[Buildroot] board/ci20 uboot problems due to fdt headers

[Buildroot] board/ci20 uboot problems due to fdt headers

[Paulo Matos]

Hello,

I am attempting to build ci20 board on an archlinux host.
I am getting several errors build uboot that look like:


/usr/bin/gcc -O2 -isystem
/home/pmatos/Projects/igalia/buildroot/output/host/include -Wall
-Wstrict-prototypes -O2 -fomit-frame-pointer -include
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include/libfdt_env.h
-idirafter
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include
-idirafter
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include2
-idirafter
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include
-I
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/lib/libfdt
-I
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/tools
-DCONFIG_SYS_TEXT_BASE=0x80000000 -DUSE_HOSTCC -D__KERNEL_STRICT_NAMES
-D_GNU_SOURCE   -o fit_image.o fit_image.c -c
In file included from ../include/libfdt.h:54,
                 from fdt_host.h:11,
                 from mkimage.h:22,
                 from aisimage.c:8:
/usr/include/libfdt_env.h:71:30: error: conflicting types for ?fdt64_t?
 typedef uint64_t FDT_BITWISE fdt64_t;
                              ^~~~~~~
In file included from <command-line>:
/home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include/libfdt_env.h:19:16:
note: previous declaration of ?fdt64_t? was here
 typedef __be64 fdt64_t;
                ^~~~~~~


This seems to be related to a previous thread that seemed to end up
saying 'we have fixed that'.

http://lists.busybox.net/pipermail/buildroot/2018-April/217653.html

Does anyone know what the status of this is? As far as I understood the
problem is related to possible include headers installed in the host
clashing with the ones distributed with buildroot.

-- 
Paulo Matos

[Buildroot] board/ci20 uboot problems due to fdt headers

[Arnout Vandecappelle]

On 28/05/2019 15:57, Paulo Matos wrote:
> Hello,
> 
> I am attempting to build ci20 board on an archlinux host.
> I am getting several errors build uboot that look like:
> 
> 
> /usr/bin/gcc -O2 -isystem
> /home/pmatos/Projects/igalia/buildroot/output/host/include -Wall
> -Wstrict-prototypes -O2 -fomit-frame-pointer -include
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include/libfdt_env.h
> -idirafter
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include
> -idirafter
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include2
> -idirafter
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include
> -I
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/lib/libfdt
> -I
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/tools
> -DCONFIG_SYS_TEXT_BASE=0x80000000 -DUSE_HOSTCC -D__KERNEL_STRICT_NAMES
> -D_GNU_SOURCE   -o fit_image.o fit_image.c -c
> In file included from ../include/libfdt.h:54,
>                  from fdt_host.h:11,
>                  from mkimage.h:22,
>                  from aisimage.c:8:
> /usr/include/libfdt_env.h:71:30: error: conflicting types for ?fdt64_t?
>  typedef uint64_t FDT_BITWISE fdt64_t;
>                               ^~~~~~~
> In file included from <command-line>:
> /home/pmatos/Projects/igalia/buildroot/output/build/uboot-dd3c1b95dac7d10b2ca5806f65e5c1050d7dd0fa/include/libfdt_env.h:19:16:
> note: previous declaration of ?fdt64_t? was here
>  typedef __be64 fdt64_t;
>                 ^~~~~~~
> 
> 
> This seems to be related to a previous thread that seemed to end up
> saying 'we have fixed that'.
> 
> http://lists.busybox.net/pipermail/buildroot/2018-April/217653.html
> 
> Does anyone know what the status of this is? As far as I understood the
> problem is related to possible include headers installed in the host
> clashing with the ones distributed with buildroot.

 Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
Buildroot version are you using? Anything more recent than 2018.02.1 should no
longer have this issue.


 Regards,
 Arnout

[Buildroot] board/ci20 uboot problems due to fdt headers

[Paulo Matos]

On 28/05/2019 20:22, Arnout Vandecappelle wrote:
>
>  Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
> Buildroot version are you using? Anything more recent than 2018.02.1 should no
> longer have this issue.
>


Apologies for not mentioning it but I am using tip of master, which is
why I was surprised when I saw this error come up again. Not doing
anything special.

On arch `pacman -Ss dtc` installs /usr/include/libfdt.h, which seems to
conflict with the ones from uboot.

Should I report this?

Regards,
-- 
Paulo Matos

[Buildroot] board/ci20 uboot problems due to fdt headers

[Arnout Vandecappelle]

On 29/05/2019 08:54, Paulo Matos wrote:
> 
> 
> On 28/05/2019 20:22, Arnout Vandecappelle wrote:
>>
>>  Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
>> Buildroot version are you using? Anything more recent than 2018.02.1 should no
>> longer have this issue.
>>
> 
> 
> Apologies for not mentioning it but I am using tip of master, which is
> why I was surprised when I saw this error come up again. Not doing
> anything special.
> 
> On arch `pacman -Ss dtc` installs /usr/include/libfdt.h, which seems to
> conflict with the ones from uboot.
> 
> Should I report this?

 You did :-)

 It can indeed easily be reproduced with the ci20 defconfig.

 The issue is that in U-Boot commit 01286329b27b27eaeda045b469d41b1d9fce545a,
$(SRCTREE) was renamed to $(srctree). So, any commit before that (i.e. any
version before 2014.04) will not match the sed pattern from our uboot.mk and
will not be fixed.

 The solution is to add an addition substitution pattern in uboot.mk that does
the same on SRCTREE instead of srctree.

 In addition, commit 0de71d507157c4bd4fddcd3a419140d2b986eed2 moved it from
libfdt to lib/libfdt. In other words, before 2010.06, the pattern matches even
less... But that's easily fixed by making the lib/ part optional.

 Oh, but actually, the fix simply doesn't work on older U-Boot versions -
Buildroot commit f437bf547ca4484 disables it if scripts/dtc/libfdt doesn't
exist... If that directory doesn't exist, the bundled libfdt.h, libfdt_env.h and
fdt.h reside in ${srctree}/include. However, that directory also contains a
bunch of files that may collide with the files from /usr/include that we want
for host compilation...

 So, I think the solution might be something like this:

define UBOOT_FIXUP_LIBFDT_INCLUDE
	$(Q)if [ -d $(@D)/scripts/dtc/libfdt ]; then \
		mkdir -p $(@D)/scripts/dtc/libfdt; \
		cd $(@D)/scripts/dts/libfdt; \
		ln -s ../../../include/fdt.h .; \
		ln -s ../../../include/libfdt*.h .; \
	fi
	$(Q)$(SED) \
		's%-I$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
		's%-I$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
		$(@D)/tools/Makefile
endef

 Of course, this was completely untested. Any takers to create a patch from this?

 Regards,
 Arnout

[Buildroot] board/ci20 uboot problems due to fdt headers

[Giulio Benetti]

Hello Arnout, Pablo,

Il 29/05/2019 22:50, Arnout Vandecappelle ha scritto:
> 
> 
> On 29/05/2019 08:54, Paulo Matos wrote:
>>
>>
>> On 28/05/2019 20:22, Arnout Vandecappelle wrote:
>>>
>>>   Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
>>> Buildroot version are you using? Anything more recent than 2018.02.1 should no
>>> longer have this issue.
>>>
>>
>>
>> Apologies for not mentioning it but I am using tip of master, which is
>> why I was surprised when I saw this error come up again. Not doing
>> anything special.
>>
>> On arch `pacman -Ss dtc` installs /usr/include/libfdt.h, which seems to
>> conflict with the ones from uboot.
>>
>> Should I report this?
> 
>   You did :-)
> 
>   It can indeed easily be reproduced with the ci20 defconfig.
> 
>   The issue is that in U-Boot commit 01286329b27b27eaeda045b469d41b1d9fce545a,
> $(SRCTREE) was renamed to $(srctree). So, any commit before that (i.e. any
> version before 2014.04) will not match the sed pattern from our uboot.mk and
> will not be fixed.
> 
>   The solution is to add an addition substitution pattern in uboot.mk that does
> the same on SRCTREE instead of srctree.
> 
>   In addition, commit 0de71d507157c4bd4fddcd3a419140d2b986eed2 moved it from
> libfdt to lib/libfdt. In other words, before 2010.06, the pattern matches even
> less... But that's easily fixed by making the lib/ part optional.
> 
>   Oh, but actually, the fix simply doesn't work on older U-Boot versions -
> Buildroot commit f437bf547ca4484 disables it if scripts/dtc/libfdt doesn't
> exist... If that directory doesn't exist, the bundled libfdt.h, libfdt_env.h and
> fdt.h reside in ${srctree}/include. However, that directory also contains a
> bunch of files that may collide with the files from /usr/include that we want
> for host compilation...
> 
>   So, I think the solution might be something like this:
> 
> define UBOOT_FIXUP_LIBFDT_INCLUDE
> 	$(Q)if [ -d $(@D)/scripts/dtc/libfdt ]; then \
> 		mkdir -p $(@D)/scripts/dtc/libfdt; \
> 		cd $(@D)/scripts/dts/libfdt; \
> 		ln -s ../../../include/fdt.h .; \
> 		ln -s ../../../include/libfdt*.h .; \
> 	fi
> 	$(Q)$(SED) \
> 		's%-I$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
> 		's%-I$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
> 		$(@D)/tools/Makefile
> endef
> 
>   Of course, this was completely untested. Any takers to create a patch from this?

Here I am :)
Going to work on it soon.

-- 
Giulio Benetti
CTO

MICRONOVA SRL
Sede: Via A. Niedda 3 - 35010 Vigonza (PD)
Tel. 049/8931563 - Fax 049/8931346
Cod.Fiscale - P.IVA 02663420285
Capitale Sociale ? 26.000 i.v.
Iscritta al Reg. Imprese di Padova N. 02663420285
Numero R.E.A. 258642

[Buildroot] board/ci20 uboot problems due to fdt headers

[Thomas De Schampheleire]

El mi?., 29 may. 2019 a las 22:55, Giulio Benetti
(<giulio.benetti@micronovasrl.com>) escribi?:
>
> Hello Arnout, Pablo,
>
> Il 29/05/2019 22:50, Arnout Vandecappelle ha scritto:
> >
> >
> > On 29/05/2019 08:54, Paulo Matos wrote:
> >>
> >>
> >> On 28/05/2019 20:22, Arnout Vandecappelle wrote:
> >>>
> >>>   Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
> >>> Buildroot version are you using? Anything more recent than 2018.02.1 should no
> >>> longer have this issue.
> >>>
> >>
> >>
> >> Apologies for not mentioning it but I am using tip of master, which is
> >> why I was surprised when I saw this error come up again. Not doing
> >> anything special.
> >>
> >> On arch `pacman -Ss dtc` installs /usr/include/libfdt.h, which seems to
> >> conflict with the ones from uboot.
> >>
> >> Should I report this?
> >
> >   You did :-)
> >
> >   It can indeed easily be reproduced with the ci20 defconfig.
> >
> >   The issue is that in U-Boot commit 01286329b27b27eaeda045b469d41b1d9fce545a,
> > $(SRCTREE) was renamed to $(srctree). So, any commit before that (i.e. any
> > version before 2014.04) will not match the sed pattern from our uboot.mk and
> > will not be fixed.
> >
> >   The solution is to add an addition substitution pattern in uboot.mk that does
> > the same on SRCTREE instead of srctree.
> >
> >   In addition, commit 0de71d507157c4bd4fddcd3a419140d2b986eed2 moved it from
> > libfdt to lib/libfdt. In other words, before 2010.06, the pattern matches even
> > less... But that's easily fixed by making the lib/ part optional.
> >
> >   Oh, but actually, the fix simply doesn't work on older U-Boot versions -
> > Buildroot commit f437bf547ca4484 disables it if scripts/dtc/libfdt doesn't
> > exist... If that directory doesn't exist, the bundled libfdt.h, libfdt_env.h and
> > fdt.h reside in ${srctree}/include. However, that directory also contains a
> > bunch of files that may collide with the files from /usr/include that we want
> > for host compilation...
> >
> >   So, I think the solution might be something like this:
> >
> > define UBOOT_FIXUP_LIBFDT_INCLUDE
> >       $(Q)if [ -d $(@D)/scripts/dtc/libfdt ]; then \
> >               mkdir -p $(@D)/scripts/dtc/libfdt; \
> >               cd $(@D)/scripts/dts/libfdt; \
> >               ln -s ../../../include/fdt.h .; \
> >               ln -s ../../../include/libfdt*.h .; \
> >       fi
> >       $(Q)$(SED) \
> >               's%-I$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
> >               's%-I$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
> >               $(@D)/tools/Makefile
> > endef
> >
> >   Of course, this was completely untested. Any takers to create a patch from this?

I encountered the same problem with a u-boot based on 2017.05.
I had to modify the suggestion above slightly:
- the if check should be negated
- typo 'dts' -> 'dtc' in the cd
- extra header file libfdt_internal.h
- escaping the '$' which has a regex meaning inside sed
- addition of '-e' on the second sed expression, first one is already
part of $(SED).

Following works for me:

define UBOOT_FIXUP_LIBFDT_INCLUDE
        $(Q)if [ ! -d $(@D)/scripts/dtc/libfdt ]; then \
                mkdir -p $(@D)/scripts/dtc/libfdt; \
                cd $(@D)/scripts/dtc/libfdt; \
                ln -s ../../../include/fdt.h .; \
                ln -s ../../../include/libfdt*.h .; \
                ln -s ../../../lib/libfdt/libfdt_internal.h .; \
        fi
        $(Q)$(SED) \

's%-I\$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
                -e
's%-I\$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%'
\
                $(@D)/tools/Makefile
endef


The seemingly unbalanced '-e' could be avoided by moving the two
substitutions in one command, separated with semicolon, as follows:

        $(Q)$(SED) \

's%-I\$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%; \

s%-I\$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
                $(@D)/tools/Makefile


>
> Here I am :)
> Going to work on it soon.

Giulio, will you test this modified change in your case?

Thanks,
Thomas

[Buildroot] board/ci20 uboot problems due to fdt headers

[Giulio Benetti]

Hello Thomas,

Il 04/06/2019 14:37, Thomas De Schampheleire ha scritto:
> El mi?., 29 may. 2019 a las 22:55, Giulio Benetti
> (<giulio.benetti@micronovasrl.com>) escribi?:
>>
>> Hello Arnout, Pablo,
>>
>> Il 29/05/2019 22:50, Arnout Vandecappelle ha scritto:
>>>
>>>
>>> On 29/05/2019 08:54, Paulo Matos wrote:
>>>>
>>>>
>>>> On 28/05/2019 20:22, Arnout Vandecappelle wrote:
>>>>>
>>>>>    Indeed, it was fixed by commit bf733342324 and its follow-up f437bf547c. Which
>>>>> Buildroot version are you using? Anything more recent than 2018.02.1 should no
>>>>> longer have this issue.
>>>>>
>>>>
>>>>
>>>> Apologies for not mentioning it but I am using tip of master, which is
>>>> why I was surprised when I saw this error come up again. Not doing
>>>> anything special.
>>>>
>>>> On arch `pacman -Ss dtc` installs /usr/include/libfdt.h, which seems to
>>>> conflict with the ones from uboot.
>>>>
>>>> Should I report this?
>>>
>>>    You did :-)
>>>
>>>    It can indeed easily be reproduced with the ci20 defconfig.
>>>
>>>    The issue is that in U-Boot commit 01286329b27b27eaeda045b469d41b1d9fce545a,
>>> $(SRCTREE) was renamed to $(srctree). So, any commit before that (i.e. any
>>> version before 2014.04) will not match the sed pattern from our uboot.mk and
>>> will not be fixed.
>>>
>>>    The solution is to add an addition substitution pattern in uboot.mk that does
>>> the same on SRCTREE instead of srctree.
>>>
>>>    In addition, commit 0de71d507157c4bd4fddcd3a419140d2b986eed2 moved it from
>>> libfdt to lib/libfdt. In other words, before 2010.06, the pattern matches even
>>> less... But that's easily fixed by making the lib/ part optional.
>>>
>>>    Oh, but actually, the fix simply doesn't work on older U-Boot versions -
>>> Buildroot commit f437bf547ca4484 disables it if scripts/dtc/libfdt doesn't
>>> exist... If that directory doesn't exist, the bundled libfdt.h, libfdt_env.h and
>>> fdt.h reside in ${srctree}/include. However, that directory also contains a
>>> bunch of files that may collide with the files from /usr/include that we want
>>> for host compilation...
>>>
>>>    So, I think the solution might be something like this:
>>>
>>> define UBOOT_FIXUP_LIBFDT_INCLUDE
>>>        $(Q)if [ -d $(@D)/scripts/dtc/libfdt ]; then \
>>>                mkdir -p $(@D)/scripts/dtc/libfdt; \
>>>                cd $(@D)/scripts/dts/libfdt; \
>>>                ln -s ../../../include/fdt.h .; \
>>>                ln -s ../../../include/libfdt*.h .; \
>>>        fi
>>>        $(Q)$(SED) \
>>>                's%-I$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
>>>                's%-I$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
>>>                $(@D)/tools/Makefile
>>> endef
>>>
>>>    Of course, this was completely untested. Any takers to create a patch from this?
> 
> I encountered the same problem with a u-boot based on 2017.05.
> I had to modify the suggestion above slightly:
> - the if check should be negated
> - typo 'dts' -> 'dtc' in the cd
> - extra header file libfdt_internal.h
> - escaping the '$' which has a regex meaning inside sed
> - addition of '-e' on the second sed expression, first one is already
> part of $(SED).
> 
> Following works for me:
> 
> define UBOOT_FIXUP_LIBFDT_INCLUDE
>          $(Q)if [ ! -d $(@D)/scripts/dtc/libfdt ]; then \
>                  mkdir -p $(@D)/scripts/dtc/libfdt; \
>                  cd $(@D)/scripts/dtc/libfdt; \
>                  ln -s ../../../include/fdt.h .; \
>                  ln -s ../../../include/libfdt*.h .; \
>                  ln -s ../../../lib/libfdt/libfdt_internal.h .; \
>          fi
>          $(Q)$(SED) \
> 
> 's%-I\$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%' \
>                  -e
> 's%-I\$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%'
> \
>                  $(@D)/tools/Makefile
> endef
> 
> 
> The seemingly unbalanced '-e' could be avoided by moving the two
> substitutions in one command, separated with semicolon, as follows:
> 
>          $(Q)$(SED) \
> 
> 's%-I\$$(srctree)/lib/libfdt%-I$$(srctree)/scripts/dtc/libfdt%; \
> 
> s%-I\$$(SRCTREE)\(/lib\)\?/libfdt%-I$$(SRCTREE)/scripts/dtc/libfdt%' \
>                  $(@D)/tools/Makefile
> 
> 
>>
>> Here I am :)
>> Going to work on it soon.
> 
> Giulio, will you test this modified change in your case?

Unfortunately I don't have time for this at the moment and I barely read 
what's the topic.
So feel free to take over this patch.

Best regards
-- 
Giulio Benetti
CTO

MICRONOVA SRL
Sede: Via A. Niedda 3 - 35010 Vigonza (PD)
Tel. 049/8931563 - Fax 049/8931346
Cod.Fiscale - P.IVA 02663420285
Capitale Sociale ? 26.000 i.v.
Iscritta al Reg. Imprese di Padova N. 02663420285
Numero R.E.A. 258642