't it be 
possible for the buildroot team to download the zip, repack it in single zips 
and host/mirror them on the buildroot mirrors? Or atleast the toolchain parts 
of the zip?

I know it's not a good option, but maybe it would ease the life of the 
buildroot devs (i.e. you and peter :P)

>Moreover, having to register on the website (name, address, phone
> number and all the stuff) is even more inconvenient.

I know that form, and even if it looks like it is mandatory, it is not, just 
click continue - even hotlinking would be possible.

Thanks,
Peter

[Buildroot] AVR32 is broken (toolchain)

[Bjørn Forsman]

Hi all,

On 22 July 2010 17:10, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> BTW, if there are people working at Atmel on this list, it'd be great
> if Atmel could put online updated version of gcc, binutils and gdb for
> the AVR32 architecture. Currently, we are stuck at gcc 4.2.x and
> binutils 2.18.
>
> On the Atmel website, it is possible to download a single zip file that
> contains tarballs for gcc, gdb, binutils and newlib. Having that in a
> single zip file is very unpractical for build systems like Buildroot.
> Moreover, having to register on the website (name, address, phone
> number and all the stuff) is even more inconvenient.
>
> Would it be possible to just drop the individual tarballs at some
> public location, with no registration required ? Or even better, get
> the AVR32 support merged in the respective upstream projects :-)

FYI: Atmel are eol'ing the AVR32AP7000 device. It is currently "Not
recommended for new design" [1].

So...I guess AVR32 Linux is kind of a dead end now... or use uClinux
on AVR32 UC3 devices...

Best regards,
Bj?rn Forsman

[1]: http://www.atmel.com/dyn/products/product_card.asp?part_id=3903

[Buildroot] AVR32 is broken (toolchain)

[Thiago A. Corrêa]

Hi,

2010/7/30 Bj?rn Forsman <bjorn.forsman@gmail.com>:
> Hi all,
>
>
> FYI: Atmel are eol'ing the AVR32AP7000 device. It is currently "Not
> recommended for new design" [1].

They have said in forums and other communications that they would EOL
because it reached the middle of it's expected life time. They still
produce them and would continue to do so for sometime.

> So...I guess AVR32 Linux is kind of a dead end now... or use uClinux
> on AVR32 UC3 devices...

But that's not an option for existing products that needs support. For
new development, people would probably look for arms or mips. We are
in the process of migrating to an ARM device from Samsung, but still
we need to be able to support the hardware that is out there. The next
batch will be our last with AP7000 but warranties, support, etc still
goes on.

Kind Regards,
    Thiago A. Correa