[PATCH 1/3] linux: add LIC_FILES_CHKSUM

[PATCH 1/3] linux: add LIC_FILES_CHKSUM

[Ilya Yanok]

Linux is licensed under GPL so add appropriate LIC_FILES_CHKSUM to
linux.inc file.

Signed-off-by: Ilya Yanok <yanok@emcraft.com>
---
 meta/recipes-kernel/linux/linux.inc |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/meta/recipes-kernel/linux/linux.inc b/meta/recipes-kernel/linux/linux.inc
index 93859bb..c23d90f 100644
--- a/meta/recipes-kernel/linux/linux.inc
+++ b/meta/recipes-kernel/linux/linux.inc
@@ -2,6 +2,8 @@ DESCRIPTION = "Linux Kernel"
 SECTION = "kernel"
 LICENSE = "GPL"
 
+LIC_FILES_CHKSUM = "file://COPYING;md5=d7810fab7487fb0aad327b76f1be7cd7"
+
 inherit kernel
 
 DEPENDS_append_em-x270 = " mtd-utils "
-- 
1.7.4

[PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Ilya Yanok]

Signed-off-by: Ilya Yanok <yanok@emcraft.com>
---
 meta/recipes-bsp/uboot/u-boot.inc |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
index 889f621..6492016 100644
--- a/meta/recipes-bsp/uboot/u-boot.inc
+++ b/meta/recipes-bsp/uboot/u-boot.inc
@@ -5,6 +5,8 @@ PRIORITY = "optional"
 LICENSE = "GPL"
 PROVIDES = "virtual/bootloader"
 
+LIC_FILES_CHKSUM = "file://COPYING;md5=1707d6db1d42237583f50183a5651ecb"
+
 inherit deploy
 
 PARALLEL_MAKE=""
-- 
1.7.4

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Darren Hart]

On 02/08/2011 12:26 PM, Ilya Yanok wrote:
> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
> ---
>   meta/recipes-bsp/uboot/u-boot.inc |    2 ++
>   1 files changed, 2 insertions(+), 0 deletions(-)
>
> diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
> index 889f621..6492016 100644
> --- a/meta/recipes-bsp/uboot/u-boot.inc
> +++ b/meta/recipes-bsp/uboot/u-boot.inc
> @@ -5,6 +5,8 @@ PRIORITY = "optional"
>   LICENSE = "GPL"
>   PROVIDES = "virtual/bootloader"
>
> +LIC_FILES_CHKSUM = "file://COPYING;md5=1707d6db1d42237583f50183a5651ecb"
> +

 From what I can tell u-boot.inc is only used by u-boot_git.bb which 
already defines LIC_FILES_CHKSUM and the appropriate GPLv2+ LICENSE 
variable.

It seems to me the LICENSE and LIC_FILES_CHKSUM should be defined in the 
same recipe as the SRC_URI. In this case, u-boot_git.bb. Is there a 
motivating reason to add it to .inc? If so, we should also remove it 
from u-boot_git.bb to avoid unnecessary duplication which leads to 
confusion down the road.

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Gary Thomas]

On 02/09/2011 10:38 AM, Darren Hart wrote:
> On 02/08/2011 12:26 PM, Ilya Yanok wrote:
>> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
>> ---
>> meta/recipes-bsp/uboot/u-boot.inc | 2 ++
>> 1 files changed, 2 insertions(+), 0 deletions(-)
>>
>> diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
>> index 889f621..6492016 100644
>> --- a/meta/recipes-bsp/uboot/u-boot.inc
>> +++ b/meta/recipes-bsp/uboot/u-boot.inc
>> @@ -5,6 +5,8 @@ PRIORITY = "optional"
>> LICENSE = "GPL"
>> PROVIDES = "virtual/bootloader"
>>
>> +LIC_FILES_CHKSUM = "file://COPYING;md5=1707d6db1d42237583f50183a5651ecb"
>> +
>
>  From what I can tell u-boot.inc is only used by u-boot_git.bb which already defines LIC_FILES_CHKSUM and the appropriate GPLv2+ LICENSE variable.
>
> It seems to me the LICENSE and LIC_FILES_CHKSUM should be defined in the same recipe as the SRC_URI. In this case, u-boot_git.bb. Is there a motivating reason to add it to .inc? If
> so, we should also remove it from u-boot_git.bb to avoid unnecessary duplication which leads to confusion down the road.
>

I agree - what if you have multiple .bb files sharing a single
.inc and the license and/or license data are different?  It
seems to me that this data should live where the actual sources
are being introduced, e.g. the .bb file.

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Ilya Yanok]

Hi Gary,

09.02.2011 20:42, Gary Thomas wrote:
>> It seems to me the LICENSE and LIC_FILES_CHKSUM should be defined in
>> the same recipe as the SRC_URI. In this case, u-boot_git.bb. Is there
>> a motivating reason to add it to .inc? If
>> so, we should also remove it from u-boot_git.bb to avoid unnecessary
>> duplication which leads to confusion down the road.
>>
>
> I agree - what if you have multiple .bb files sharing a single
> .inc and the license and/or license data are different? It
> seems to me that this data should live where the actual sources
> are being introduced, e.g. the .bb file.

I think that the main license files is going to be the same for all 
forks (well, at least for GPLed projects). And if the specific recipe 
adds some code with different license we can add LIC_FILES_CHKSUM 
entries in this specific recipe, can't we?

Regards, Ilya.

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Ilya Yanok]

Hi Darren,

09.02.2011 20:38, Darren Hart wrote:
>> +LIC_FILES_CHKSUM = "file://COPYING;md5=1707d6db1d42237583f50183a5651ecb"
>> +
>
>  From what I can tell u-boot.inc is only used by u-boot_git.bb which
> already defines LIC_FILES_CHKSUM and the appropriate GPLv2+ LICENSE
> variable.

Well, that's true. Actually I'm adding my own U-Boot recipe but want to 
use u-boot.inc.

> It seems to me the LICENSE and LIC_FILES_CHKSUM should be defined in the
> same recipe as the SRC_URI. In this case, u-boot_git.bb. Is there a
> motivating reason to add it to .inc? If so, we should also remove it

My idea was that main COPYING file will be present in virtually any 
U-Boot version, so we want LIC_FILES_CHKSUM for this file to be in 
common include file.

> from u-boot_git.bb to avoid unnecessary duplication which leads to
> confusion down the road.

Yes, I think we should remove LIC_FILES_CHKSUM from u-boot_git.bb. 
Should I remove it and repost the patch?

Regards, Ilya.

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Darren Hart]

On 02/09/2011 02:58 PM, Ilya Yanok wrote:
> Hi Darren,
>
> 09.02.2011 20:38, Darren Hart wrote:
>>> +LIC_FILES_CHKSUM =
>>> "file://COPYING;md5=1707d6db1d42237583f50183a5651ecb"
>>> +
>>
>> From what I can tell u-boot.inc is only used by u-boot_git.bb which
>> already defines LIC_FILES_CHKSUM and the appropriate GPLv2+ LICENSE
>> variable.
>
> Well, that's true. Actually I'm adding my own U-Boot recipe but want to
> use u-boot.inc.


I figured, otherwise you would have hit a build failure with 
u-boot_git.bb (since it's CHKSUM was out of date and overwrote the one 
in .inc).


>> It seems to me the LICENSE and LIC_FILES_CHKSUM should be defined in the
>> same recipe as the SRC_URI. In this case, u-boot_git.bb. Is there a
>> motivating reason to add it to .inc? If so, we should also remove it
>
> My idea was that main COPYING file will be present in virtually any
> U-Boot version, so we want LIC_FILES_CHKSUM for this file to be in
> common include file.


True, but it may suffer from line-wrapping or other subtle changes which 
will break the CHKSUMs.


>> from u-boot_git.bb to avoid unnecessary duplication which leads to
>> confusion down the road.
>
> Yes, I think we should remove LIC_FILES_CHKSUM from u-boot_git.bb.
> Should I remove it and repost the patch?


I've sent a couple patches to address this and some other existing 
issues. You were CC'd and Richard has merged with master. This should 
get us to a state that works for all known usage, holler if they cause 
you any undue pain. :-)

Thanks,

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Ilya Yanok]

Hi Darren,

10.02.2011 2:04, Darren Hart wrote:
>> My idea was that main COPYING file will be present in virtually any
>> U-Boot version, so we want LIC_FILES_CHKSUM for this file to be in
>> common include file.
>
>
> True, but it may suffer from line-wrapping or other subtle changes which
> will break the CHKSUMs.

Agreed, this can be an issue.

>>> from u-boot_git.bb to avoid unnecessary duplication which leads to
>>> confusion down the road.
>>
>> Yes, I think we should remove LIC_FILES_CHKSUM from u-boot_git.bb.
>> Should I remove it and repost the patch?
>
>
> I've sent a couple patches to address this and some other existing
> issues. You were CC'd and Richard has merged with master. This should
> get us to a state that works for all known usage, holler if they cause
> you any undue pain. :-)

Uh.. Somehow I can find these patches neither in my mailbox nor in the 
list archive... Could you please give me some link?

Regards, Ilya.

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Darren Hart]

On 02/09/2011 03:25 PM, Ilya Yanok wrote:
> Hi Darren,
>
> 10.02.2011 2:04, Darren Hart wrote:
>>> My idea was that main COPYING file will be present in virtually any
>>> U-Boot version, so we want LIC_FILES_CHKSUM for this file to be in
>>> common include file.
>>
>>
>> True, but it may suffer from line-wrapping or other subtle changes which
>> will break the CHKSUMs.
>
> Agreed, this can be an issue.
>
>>>> from u-boot_git.bb to avoid unnecessary duplication which leads to
>>>> confusion down the road.
>>>
>>> Yes, I think we should remove LIC_FILES_CHKSUM from u-boot_git.bb.
>>> Should I remove it and repost the patch?
>>
>>
>> I've sent a couple patches to address this and some other existing
>> issues. You were CC'd and Richard has merged with master. This should
>> get us to a state that works for all known usage, holler if they cause
>> you any undue pain. :-)
>
> Uh.. Somehow I can find these patches neither in my mailbox nor in the
> list archive... Could you please give me some link?
>

Hrm, this appears to be a problem on my end - I sent email from a build 
system which I thought was configured properly, the mail did make it to 
me... but I don't see it on the archives either. Thanks for pointing 
that out.

The patches are the top two here:

http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=dvhart/uboot


-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Ilya Yanok]

Hi Darren,

10.02.2011 2:38, Darren Hart wrote:
> Hrm, this appears to be a problem on my end - I sent email from a build
> system which I thought was configured properly, the mail did make it to
> me... but I don't see it on the archives either. Thanks for pointing
> that out.
>
> The patches are the top two here:
>
> http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=dvhart/uboot

Thanks! So the policy is "LIC_FILES_CHKSUM should be near SRC_URI", right?

Regards, Ilya.

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Darren Hart]

On 02/09/2011 03:43 PM, Ilya Yanok wrote:
> Hi Darren,
>
> 10.02.2011 2:38, Darren Hart wrote:
>> Hrm, this appears to be a problem on my end - I sent email from a build
>> system which I thought was configured properly, the mail did make it to
>> me... but I don't see it on the archives either. Thanks for pointing
>> that out.
>>
>> The patches are the top two here:
>>
>> http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=dvhart/uboot
>
> Thanks! So the policy is "LIC_FILES_CHKSUM should be near SRC_URI", right?

Well, Gary and I agree on that, but I haven't heard an official 
statement regarding this. It certainly seems to me to be the only 
maintainable way to do it.

Richard, any thoughts?

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 2/3] u-boot: add LIC_FILES_CHKSUM

[Richard Purdie]

On Wed, 2011-02-09 at 16:06 -0800, Darren Hart wrote:
> On 02/09/2011 03:43 PM, Ilya Yanok wrote:
> > Hi Darren,
> >
> > 10.02.2011 2:38, Darren Hart wrote:
> >> Hrm, this appears to be a problem on my end - I sent email from a build
> >> system which I thought was configured properly, the mail did make it to
> >> me... but I don't see it on the archives either. Thanks for pointing
> >> that out.
> >>
> >> The patches are the top two here:
> >>
> >> http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=dvhart/uboot
> >
> > Thanks! So the policy is "LIC_FILES_CHKSUM should be near SRC_URI", right?
> 
> Well, Gary and I agree on that, but I haven't heard an official 
> statement regarding this. It certainly seems to me to be the only 
> maintainable way to do it.
> 
> Richard, any thoughts?

I agree that the two are very linked and should generally stick
together.

Cheers,

Richard

[PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Ilya Yanok]

gcc on PowerPC is currently compiled with -Os optimization disabled so
we have to use -O2 on PowerPC instead.

Signed-off-by: Ilya Yanok <yanok@emcraft.com>
---
 meta/recipes-bsp/uboot/u-boot.inc |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
index 6492016..7694c77 100644
--- a/meta/recipes-bsp/uboot/u-boot.inc
+++ b/meta/recipes-bsp/uboot/u-boot.inc
@@ -22,7 +22,12 @@ do_compile () {
 	unset CFLAGS
 	unset CPPFLAGS
 	oe_runmake ${UBOOT_MACHINE}
-	oe_runmake all
+	# -Os is disabled on PowerPC
+	if [ ${TARGET_ARCH} == "powerpc" ] ; then
+		oe_runmake OPTFLAGS=-O2 all
+	else
+		oe_runmake all
+	fi
 }
 
 do_deploy () {
-- 
1.7.4

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Darren Hart]

On 02/08/2011 12:26 PM, Ilya Yanok wrote:
> gcc on PowerPC is currently compiled with -Os optimization disabled so
> we have to use -O2 on PowerPC instead.
>
> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
> ---
>   meta/recipes-bsp/uboot/u-boot.inc |    7 ++++++-
>   1 files changed, 6 insertions(+), 1 deletions(-)
>
> diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
> index 6492016..7694c77 100644
> --- a/meta/recipes-bsp/uboot/u-boot.inc
> +++ b/meta/recipes-bsp/uboot/u-boot.inc
> @@ -22,7 +22,12 @@ do_compile () {
>   	unset CFLAGS
>   	unset CPPFLAGS
>   	oe_runmake ${UBOOT_MACHINE}
> -	oe_runmake all
> +	# -Os is disabled on PowerPC
> +	if [ ${TARGET_ARCH} == "powerpc" ] ; then
> +		oe_runmake OPTFLAGS=-O2 all
> +	else
> +		oe_runmake all
> +	fi


This is already disabled in u-boot.inc with the following line:

# GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"

Resetting OPTFLAGS to '' removes the -Os from the variable in the 
Makefile. This isn't just a PowerPC problem, it also manifested on ARM, 
I believe it is a GCC 4.5.1 issue.

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Adrian Alonso]

[-- Attachment #1: Type: text/plain, Size: 1964 bytes --]

I observe that with current gcc for powerpc uboot only builds with
OPTFLAGS=-O2, unsetting or using other OPT level the build fails
with undefined symbols.

I override the OPTFLAGS in the dedicated recipe for uboot under meta-xilinx
BSP
but for powerpc targets in main meta layer Ilya approach does the work.

On Wed, Feb 9, 2011 at 11:42 AM, Darren Hart <dvhart@linux.intel.com> wrote:

> On 02/08/2011 12:26 PM, Ilya Yanok wrote:
>
>> gcc on PowerPC is currently compiled with -Os optimization disabled so
>> we have to use -O2 on PowerPC instead.
>>
>> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
>> ---
>>  meta/recipes-bsp/uboot/u-boot.inc |    7 ++++++-
>>  1 files changed, 6 insertions(+), 1 deletions(-)
>>
>> diff --git a/meta/recipes-bsp/uboot/u-boot.inc
>> b/meta/recipes-bsp/uboot/u-boot.inc
>> index 6492016..7694c77 100644
>> --- a/meta/recipes-bsp/uboot/u-boot.inc
>> +++ b/meta/recipes-bsp/uboot/u-boot.inc
>> @@ -22,7 +22,12 @@ do_compile () {
>>        unset CFLAGS
>>        unset CPPFLAGS
>>        oe_runmake ${UBOOT_MACHINE}
>> -       oe_runmake all
>> +       # -Os is disabled on PowerPC
>> +       if [ ${TARGET_ARCH} == "powerpc" ] ; then
>> +               oe_runmake OPTFLAGS=-O2 all
>> +       else
>> +               oe_runmake all
>> +       fi
>>
>
>
> This is already disabled in u-boot.inc with the following line:
>
> # GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
> EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"
>
> Resetting OPTFLAGS to '' removes the -Os from the variable in the Makefile.
> This isn't just a PowerPC problem, it also manifested on ARM, I believe it
> is a GCC 4.5.1 issue.
>
> --
> Darren Hart
> Intel Open Source Technology Center
> Yocto Project - Linux Kernel
>
> _______________________________________________
> poky mailing list
> poky@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/poky
>

[-- Attachment #2: Type: text/html, Size: 2820 bytes --]

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Darren Hart]

On 02/09/2011 10:05 AM, Adrian Alonso wrote:
> I observe that with current gcc for powerpc uboot only builds with
> OPTFLAGS=-O2, unsetting or using other OPT level the build fails
> with undefined symbols.
>
> I override the OPTFLAGS in the dedicated recipe for uboot under
> meta-xilinx BSP
> but for powerpc targets in main meta layer Ilya approach does the work.

I see - shouldn't be reviewing patches with a head cold. I wonder if the 
-O2 would work with ARM as well. If so, we could just set OPTFLAGS to 
-O2 instead of '' for all architectures. I'd much prefer that if 
possible, I'll test on ARM.

--
Darren

>
> On Wed, Feb 9, 2011 at 11:42 AM, Darren Hart <dvhart@linux.intel.com
> <mailto:dvhart@linux.intel.com>> wrote:
>
>     On 02/08/2011 12:26 PM, Ilya Yanok wrote:
>
>         gcc on PowerPC is currently compiled with -Os optimization
>         disabled so
>         we have to use -O2 on PowerPC instead.
>
>         Signed-off-by: Ilya Yanok<yanok@emcraft.com
>         <mailto:yanok@emcraft.com>>
>         ---
>           meta/recipes-bsp/uboot/u-boot.inc |    7 ++++++-
>           1 files changed, 6 insertions(+), 1 deletions(-)
>
>         diff --git a/meta/recipes-bsp/uboot/u-boot.inc
>         b/meta/recipes-bsp/uboot/u-boot.inc
>         index 6492016..7694c77 100644
>         --- a/meta/recipes-bsp/uboot/u-boot.inc
>         +++ b/meta/recipes-bsp/uboot/u-boot.inc
>         @@ -22,7 +22,12 @@ do_compile () {
>                 unset CFLAGS
>                 unset CPPFLAGS
>                 oe_runmake ${UBOOT_MACHINE}
>         -       oe_runmake all
>         +       # -Os is disabled on PowerPC
>         +       if [ ${TARGET_ARCH} == "powerpc" ] ; then
>         +               oe_runmake OPTFLAGS=-O2 all
>         +       else
>         +               oe_runmake all
>         +       fi
>
>
>
>     This is already disabled in u-boot.inc with the following line:
>
>     # GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
>     EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"
>
>     Resetting OPTFLAGS to '' removes the -Os from the variable in the
>     Makefile. This isn't just a PowerPC problem, it also manifested on
>     ARM, I believe it is a GCC 4.5.1 issue.
>
>     --
>     Darren Hart
>     Intel Open Source Technology Center
>     Yocto Project - Linux Kernel
>
>     _______________________________________________
>     poky mailing list
>     poky@yoctoproject.org <mailto:poky@yoctoproject.org>
>     https://lists.yoctoproject.org/listinfo/poky
>
>


-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Koen Kooi]

Op 9 feb 2011, om 18:42 heeft Darren Hart het volgende geschreven:

> On 02/08/2011 12:26 PM, Ilya Yanok wrote:
>> gcc on PowerPC is currently compiled with -Os optimization disabled so
>> we have to use -O2 on PowerPC instead.
>> 
>> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
>> ---
>>  meta/recipes-bsp/uboot/u-boot.inc |    7 ++++++-
>>  1 files changed, 6 insertions(+), 1 deletions(-)
>> 
>> diff --git a/meta/recipes-bsp/uboot/u-boot.inc b/meta/recipes-bsp/uboot/u-boot.inc
>> index 6492016..7694c77 100644
>> --- a/meta/recipes-bsp/uboot/u-boot.inc
>> +++ b/meta/recipes-bsp/uboot/u-boot.inc
>> @@ -22,7 +22,12 @@ do_compile () {
>>  	unset CFLAGS
>>  	unset CPPFLAGS
>>  	oe_runmake ${UBOOT_MACHINE}
>> -	oe_runmake all
>> +	# -Os is disabled on PowerPC
>> +	if [ ${TARGET_ARCH} == "powerpc" ] ; then
>> +		oe_runmake OPTFLAGS=-O2 all
>> +	else
>> +		oe_runmake all
>> +	fi
> 
> 
> This is already disabled in u-boot.inc with the following line:
> 
> # GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
> EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"
> 
> Resetting OPTFLAGS to '' removes the -Os from the variable in the Makefile. This isn't just a PowerPC problem, it also manifested on ARM, I believe it is a GCC 4.5.1 issue.

4.5.3 + linaro patches has similar -Os problems on armv7a at least, fwiw

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Wolfgang Denk]

Dear Koen Kooi,

In message <B4A7B800-CC7B-439C-9F36-7410D3B6D117@dominion.thruhere.net> you wrote:
> 
> 4.5.3 + linaro patches has similar -Os problems on armv7a at least, fwiw

For reference: we have a list of known compiler issues here:
http://www.denx.de/wiki/U-Boot/GccBugsAndIssues

For some bugs exist patches, but so far these have not been mainlined
yet (nor do I see all of them in the Yocto or Linaro tool chains).

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de
It seems intuitively obvious to me, which  means  that  it  might  be
wrong.                                                 -- Chris Torek

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Ilya Yanok]

Hi Darren,

09.02.2011 20:42, Darren Hart wrote:
> This is already disabled in u-boot.inc with the following line:
>
> # GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
> EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"

Yes, sorry, I've useed not up-to-date version of metadata and missed 
this fix. But it has to be '-O2' still.

> Resetting OPTFLAGS to '' removes the -Os from the variable in the
> Makefile. This isn't just a PowerPC problem, it also manifested on ARM,
> I believe it is a GCC 4.5.1 issue.

Well, the difference here is that on PowerPC we have -Os option disabled 
completely. Wolfgang tried to summarize know gcc bugs and seems like -Os 
bugs on ARM and PowerPC have different nature (it looks like bugs on ARM 
are already fixed while on the ones PowerPC are not).
Then I think it's a bad idea to remove -Os everywhere: U-Boot really 
wants to be built with -Os optimization so we might want to change 
optimization level to -O2 only for targets where it's absolutely 
required. Actually I was able to build and successfully run U-Boot on 
beagleboard with -Os so it doesn't look like it's completely broken on ARM.

Regards, Ilya.

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Darren Hart]

On 02/09/2011 03:21 PM, Ilya Yanok wrote:
> Hi Darren,
>
> 09.02.2011 20:42, Darren Hart wrote:
>> This is already disabled in u-boot.inc with the following line:
>>
>> # GCC 4.5.1 builds unusable binaries using -Os, remove it from OPTFLAGS
>> EXTRA_OEMAKE = "CROSS_COMPILE=${TARGET_PREFIX} OPTFLAGS=''"
>
> Yes, sorry, I've useed not up-to-date version of metadata and missed
> this fix. But it has to be '-O2' still.

Understood, -O2 is in master now for all arches.

>
>> Resetting OPTFLAGS to '' removes the -Os from the variable in the
>> Makefile. This isn't just a PowerPC problem, it also manifested on ARM,
>> I believe it is a GCC 4.5.1 issue.
>
> Well, the difference here is that on PowerPC we have -Os option disabled
> completely. Wolfgang tried to summarize know gcc bugs and seems like -Os
> bugs on ARM and PowerPC have different nature (it looks like bugs on ARM
> are already fixed while on the ones PowerPC are not).
> Then I think it's a bad idea to remove -Os everywhere: U-Boot really
> wants to be built with -Os optimization so we might want to change
> optimization level to -O2 only for targets where it's absolutely
> required. Actually I was able to build and successfully run U-Boot on
> beagleboard with -Os so it doesn't look like it's completely broken on ARM.

Interesting - is this using the current recipe in poky? It was building 
for the beagleboard that I hit the bug which caused me to clear -Os from 
the EXTRA_OEMAKE arguments.


-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel

Re: [PATCH 3/3] u-boot: compile with -O2 on PowerPC

[Ilya Yanok]

Hi Darren,

10.02.2011 2:34, Darren Hart wrote:
>> Then I think it's a bad idea to remove -Os everywhere: U-Boot really
>> wants to be built with -Os optimization so we might want to change
>> optimization level to -O2 only for targets where it's absolutely
>> required. Actually I was able to build and successfully run U-Boot on
>> beagleboard with -Os so it doesn't look like it's completely broken on
>> ARM.
>
> Interesting - is this using the current recipe in poky? It was building
> for the beagleboard that I hit the bug which caused me to clear -Os from
> the EXTRA_OEMAKE arguments.

I'm using slightly outdated Poky tree (commit 4cd1118) with some my 
additions and my own U-Boot 2011.03-rc1 recipe (nothing special, I've 
just changed SRC_URI and supported machine targets). I'll try to rebase 
to master and retest in a couple of days.

Regards, Ilya.

Re: [PATCH 1/3] linux: add LIC_FILES_CHKSUM

[Richard Purdie]

On Tue, 2011-02-08 at 21:26 +0100, Ilya Yanok wrote:
> Linux is licensed under GPL so add appropriate LIC_FILES_CHKSUM to
> linux.inc file.
> 
> Signed-off-by: Ilya Yanok <yanok@emcraft.com>

Thanks for these, I merged patches 1-3 into master. I also updated the
LICENSE field to specify the GPL version for these recipes.

Cheers,

Richard

Re: [PATCH 1/3] linux: add LIC_FILES_CHKSUM

[Darren Hart]

On 02/08/2011 12:26 PM, Ilya Yanok wrote:
> Linux is licensed under GPL so add appropriate LIC_FILES_CHKSUM to
> linux.inc file.
>
> Signed-off-by: Ilya Yanok<yanok@emcraft.com>
> ---
>   meta/recipes-kernel/linux/linux.inc |    2 ++
>   1 files changed, 2 insertions(+), 0 deletions(-)
>
> diff --git a/meta/recipes-kernel/linux/linux.inc b/meta/recipes-kernel/linux/linux.inc
> index 93859bb..c23d90f 100644
> --- a/meta/recipes-kernel/linux/linux.inc
> +++ b/meta/recipes-kernel/linux/linux.inc
> @@ -2,6 +2,8 @@ DESCRIPTION = "Linux Kernel"
>   SECTION = "kernel"
>   LICENSE = "GPL"
>
> +LIC_FILES_CHKSUM = "file://COPYING;md5=d7810fab7487fb0aad327b76f1be7cd7"
> +

This is consistent with other .inc files (such as linux-yocto.inc). I'm 
fine with it, but wanted to mention that linux.inc is going to be going 
trough some serious pruning and refactoring in the near future.

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel