[Buildroot] Build failure in icu-59.1

[Buildroot] Build failure in icu-59.1

[Romain Naour]

Hi Ross,

Please, keep the list in Cc.

Le 30/09/2017 ? 14:33, grunpferd at netscape.net a ?crit?:
> 
> G'day Romain,
> 
> ?
>>>> glibc has been updated to 2.26 recently and the xlocale.h header file has been
>>>> removed.
>>>> See https://sourceware.org/glibc/wiki/Release/2.26#Packaging_Changes
>>>>?
>>>> I tested several packages with this new version but it's difficult to test every
>>>> single packages.
>>>?
>>>> Can you report to the upstream project or if it's already fixed upstream, send
>>>> the patch to the ml?
>>>?
>>> I have have lodged bug report with ICU project. As far as I can tell it has not
>>> been fixed or reported.
>>
>>?http://bugs.icu-project.org/trac/ticket/13385
> 
>>>?
>>> I think your analysis is correct, the advance to glibc 2.26 has flushed out this
>>> problem.
>>> How this gets fixed may depend on the ICU project that has icu 59.1 as their
>>> latest release.
>>> In the mean time buildroot might have to handle this by itself.
> 
>> glibc suggest to replace xlocale.h by locale.h.
>> Can you send a patch doing so ?
>>
>> See http://nightly.buildroot.org/manual.html#submitting-patches
> 
> I have checked that using locale.h instead of xlocale.h does indeed fix the issue.
> 
> Not quite sure how to tackle the patch however!
> 
> Currently digitlst.cpp
> 
> uses:
> #if U_USE_STRTOD_L
> # if U_PLATFORM_USES_ONLY_WIN32_API || U_PLATFORM == U_PF_CYGWIN
> # ??include <locale.h>
> # else
> # ??include <xlocale.h>
> # endif
> #endif
> 
> 
> so I presume something like:
> 
> #if U_USE_STRTOD_L
> # if U_PLATFORM_USES_ONLY_WIN32_API || U_PLATFORM == U_PF_CYGWIN?
> ||??__GLIBC_PREREQ(2, 26)
> # ??include <locale.h>
> # else
> # ??include <xlocale.h>
> # endif
> #endif
> 
> I am just not sure how to handle that GLIBC_2.26 part.
> Can we use?__GLIBC_PREREQ(?maj, min)?
> 
> Keeping in mind that glibc <= 2.25 will work while using xlocale.h

xlocale.h shouldn't be used by icu in the first place.

Maybe you can try this patch:
https://git.yoctoproject.org/cgit.cgi/poky/plain/meta/recipes-support/icu/icu/0001-i18n-Drop-include-xlocale.h.patch

Best regards,
Romain

> 
> Regards,
> 
> Ross
> 
> 

[Buildroot] Build failure in icu-59.1

[grunpferd at netscape.net]

G'day Romain,

Le 30/09/2017 ? 14:33, grunpferd at netscape.net a ?crit :
> 
> G'day Romain,
> 
>  
>>>> glibc has been updated to 2.26 recently and the xlocale.h header file has been
>>>> removed.
>>>> See https://sourceware.org/glibc/wiki/Release/2.26#Packaging_Changes
>>>> 
>>>> I tested several packages with this new version but it's difficult to test every
>>>> single packages.
>>> 
>>>> Can you report to the upstream project or if it's already fixed upstream, send
>>>> the patch to the ml?
>>> 
>>> I have have lodged bug report with ICU project. As far as I can tell it has not
>>> been fixed or reported.
>>
>> http://bugs.icu-project.org/trac/ticket/13385
> 
>>> 
>>> I think your analysis is correct, the advance to glibc 2.26 has flushed out this
>>> problem.
>>> How this gets fixed may depend on the ICU project that has icu 59.1 as their
>>> latest release.
>>> In the mean time buildroot might have to handle this by itself.
> 
>> glibc suggest to replace xlocale.h by locale.h.
>> Can you send a patch doing so ?
>>
>> See http://nightly.buildroot.org/manual.html#submitting-patches
> 
> I have checked that using locale.h instead of xlocale.h does indeed fix the issue.
> 
> Not quite sure how to tackle the patch however!
> 
> Currently digitlst.cpp
> 
> uses:
> #if U_USE_STRTOD_L
> # if U_PLATFORM_USES_ONLY_WIN32_API || U_PLATFORM == U_PF_CYGWIN
> #   include <locale.h>
> # else
> #   include <xlocale.h>
> # endif
> #endif
> 
> 
> so I presume something like:
> 
> #if U_USE_STRTOD_L
> # if U_PLATFORM_USES_ONLY_WIN32_API || U_PLATFORM == U_PF_CYGWIN 
> ||  __GLIBC_PREREQ(2, 26)
> #   include <locale.h>
> # else
> #   include <xlocale.h>
> # endif
> #endif
> 
> I am just not sure how to handle that GLIBC_2.26 part.
> Can we use __GLIBC_PREREQ( maj, min)?
> 
> Keeping in mind that glibc <= 2.25 will work while using xlocale.h

xlocale.h shouldn't be used by icu in the first place.

Maybe you can try this patch:
https://git.yoctoproject.org/cgit.cgi/poky/plain/meta/recipes-support/icu/icu/0001-i18n-Drop-include-xlocale.h.patch

Yes that patch does indeed remove xlocale.h and simply replaces it with locale.h


The resultant code then builds as expected.
So picking up that patch looks like the right thing to do.


Regards,


Ross
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20170930/00c6198d/attachment.html>

[Buildroot] Build failure in icu-59.1

[Romain Naour]

Hi,

Le 30/09/2017 ? 12:10, grunpferd at netscape.net a ?crit?:
> 
> Thanks Romain,
> 
>> glibc has been updated to 2.26 recently and the xlocale.h header file has been
>> removed.
>> See https://sourceware.org/glibc/wiki/Release/2.26#Packaging_Changes
>>?
>> I tested several packages with this new version but it's difficult to test every
>> single packages.
> 
>> Can you report to the upstream project or if it's already fixed upstream, send
>> the patch to the ml?
> 
> I have have lodged bug report with ICU project. As far as I can tell it has not
> been fixed or reported.

http://bugs.icu-project.org/trac/ticket/13385

> 
> I think your analysis is correct, the advance to glibc 2.26 has flushed out this
> problem.
> How this gets fixed may depend on the ICU project that has icu 59.1 as their
> latest release.
> In the mean time buildroot might have to handle this by itself.

glibc suggest to replace xlocale.h by locale.h.
Can you send a patch doing so ?

See http://nightly.buildroot.org/manual.html#submitting-patches

Best regards,
Romain

> 
> 
> Regards,
> 
> Ross
> 

[Buildroot] Build failure in icu-59.1

[grunpferd at netscape.net]

Thanks Romain,


> glibc has been updated to 2.26 recently and the xlocale.h header file has been
> removed.
> See https://sourceware.org/glibc/wiki/Release/2.26#Packaging_Changes
> 
> I tested several packages with this new version but it's difficult to test every
> single packages.

> Can you report to the upstream project or if it's already fixed upstream, send
> the patch to the ml?


I have have lodged bug report with ICU project. As far as I can tell it has not been fixed or reported.


I think your analysis is correct, the advance to glibc 2.26 has flushed out this problem.
How this gets fixed may depend on the ICU project that has icu 59.1 as their latest release.
In the mean time buildroot might have to handle this by itself.




Regards,


Ross


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20170930/d98236af/attachment.html>

[Buildroot] Build failure in icu-59.1

[Romain Naour]

Hi Ross,

Le 30/09/2017 ? 04:46, grunpferd at netscape.net a ?crit?:
> There seems to be some problem building?icu-59.1?that has crept in here.
> 
> This used to work nicely.
> 
> The following errors get reported during compiling?icu-59.1
> 
> 
> make[2]: Entering directory
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/common'
> make[2]: Nothing to be done for 'all'.
> make[2]: Leaving directory
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/common'
> /usr/bin/make[1]: Making `all' in `i18n'
> make[2]: Entering directory
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/i18n'
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?fmtable_cnv.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?digitlst.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?choicfmt.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?datefmt.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?smpdtfmt.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?dtfmtsym.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?reldtfmt.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?udat.cpp
> ??arm-buildroot-linux-gnueabihf-g++ ????... ?dtptngen.cpp
> digitlst.cpp:67:24: fatal error: xlocale.h: No such file or directory
> # ??include <xlocale.h>
> ???????????????????????^
> compilation terminated.
> *** Failed compilation command follows:
> ----------------------------------------------------------
> /home/rfg/w/buildroot_p_gcc6_4.13/output/host/bin/arm-buildroot-linux-gnueabihf-g++
> -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FIL
> E_OFFSET_BITS=64 -D_REENTRANT -DU_HAVE_ELF_H=1 -DU_HAVE_ATOMIC=1
> -DU_HAVE_STRTOD_L=1 -I. -I../common -DU_ATTRIBUTE_DEPRECATED= -DU_I
> 18N_IMPLEMENTATION -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE
> -D_FILE_OFFSET_BITS=64 -O3 -pedantic -Wpointer-arith -Wwrite-strings -W
> no-long-long -std=c++11 -c -DPIC -fPIC -o digitlst.o digitlst.cpp
> --- ( rebuild with "/usr/bin/make VERBOSE=1 all" to show all parameters ) --------
> ../config/mh-linux:53: recipe for target 'digitlst.o' failed
> make[2]: *** [digitlst.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> make[2]: Leaving directory
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/i18n'
> Makefile:147: recipe for target 'all-recursive' failed
> make[1]: *** [all-recursive] Error 2
> make[1]: Leaving directory
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source'
> package/pkg-generic.mk:234: recipe for target
> '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/.stamp_built' failed
> make: *** [/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/.stamp_built]
> Error 2
> 
> 
> So it looks like it can't locate xlocale.h.
> 
> 
> I have attached the .config for this build.

Thanks for the report!

glibc has been updated to 2.26 recently and the xlocale.h header file has been
removed.
See https://sourceware.org/glibc/wiki/Release/2.26#Packaging_Changes

I tested several packages with this new version but it's difficult to test every
single packages.

Can you report to the upstream project or if it's already fixed upstream, send
the patch to the ml?

Note: For now, toolchains used in autobuilders still use glibc 2.25. That's why
we didn't noticed this issue before.

Best regards,
Romain

> 
> 
> Regards,
> 
> Ross
> 
> 
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot
> 

[Buildroot] Build failure in icu-59.1

[grunpferd at netscape.net]

There seems to be some problem building icu-59.1 that has crept in here.



This used to work nicely.


The following errors get reported during compiling icu-59.1




make[2]: Entering directory '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/common'
make[2]: Nothing to be done for 'all'.
make[2]: Leaving directory '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/common'
/usr/bin/make[1]: Making `all' in `i18n'
make[2]: Entering directory '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/i18n'
   arm-buildroot-linux-gnueabihf-g++     ...  fmtable_cnv.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  digitlst.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  choicfmt.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  datefmt.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  smpdtfmt.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  dtfmtsym.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  reldtfmt.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  udat.cpp
   arm-buildroot-linux-gnueabihf-g++     ...  dtptngen.cpp
digitlst.cpp:67:24: fatal error: xlocale.h: No such file or directory
 #   include <xlocale.h>
                        ^
compilation terminated.
*** Failed compilation command follows: ----------------------------------------------------------
/home/rfg/w/buildroot_p_gcc6_4.13/output/host/bin/arm-buildroot-linux-gnueabihf-g++ -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FIL
E_OFFSET_BITS=64 -D_REENTRANT -DU_HAVE_ELF_H=1 -DU_HAVE_ATOMIC=1 -DU_HAVE_STRTOD_L=1 -I. -I../common -DU_ATTRIBUTE_DEPRECATED= -DU_I
18N_IMPLEMENTATION -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -O3 -pedantic -Wpointer-arith -Wwrite-strings -W
no-long-long -std=c++11 -c -DPIC -fPIC -o digitlst.o digitlst.cpp
--- ( rebuild with "/usr/bin/make VERBOSE=1 all" to show all parameters ) --------
../config/mh-linux:53: recipe for target 'digitlst.o' failed
make[2]: *** [digitlst.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source/i18n'
Makefile:147: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 2
make[1]: Leaving directory '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/source'
package/pkg-generic.mk:234: recipe for target '/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/.stamp_built' failed
make: *** [/home/rfg/w/buildroot_p_gcc6_4.13/output/build/icu-59.1/.stamp_built] Error 2



So it looks like it can't locate xlocale.h.




I have attached the .config for this build.




Regards,


Ross
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20170929/eacae2ff/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: .config
Type: application/octet-stream
Size: 118400 bytes
Desc: not available
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20170929/eacae2ff/attachment.obj>