[Buildroot] Configuring host-lzo fails

[Buildroot] Configuring host-lzo fails

[Peter Hüwe]

I've started a new thread, since it seems not related to avr32 in particular.

I'm using the atngw100_defconfig on my 64-bit gentoo system with a gcc 4.1.2

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

TP > It sounds like the autoconf host-config.cache has been filled with an
TP > empty value for --target.
 
TP > Can you show the contents
TP > of /avr/buildroot/output/build/host-config.cache after the failure, and
TP> can you post the *full* build log somewhere ?

I've gzipped the whole build log (from a fresh git clone) and uploaded it to 
http://www.hs-augsburg.de/~phuewe/typescript2.gz (498K, unzipped 8,8Mb)

The contens of the host-config.cache can be found at
http://paste.pocoo.org/show/242433/


Thanks,
Peter

[Buildroot] Configuring host-lzo fails

[Thomas Petazzoni]

Hello Peter,

On Tue, 27 Jul 2010 23:33:09 +0200
Peter H?we <PeterHuewe@gmx.de> wrote:

> I've started a new thread, since it seems not related to avr32 in
> particular.

Yep.

> I'm using the atngw100_defconfig on my 64-bit gentoo system with a
> gcc 4.1.2

I was also able to reproduce the issue.

> I've gzipped the whole build log (from a fresh git clone) and
> uploaded it to http://www.hs-augsburg.de/~phuewe/typescript2.gz
> (498K, unzipped 8,8Mb)
> 
> The contens of the host-config.cache can be found at
> http://paste.pocoo.org/show/242433/

Thanks. As I could reproduce, I've started to investigate the problem.
I've summarized my findings at
http://free-electrons.com/~thomas/pub/autoconf-issue.txt, but I don't
have a clear solution yet.

As a temporary fix, you can probably try something like :

HOST_LZO_CONF_OPT = --target=$(GNU_HOST_NAME)

in package/lzo/lzo.mk.

Regards,

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

[Buildroot] Configuring host-lzo fails

[Peter Huewe]

Hi Thomas, 
> I was also able to reproduce the issue. 
good - makes it easier to track down ;)

> Thanks. As I could reproduce, I've started to investigate the problem.
> I've summarized my findings at
> http://free-electrons.com/~thomas/pub/autoconf-issue.txt, but I don't
> have a clear solution yet.
> 
> As a temporary fix, you can probably try something like :
> 
> HOST_LZO_CONF_OPT = --target=$(GNU_HOST_NAME)
> 
> in package/lzo/lzo.mk.
> 

I tried your temporary and it works - however I made following observation.
0)  rm -rf /buildroot/output/build/host-lzo-2.03/  buildroot/output/build/host-config.cache
1) Add HOST_LZO_CONF_OPT = --target=$(GNU_HOST_NAME) to package/lzo/lzo.mk
2) make
--> host lzo builds fine

3) rm -rf /buildroot/output/build/host-lzo-2.03/  buildroot/output/build/host-config.cache
4) remove HOST_LZO_CONF_OPT = --target=$(GNU_HOST_NAME) from package/lzo/lzo.mk
5) make
--> host lzo builds perfectly fine too!

Any Ideas? 
Since I've removed the cache, we don't have a cached value here.


Thanks,
Peter


btw.: the german umlaut in my name (?) is usually converted to ue, strange german language thing ;)

[Buildroot] Configuring host-lzo fails

[Peter Hüwe]

Am Mittwoch 28 Juli 2010 00:20:05 schrieb Peter Huewe:
Hi Thomas,

please also note that Ossy has reported this issue on Friday at 11:18pm CEST

"[Buildroot] lzof 2.03 fails building with activated configuration cache"

Ossy > I was wondering why the ACC Conformance check of te compiler fails for 
Ossy > lzof and is successful for some others. I switched of the configration 
Ossy > cache and the lzof package did NOT compile.
Ossy > Please see my config attached. For reproducing this error, turn on the 
Ossy > config cache again.

Thanks,
Peter

[Buildroot] Configuring host-lzo fails

[Ossy]

Am 28.07.2010 02:14, schrieb Peter H?we:
> Am Mittwoch 28 Juli 2010 00:20:05 schrieb Peter Huewe:
> Hi Thomas,
>
> please also note that Ossy has reported this issue on Friday at 11:18pm CEST
>
> "[Buildroot] lzof 2.03 fails building with activated configuration cache"
>
> Ossy>  I was wondering why the ACC Conformance check of te compiler fails for
> Ossy>  lzof and is successful for some others. I switched of the configration
> Ossy>  cache and the lzof package did NOT compile.
> Ossy>  Please see my config attached. For reproducing this error, turn on the
> Ossy>  config cache again.
>
True. I already ran into the same trap. Just to correct my mail from 
Friday: The package does NOT compile with ACTIVATED configuration cache. 
I switched the cache off and it was fine.

Regards,
Marcus

[Buildroot] Configuring host-lzo fails

[Thomas Petazzoni]

Hello Peter,

On Tue, 27 Jul 2010 23:33:09 +0200
Peter H?we <PeterHuewe@gmx.de> wrote:

> TP > It sounds like the autoconf host-config.cache has been filled
> with an TP > empty value for --target.
>  
> TP > Can you show the contents
> TP > of /avr/buildroot/output/build/host-config.cache after the
> failure, and
> TP> can you post the *full* build log somewhere ?
> 
> I've gzipped the whole build log (from a fresh git clone) and
> uploaded it to http://www.hs-augsburg.de/~phuewe/typescript2.gz
> (498K, unzipped 8,8Mb)
> 
> The contens of the host-config.cache can be found at
> http://paste.pocoo.org/show/242433/

I finally took the time to track and fix this problem of host-lzo
configuration with the shared configuration cache.

The problem was due to the test in ./configure for the target system
name being incorrectly written. Newer versions of autoconf do not
generate such an incorrect code, so the solution was simply to
autoreconfigure the package. Unfortunately, it wasn't so simple, as the
package required several adaptations before being autoreconfigurable.

This is fixed in my for-2010.08 branch, the specific commit being
http://git.buildroot.net/~tpetazzoni/git/buildroot/commit/?h=for-2010.08&id=fee182d20d9b1876436c9673744ae6312a40e889.

If you have the time to test this branch and report whether it also
fixes the problem for you, it'd be nice.

Regards,

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

[Buildroot] Configuring host-lzo fails

[Ossy]

Am 13.08.2010 12:26, schrieb Thomas Petazzoni:
> This is fixed in my for-2010.08 branch, the specific commit being
> http://git.buildroot.net/~tpetazzoni/git/buildroot/commit/?h=for-2010.08&id=fee182d20d9b1876436c9673744ae6312a40e889.
>
> If you have the time to test this branch and report whether it also
> fixes the problem for you, it'd be nice.
Hi,

I applied this patch
http://git.buildroot.net/~tpetazzoni/git/buildroot/patch/?id=fee182d20d9b1876436c9673744ae6312a40e889
to the current git tree on my drive. This time I did not encounter a 
problem with host-lzo, but host-zlib ran into an error ("zlib.a not 
present"). After make flush with a consecutive make, the systembuild 
went on without problems.

I'm not sure if this is a problem "behind" host-lzo (which was solved by 
the former make flush for host-lzo) or a new problem caused by the patch 
fee182d20d9b1876436c9673744ae6312a40e889.

Regards,
Marcus

[Buildroot] Configuring host-lzo fails

[Ossy]

Am 18.08.2010 23:34, schrieb Ossy:
> Am 13.08.2010 12:26, schrieb Thomas Petazzoni:
>> This is fixed in my for-2010.08 branch, the specific commit being
>> http://git.buildroot.net/~tpetazzoni/git/buildroot/commit/?h=for-2010.08&id=fee182d20d9b1876436c9673744ae6312a40e889.
>>
>>
>> If you have the time to test this branch and report whether it also
>> fixes the problem for you, it'd be nice.
>
 > [host-zlib error]
>
The same patch applied, I ran into an error with expat today:
configure: error: no acceptable ld found in $PATH
make: *** 
[/home/ossy/buildroot/buildroot-dev/output/build/expat-2.0.1/.stamp_configured] 
Fehler 1

After make flush; make the system build went on without problems.

Regards,
Marcus

[Buildroot] Configuring host-lzo fails

[Ossy]

Am 21.08.2010 16:41, schrieb Ossy:
> Am 18.08.2010 23:34, schrieb Ossy:
>> Am 13.08.2010 12:26, schrieb Thomas Petazzoni:
>>> This is fixed in my for-2010.08 branch, the specific commit being
>>> http://git.buildroot.net/~tpetazzoni/git/buildroot/commit/?h=for-2010.08&id=fee182d20d9b1876436c9673744ae6312a40e889.
>>>
>>>
>>>
>>> If you have the time to test this branch and report whether it also
>>> fixes the problem for you, it'd be nice.
>>
>  > [host-zlib error]
>>
> The same patch applied, I ran into an error with expat today:
> configure: error: no acceptable ld found in $PATH
> make: ***
> [/home/ossy/buildroot/buildroot-dev/output/build/expat-2.0.1/.stamp_configured]
> Fehler 1
>
> After make flush; make the system build went on without problems.
>
The same whith which-2.20:
/buildroot/buildroot-dev/output/staging -DHAVE_CONFIG_H -I. -I..     -Os 
-pipe -Os  -mtune=arm926ej-s -march=armv5te -mabi=aapcs-linux 
-msoft-float -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE 
-D_FILE_OFFSET_BITS=64 
-I/home/ossy/buildroot/buildroot-dev/output/staging/usr/include 
-I/home/ossy/buildroot/buildroot-dev/output/staging/include -c shell.c
Unknown depmode none
make[3]: *** [tilde.o] Fehler 1

I had to make flush and rm -r output/build/which-2.20/
After flushing and removing the system build went on. Including which.

[Buildroot] Configuring host-lzo fails

[Thomas Petazzoni]

On Sat, 21 Aug 2010 16:41:30 +0200
Ossy <ossy1980@gmx.net> wrote:

> The same patch applied, I ran into an error with expat today:
> configure: error: no acceptable ld found in $PATH
> make: *** 
> [/home/ossy/buildroot/buildroot-dev/output/build/expat-2.0.1/.stamp_configured] 
> Fehler 1
> 
> After make flush; make the system build went on without problems.

When something doesn't work, then you do "make flush" and then it
works, it means that you've hit a configuration cache issue.

I'm more and more tempted to disable this configuration cache feature.
According to the autoconf maintainers, it could theorically work, but
in practice, there's no real convention on the variable name/value
stored in the configuration cache, so re-using the cache between
different packages is going to cause a lot of troubles. As we are
seeing.

Regards,

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

[Buildroot] Configuring host-lzo fails

[Khem Raj]

On (21/08/10 18:01), Thomas Petazzoni wrote:
> On Sat, 21 Aug 2010 16:41:30 +0200
> Ossy <ossy1980@gmx.net> wrote:
> 
> > The same patch applied, I ran into an error with expat today:
> > configure: error: no acceptable ld found in $PATH
> > make: *** 
> > [/home/ossy/buildroot/buildroot-dev/output/build/expat-2.0.1/.stamp_configured] 
> > Fehler 1
> > 
> > After make flush; make the system build went on without problems.
> 
> When something doesn't work, then you do "make flush" and then it
> works, it means that you've hit a configuration cache issue.
> 
> I'm more and more tempted to disable this configuration cache feature.
> According to the autoconf maintainers, it could theorically work, but
> in practice, there's no real convention on the variable name/value
> stored in the configuration cache, so re-using the cache between
> different packages is going to cause a lot of troubles. As we are
> seeing.

yes thats very true. 

-Khem