[U-Boot] undefined reference to `environment'

[U-Boot] undefined reference to `environment'

[Joe Hershberger]

Hi Simon / Tom,

It seems the SH4 build is failing with the current trunk. Or at least
my test build is failing and the change seems to have nothing to do
with the failure.

https://travis-ci.org/jhershbe/u-boot/jobs/269330102

Cheers,
-Joe

[U-Boot] undefined reference to `environment'

[Simon Glass]

+Thomas

Hi Joe,

On 29 August 2017 at 06:59, Joe Hershberger <joe.hershberger@ni.com> wrote:
> Hi Simon / Tom,
>
> It seems the SH4 build is failing with the current trunk. Or at least
> my test build is failing and the change seems to have nothing to do
> with the failure.
>
> https://travis-ci.org/jhershbe/u-boot/jobs/269330102

A bisect shows this came in with:

f40ad66f (refs/bisect/bad) arch/sh: don't bring common/env_embedded.o
into the link

Thomas, could you please take a look?

>
> Cheers,
> -Joe

Regards,
Simon

[U-Boot] undefined reference to `environment'

[Thomas Petazzoni]

Hello,

On Wed, 30 Aug 2017 04:16:39 +0800, Simon Glass wrote:

> Hi Joe,
> 
> On 29 August 2017 at 06:59, Joe Hershberger <joe.hershberger@ni.com> wrote:
> > Hi Simon / Tom,
> >
> > It seems the SH4 build is failing with the current trunk. Or at least
> > my test build is failing and the change seems to have nothing to do
> > with the failure.
> >
> > https://travis-ci.org/jhershbe/u-boot/jobs/269330102  
> 
> A bisect shows this came in with:
> 
> f40ad66f (refs/bisect/bad) arch/sh: don't bring common/env_embedded.o
> into the link
> 
> Thomas, could you please take a look?

Sure, I'll have a look. I guess I should be able to reproduce by
building ms7722se_defconfig ?

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[U-Boot] undefined reference to `environment'

[Thomas Petazzoni]

Hello,

On Wed, 30 Aug 2017 04:16:39 +0800, Simon Glass wrote:

> A bisect shows this came in with:
> 
> f40ad66f (refs/bisect/bad) arch/sh: don't bring common/env_embedded.o
> into the link
> 
> Thomas, could you please take a look?

env/embedded.o is missing in the link. It is added to extra-y by:

extra-$(CONFIG_ENV_IS_IN_FLASH) += embedded.o

I dumped extra-y right after this line, and it does contain embedded.o.
However, it does not get added to the link stage.

Why is this an extra-$() rather than an obj-$() ?

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[U-Boot] undefined reference to `environment'

[Tom Rini]

On Tue, Aug 29, 2017 at 4:50 PM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Hello,
>
> On Wed, 30 Aug 2017 04:16:39 +0800, Simon Glass wrote:
>
>> A bisect shows this came in with:
>>
>> f40ad66f (refs/bisect/bad) arch/sh: don't bring common/env_embedded.o
>> into the link
>>
>> Thomas, could you please take a look?
>
> env/embedded.o is missing in the link. It is added to extra-y by:
>
> extra-$(CONFIG_ENV_IS_IN_FLASH) += embedded.o
>
> I dumped extra-y right after this line, and it does contain embedded.o.
> However, it does not get added to the link stage.
>
> Why is this an extra-$() rather than an obj-$() ?

So there's something extra funny going on in the linking.  See
https://patchwork.ozlabs.org/patch/806263/ which I have included, but
haven't been able to push out.

-- 
Tom

[U-Boot] undefined reference to `environment'

[Thomas Petazzoni]

Hello,

On Tue, 29 Aug 2017 17:18:24 -0400, Tom Rini wrote:

> So there's something extra funny going on in the linking.  See
> https://patchwork.ozlabs.org/patch/806263/ which I have included, but
> haven't been able to push out.

At this point, I don't think it's the correct fix. I don't see why it
should be the responsibility of the linker script to include
env/embedded.o. Almost no other architecture has env/embedded.o
referenced in its linker script. Only a few m68k platforms, with custom
linker script in their include/configs/.h file include env/embedded.o
in their linker script, plus one Freescale platform.

Again, changing extra- to obj- in env/Makefile also fixes the issue.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[U-Boot] undefined reference to `environment'

[Bin Meng]

On Wed, Aug 30, 2017 at 6:07 AM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Hello,
>
> On Tue, 29 Aug 2017 17:18:24 -0400, Tom Rini wrote:
>
>> So there's something extra funny going on in the linking.  See
>> https://patchwork.ozlabs.org/patch/806263/ which I have included, but
>> haven't been able to push out.
>
> At this point, I don't think it's the correct fix. I don't see why it
> should be the responsibility of the linker script to include
> env/embedded.o. Almost no other architecture has env/embedded.o
> referenced in its linker script. Only a few m68k platforms, with custom
> linker script in their include/configs/.h file include env/embedded.o
> in their linker script, plus one Freescale platform.
>
> Again, changing extra- to obj- in env/Makefile also fixes the issue.

Yep, looks to me obj- is the correct way to fix.

Regards,
Bin

[U-Boot] undefined reference to `environment'

[Thomas Petazzoni]

Hello,

On Tue, 29 Aug 2017 17:18:24 -0400, Tom Rini wrote:

> > env/embedded.o is missing in the link. It is added to extra-y by:
> >
> > extra-$(CONFIG_ENV_IS_IN_FLASH) += embedded.o
> >
> > I dumped extra-y right after this line, and it does contain embedded.o.
> > However, it does not get added to the link stage.
> >
> > Why is this an extra-$() rather than an obj-$() ?  
> 
> So there's something extra funny going on in the linking.  See
> https://patchwork.ozlabs.org/patch/806263/ which I have included, but
> haven't been able to push out.

How did you end up solving this problem? As I said, I don't think
https://patchwork.ozlabs.org/patch/806263/ is the correct solution.

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[U-Boot] undefined reference to `environment'

[Tom Rini]

On Thu, Sep 14, 2017 at 12:29:11AM +0200, Thomas Petazzoni wrote:

> Hello,
> 
> On Tue, 29 Aug 2017 17:18:24 -0400, Tom Rini wrote:
> 
> > > env/embedded.o is missing in the link. It is added to extra-y by:
> > >
> > > extra-$(CONFIG_ENV_IS_IN_FLASH) += embedded.o
> > >
> > > I dumped extra-y right after this line, and it does contain embedded.o.
> > > However, it does not get added to the link stage.
> > >
> > > Why is this an extra-$() rather than an obj-$() ?  
> > 
> > So there's something extra funny going on in the linking.  See
> > https://patchwork.ozlabs.org/patch/806263/ which I have included, but
> > haven't been able to push out.
> 
> How did you end up solving this problem? As I said, I don't think
> https://patchwork.ozlabs.org/patch/806263/ is the correct solution.

It wasn't my favorite patch but I wouldn't have let the original in if
I had noticed the fail-to-build in the first place.  So coming up with a
better rework (and understanding of) why only that combination of
settings was failing without forcing env/embedded.o being linked in is
on the TODO list somewhere.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20170913/dd5cb280/attachment.sig>