Heads up: staging changes are going to land

Heads up: staging changes are going to land

[Koen Kooi]

Hi,

The changed described in 
http://lists.linuxtogo.org/pipermail/openembedded-devel/2008-July/005819.html 
are about to get committed.

The patch can be previewed at 
http://dominion.thruhere.net/koen/OE/staging-patch.txt

regards,

Koen

Re: Heads up: staging changes are going to land

[Richard Purdie]

Hi,

On Thu, 2008-10-02 at 20:53 +0200, Koen Kooi wrote:
> The changed described in 
> http://lists.linuxtogo.org/pipermail/openembedded-devel/2008-July/005819.html 
> are about to get committed.
> 
> The patch can be previewed at 
> http://dominion.thruhere.net/koen/OE/staging-patch.txt

I had a quick glance through that patch and you should probably rethink
the bits below:

+BUILDSDK_CPPFLAGS = "-isystem${STAGING_INCDIR}"
 export CPPFLAGS = "${TARGET_CPPFLAGS}"
 export TARGET_CPPFLAGS = "-isystem${STAGING_DIR_TARGET}${layout_includedir}"
 
 export BUILD_CFLAGS = "${BUILD_CPPFLAGS} ${BUILD_OPTIMIZATION}"
+BUILDSDK_CFLAGS = "${BUILDSDK_CPPFLAGS} ${BUILD_OPTIMIZATION}"
 export CFLAGS = "${TARGET_CFLAGS}"
 export TARGET_CFLAGS = "${TARGET_CPPFLAGS} ${SELECTED_OPTIMIZATION}"


These SDK things are a bit nuts and I suspect should be removed. I added
them to fix some problem in Poky and I think they may not be needed now
but I'm not 100% and need to find time to check. OE probably doesn't
want them, at least yet.

 
@@ -404,7 +414,7 @@ FULL_OPTIMIZATION = "-fexpensive-optimiz
 ##################################################################
 
 FULL_OPTIMIZATION = "-fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2"
-DEBUG_OPTIMIZATION = "-O -g"
+DEBUG_OPTIMIZATION = "-O -fno-omit-frame-pointer -g"


I think OE changed this for a reason and Poky never merged that change?
Maybe I don't remember right though?

Cheers,

Richard

Re: Heads up: staging changes are going to land

[Koen Kooi]

On 03-10-2008 00:25, Richard Purdie wrote:
> Hi,
>
> On Thu, 2008-10-02 at 20:53 +0200, Koen Kooi wrote:
>> The changed described in
>> http://lists.linuxtogo.org/pipermail/openembedded-devel/2008-July/005819.html
>> are about to get committed.
>>
>> The patch can be previewed at
>> http://dominion.thruhere.net/koen/OE/staging-patch.txt
>
> I had a quick glance through that patch and you should probably rethink
> the bits below:
>
> +BUILDSDK_CPPFLAGS = "-isystem${STAGING_INCDIR}"
>   export CPPFLAGS = "${TARGET_CPPFLAGS}"
>   export TARGET_CPPFLAGS = "-isystem${STAGING_DIR_TARGET}${layout_includedir}"
>
>   export BUILD_CFLAGS = "${BUILD_CPPFLAGS} ${BUILD_OPTIMIZATION}"
> +BUILDSDK_CFLAGS = "${BUILDSDK_CPPFLAGS} ${BUILD_OPTIMIZATION}"
>   export CFLAGS = "${TARGET_CFLAGS}"
>   export TARGET_CFLAGS = "${TARGET_CPPFLAGS} ${SELECTED_OPTIMIZATION}"
>
>
> These SDK things are a bit nuts and I suspect should be removed. I added
> them to fix some problem in Poky and I think they may not be needed now
> but I'm not 100% and need to find time to check. OE probably doesn't
> want them, at least yet.

OK

>
> @@ -404,7 +414,7 @@ FULL_OPTIMIZATION = "-fexpensive-optimiz
>   ##################################################################
>
>   FULL_OPTIMIZATION = "-fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2"
> -DEBUG_OPTIMIZATION = "-O -g"
> +DEBUG_OPTIMIZATION = "-O -fno-omit-frame-pointer -g"
>
>
> I think OE changed this for a reason and Poky never merged that change?
> Maybe I don't remember right though?

AIUI you need -fno-omit-frame-pointer on ARM to do profiling, right?

regards,

Koen

Re: Heads up: staging changes are going to land

[Phil Blundell]

On Fri, 2008-10-03 at 09:12 +0200, Koen Kooi wrote:
> AIUI you need -fno-omit-frame-pointer on ARM to do profiling, right?

You shouldn't do; I can't think of any reason why the profile runtime
would require frame pointers.  What goes wrong without that flag?

p.

Re: Heads up: staging changes are going to land

[Stanislav Brabec]

Koen Kooi wrote:

> >   FULL_OPTIMIZATION = "-fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2"
> > -DEBUG_OPTIMIZATION = "-O -g"
> > +DEBUG_OPTIMIZATION = "-O -fno-omit-frame-pointer -g"
> >
> >
> > I think OE changed this for a reason and Poky never merged that change?
> > Maybe I don't remember right though?
> 
> AIUI you need -fno-omit-frame-pointer on ARM to do profiling, right?

Debugging as well. I just gathered, that -fomit-frame-pointer makes -dbg
packages totally unusable. Even getting a plain stack backtrace is
impossible.

If -fomit-frame-pointer is used, writing of any -dbg package makes no
sense.


________________________________________________________________________
Stanislav Brabec
http://www.penguin.cz/~utx/zaurus

Re: Heads up: staging changes are going to land

[Koen Kooi]

On 03-10-2008 00:25, Richard Purdie wrote:

> @@ -404,7 +414,7 @@ FULL_OPTIMIZATION = "-fexpensive-optimiz
>   ##################################################################
>
>   FULL_OPTIMIZATION = "-fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2"
> -DEBUG_OPTIMIZATION = "-O -g"
> +DEBUG_OPTIMIZATION = "-O -fno-omit-frame-pointer -g"
>
>
> I think OE changed this for a reason and Poky never merged that change?
> Maybe I don't remember right though?

mtn log leads it back to my bitkeeper import from 3 years ago, svn says 
revision 1152 changed it in poky: 
http://svn.o-hand.com/view/poky?rev=1152&view=rev
So I suspect it's indeed about getting usefull profiling out of ARM 
binaries :)

regards,

Koen

Re: Heads up: staging changes are going to land

[Richard Purdie]

On Fri, 2008-10-03 at 09:23 +0200, Koen Kooi wrote:
> On 03-10-2008 00:25, Richard Purdie wrote:
> 
> > @@ -404,7 +414,7 @@ FULL_OPTIMIZATION = "-fexpensive-optimiz
> >   ##################################################################
> >
> >   FULL_OPTIMIZATION = "-fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2"
> > -DEBUG_OPTIMIZATION = "-O -g"
> > +DEBUG_OPTIMIZATION = "-O -fno-omit-frame-pointer -g"
> >
> >
> > I think OE changed this for a reason and Poky never merged that change?
> > Maybe I don't remember right though?
> 
> mtn log leads it back to my bitkeeper import from 3 years ago, svn says 
> revision 1152 changed it in poky: 
> http://svn.o-hand.com/view/poky?rev=1152&view=rev
> So I suspect it's indeed about getting usefull profiling out of ARM 
> binaries :)

It is needed to get decent profiling on ARM, yes. Feel free to merge
that one then :)

Cheers,

Richard

Re: Heads up: staging changes are going to land

[Leon Woestenberg]

Hello Koen,

this change in that patch:

+CROSS_DIR = "${TMPDIR}/cross/${BASE_PACKAGE_ARCH}"

makes my OE PATH look like this:

<...>/build/tmp/cross/ppce300c3/bin

but my cross toolchain did not end up there, instead it ended up in
the usual place: <...>/build/tmp/cross/bin/

so I am getting:

NOTE: make -j5 -e MAKEFLAGS= include/linux/version.h CC=ccache
powerpc-angstrom-linux-gcc  LD=powerpc-angstrom-linux-ld
powerpc-angstrom-linux-gcc: No such file or directory

So obviously for me the CROSS_DIR is not respected somehow for the toolchain.

Any ideas where it could leak away?

I did a scratch build against HEAD using the latest stable bitbake
tree checkout, for MACHINE=mpc8313erdb.

Some other details might still exist at
http://tinderbox.openembedded.net/builders/likewise/

Regards,

Leon.

Re: Heads up: staging changes are going to land

[Koen Kooi]

On 06-10-2008 21:10, Leon Woestenberg wrote:
> Hello Koen,
>
> this change in that patch:
>
> +CROSS_DIR = "${TMPDIR}/cross/${BASE_PACKAGE_ARCH}"
>
> makes my OE PATH look like this:
>
> <...>/build/tmp/cross/ppce300c3/bin
>
> but my cross toolchain did not end up there, instead it ended up in
> the usual place:<...>/build/tmp/cross/bin/

My cross looks like this:

koen@dominion:/OE/angstrom-dev/cross$ ls
arm-oabi/  armv4t/  armv5te/  armv6/  armv7a/  avr32/  ppc405/  ppc603e/

  > so I am getting:
>
> NOTE: make -j5 -e MAKEFLAGS= include/linux/version.h CC=ccache
> powerpc-angstrom-linux-gcc  LD=powerpc-angstrom-linux-ld
> powerpc-angstrom-linux-gcc: No such file or directory
>
> So obviously for me the CROSS_DIR is not respected somehow for the toolchain.
>
> Any ideas where it could leak away?
>
> I did a scratch build against HEAD using the latest stable bitbake
> tree checkout, for MACHINE=mpc8313erdb.
>
> Some other details might still exist at
> http://tinderbox.openembedded.net/builders/likewise/

I'm starting a build with that machine to see if I can reproduce it.

regards,

Koen

Re: Heads up: staging changes are going to land

[Koen Kooi]

On 06-10-2008 21:24, Koen Kooi wrote:
> On 06-10-2008 21:10, Leon Woestenberg wrote:
>> Hello Koen,
>>
>> this change in that patch:
>>
>> +CROSS_DIR = "${TMPDIR}/cross/${BASE_PACKAGE_ARCH}"
>>
>> makes my OE PATH look like this:
>>
>> <...>/build/tmp/cross/ppce300c3/bin
>>
>> but my cross toolchain did not end up there, instead it ended up in
>> the usual place:<...>/build/tmp/cross/bin/
>
> My cross looks like this:
>
> koen@dominion:/OE/angstrom-dev/cross$ ls
> arm-oabi/ armv4t/ armv5te/ armv6/ armv7a/ avr32/ ppc405/ ppc603e/
>
>  > so I am getting:
>>
>> NOTE: make -j5 -e MAKEFLAGS= include/linux/version.h CC=ccache
>> powerpc-angstrom-linux-gcc LD=powerpc-angstrom-linux-ld
>> powerpc-angstrom-linux-gcc: No such file or directory
>>
>> So obviously for me the CROSS_DIR is not respected somehow for the
>> toolchain.
>>
>> Any ideas where it could leak away?
>>
>> I did a scratch build against HEAD using the latest stable bitbake
>> tree checkout, for MACHINE=mpc8313erdb.
>>
>> Some other details might still exist at
>> http://tinderbox.openembedded.net/builders/likewise/
>
> I'm starting a build with that machine to see if I can reproduce it.

I now have:

koen@dominion:/OE/angstrom-dev/cross$ ls
arm-oabi/  armv4t/  armv5te/  armv6/  armv7a/  avr32/  ppc405/  ppc603e/ 
  ppce300c3/
koen@dominion:/OE/angstrom-dev/cross$ find ppce300c3/ | grep -- -gcc
ppce300c3/bin/powerpc-angstrom-linux-gcc
ppce300c3/bin/powerpc-angstrom-linux-gccbug
ppce300c3/bin/powerpc-angstrom-linux-gcc-4.2.4

I suspect packaged-staging is using the old staging packages (as can be 
seen from the do_setscene in the gcc tinderbox log). Could you try it 
without keeping pstage?

regards,

Koen

Re: Heads up: staging changes are going to land

[Leon Woestenberg]

Hello Koen,

On Mon, Oct 6, 2008 at 9:38 PM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> I suspect packaged-staging is using the old staging packages (as can be seen
> from the do_setscene in the gcc tinderbox log). Could you try it without
> keeping pstage?
>

Hmm, normally my Makefile clean target clean the complete TMPDIR.
Maybe this project is misconfigured, thanks for the heads-up on the
setscene step.

Thanks,
-- 
Leon

Re: Heads up: staging changes are going to land

[Leon Woestenberg]

Hello Koen,

On Tue, Oct 7, 2008 at 1:01 AM, Leon Woestenberg
<leon.woestenberg@gmail.com> wrote:
> On Mon, Oct 6, 2008 at 9:38 PM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
>> I suspect packaged-staging is using the old staging packages (as can be seen
>
> Maybe this project is misconfigured, thanks for the heads-up on the
> setscene step.
>
Affirmed. I had the packaged staging deployment dir overriden to exist
outside TMPDIR.

Sorry for the fuzz, thanks for the patch!

Regards,
-- 
Leon

Re: Heads up: staging changes are going to land

[Koen Kooi]

On 07-10-2008 10:29, Leon Woestenberg wrote:
> Hello Koen,
>
> On Tue, Oct 7, 2008 at 1:01 AM, Leon Woestenberg
> <leon.woestenberg@gmail.com>  wrote:
>> On Mon, Oct 6, 2008 at 9:38 PM, Koen Kooi<k.kooi@student.utwente.nl>  wrote:
>>> I suspect packaged-staging is using the old staging packages (as can be seen
>> Maybe this project is misconfigured, thanks for the heads-up on the
>> setscene step.
>>
> Affirmed. I had the packaged staging deployment dir overriden to exist
> outside TMPDIR.
>
> Sorry for the fuzz, thanks for the patch!

It's always nice to know that the problem is somewhere else :)

regards,

Koen