libxml-parser-perl failure
Nitin, Richard:
I was looking into the libxml failure today, and found a couple of different items that have changed in the newer version. First is the addition of an arch_check function in the Extuils/MM_Any.pm and used MakeMaker.pm. This check seems to use the archlib, which is set to ${STAGING_LIBDIR}/perl/${PV} by poky, but expects the final directory to be the ARCH, which is why we get 5.12.2 for the Arch!
Why do we set archlib to the ${PV} instead of ${PV}/${target}
The next issue that I saw had to do with Errno.pm and another architecture check, this time the errors:
ERROR from evaluation of
/vol/1/sgw/builds/ppc/tmp/work/ppc603e-poky-linux/libxml-parser-perl-2.36-r0/XML-Parser-2.36/Expat/Makefile.PL: Errno architecture (x86_64-linux-thread-multi-2.6.34.7-56.fc13.x86_64) does not match executable architecture (powerpc-linux-2.6.37-rc5-yocto-standard+) at /vol/1/sgw/builds/ppc/tmp/sysroots/x86_64-linux/usr/lib/perl/5.12.2/Errno.pm line 11.
This looks like another check that will need some work
Sau!
Hi,
Quoted 23 lines · Saul Wold
Nitin, Richard:
I was looking into the libxml failure today, and found a couple of different items that have changed in the newer version. First is the addition of an
arch_checkfunction in the Extuils/MM_Any.pm and used MakeMaker.pm. This check seems to use the archlib, which is set to ${STAGING_LIBDIR}/perl/${PV} by poky, but expects the final directory to be the ARCH, which is why we get 5.12.2 for the Arch!Why do we set archlib to the ${PV} instead of ${PV}/${target}
The next issue that I saw had to do with Errno.pm and another architecture check, this time the errors:
ERROR from evaluation of
/vol/1/sgw/builds/ppc/tmp/work/ppc603e-poky-linux/libxml-parser-perl-2.36-r0/XML-Parser-2.36/Expat/Makefile.PL:
Errno architecture (x86_64-linux-thread-multi-2.6.34.7-56.fc13.x86_64)
does not match executable architecture
(powerpc-linux-2.6.37-rc5-yocto-standard+) at /vol/1/sgw/builds/ppc/tmp/sysroots/x86_64-linux/usr/lib/perl/5.12.2/Errno.pm
line 11.
Quoted 1 line
This looks like another check that will need some work
This seemed to be introduced by my change to wrap the perl binary with a wrapper script. I've updated the wrapper to append to PERL5LIB rather than overwrite it which I think is why these failures were appearing. They change the lib search path to pass data around, ugly but we can work around it.
We still need to check the perl binary does work ok with sstate relocation after this.
Cheers,
Richard