Re: State of bitbake world, Failed tasks 2014-06-02

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2014-06-10 at 13:40 +0200, Martin Jansa wrote:
> On Tue, Jun 10, 2014 at 12:07:22PM +0200, Martin Jansa wrote:
> > On Mon, Jun 02, 2014 at 10:49:50AM +0200, Martin Jansa wrote:
> > > This build still contains huge patch setting B = S or inherit for
> > > autotools-brokensep for all recipes which were failing in earlier
> > > master builds.
> > 
> > http://www.openembedded.org/wiki/Bitbake_World_Status
> > 
> > == Failed tasks 2014-06-10 ==
> > 
> > === common (17) ===
> >     * meta-openembedded/meta-efl/recipes-efl/efl/efl_1.10.0.bb, do_compile
> 
> I'll look into this one (as it's caused by my pending commit and doesn't fail on my local machine)
> 
> >     * meta-openembedded/meta-gnome/recipes-gnome/libgsf/libgsf_1.14.26.bb, do_compile
> >     * meta-openembedded/meta-networking/recipes-support/curlpp/curlpp_0.7.3.bb, do_compile
> >     * meta-openembedded/meta-networking/recipes-support/lksctp-tools/lksctp-tools_1.0.16.bb, do_compile
> >     * meta-openembedded/meta-oe/recipes-connectivity/libnet/libnet_1.1.6.bb, do_configure
> >     * meta-openembedded/meta-oe/recipes-devtools/python/python-cython_0.20.1.bb, do_package
> >     * meta-openembedded/meta-oe/recipes-devtools/python/python-pyyaml_3.10.bb, do_compile
> >     * meta-openembedded/meta-oe/recipes-multimedia/id3lib/id3lib_3.8.3.bb, do_configure
> >     * meta-openembedded/meta-oe/recipes-navigation/omgps/omgps_svn.bb, do_compile
> >     * meta-openembedded/meta-oe/recipes-support/daemonize/daemonize_git.bb, do_configure
> >     * meta-openembedded/meta-oe/recipes-support/eject/eject_2.1.5.bb, do_compile
> >     * meta-openembedded/meta-xfce/recipes-apps/xfce4-notifyd/xfce4-notifyd_0.2.2.bb, do_compile
> >     * meta-openembedded/meta-xfce/recipes-apps/xfce4-screenshooter/xfce4-screenshooter_1.8.1.bb, do_compile
> >     * meta-qt5/recipes-qt/maliit/maliit-framework-qt5_git.bb, do_compile
> 
> Fixed in master-next and jansa/master branches (I forgot to push it
> before starting new build).
> 
> >     * meta-qt5/recipes-qt/qt5/qtwayland_git.bb, do_install
> 
> Fixed in master-next and jansa/master branches (I forgot to push it
> before starting new build).
> 
> >     * meta-smartphone/meta-fso/recipes-freesmartphone/python/python-pyrtc_git.bb, do_compile
> >     * meta-smartphone/meta-shr/recipes-shr/shr/python-phoneutils_git.bb, do_compile
> > 
> > === common-x86 (3) ===
> >     * meta-browser/recipes-browser/chromium/chromium_35.0.1916.114.bb, do_compile
> >     * meta-browser/recipes-mozilla/firefox/firefox_10.0.11esr.bb, do_compile
> >     * openembedded-core/meta/recipes-bsp/grub/grub_2.00.bb, do_package
> 
> RP asked for more details about this one (and volunteered to look into
> it - thanks!)
> 
> OE qemux86-64@ ~/oe/shr-core-branches/shr-core $ '/home/jenkins/oe/shr-core-branches/shr-core/tmp-eglibc/sysroots/x86_64-linux/usr/lib/rpm/bin/debugedit' -b '/home/jenkins/oe/shr-core-branches/shr-core/tmp-eglibc/work/core2-64-oe-linux' -d '/usr/src/debug' -i -l '/home/jenkins/oe/shr-core-branches/shr-core/tmp-eglibc/work/core2-64-oe-linux/grub/2.00-r1/debugsources.list' '/home/jenkins/oe/shr-core-branches/shr-core/tmp-eglibc/work/core2-64-oe-linux/grub/2.00-r1/package/usr/lib/grub/i386-pc/uhci.module'
> /home/jenkins/oe/shr-core-branches/shr-core/tmp-eglibc/sysroots/x86_64-linux/usr/lib/rpm/bin/debugedit: elf_update: invalid section alignment
> 
> workdir, rpm-sstate and few more logs included in 58MB 
> http://logs.nslu2-linux.org/buildlogs/oe/oe-shr-core-branches//log.world.20140610_021320.log/grub.tar.bz2

To update on this, "readelf -e uhci.module" shows:

Section Headers:
  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [12] .note.GNU-stack   PROGBITS        00000000 0010ce 000000 00      0   0  1

in a good build and:

Section Headers:
  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [24] .note.GNU-stack   PROGBITS        00000000 009180 000000 00      0   0  0

in a bad build. The problem is the "Al" (alignment) change from 1 to 0.
If its 0, debugedit complains.

It comes down to the linker. The command you can use is:

ld --build-id --eh-frame-hdr --hash-style=gnu -m elf_i386 -o uhci.module -O1 --as-needed --build-id=none -N -r -d uhci_module-uhci.o

and with bfd, it works, with gold it does not.

I haven't had a chance to dig into what gold is doing differently as yet.

I did change the libelf check to ignore the alignment if the section
size was zero and that let things build successfully. That probably
isn't the real issue though.

So the workaround could be to disable gold for grub and we need to look
further into what gold is doing here. Someone with some idea about elf
formant and linkers would probably be better than me at this point but
I'll try and have a look in due course if nobody beats me to it.

Cheers,

Richard