Using master d9ff2f897aa271e6b2d7e4dcaf8b8d19de513b50 If you build a completely new tree with SSTATE_MIRRORS set to a valid staging tree, most of the initial steps produce this error: ERROR: ld.so: object 'libpseudo.so' from LD_PRELOAD cannot be preloaded: ignored. ERROR: ld.so: object 'libpseudo.so' from LD_PRELOAD cannot be preloaded: ignored. I think this is because the LD_PRELOAD variable is always present, pointing into the sysroots tree, even before that tree has been created. From meta/conf/meta.conf: LD_PRELOAD=libpseudo.so LD_LIBRARY_PATH=${STAGING_LIBDIR_NATIVE}/pseudo/lib64:${STAGING_LIBDIR_NATIVE}/pseudo/lib This is trivial to reproduce. Just build a tree from scratch, without using SSTATE_MIRRORS. Then build another, again from scratch, with SSTATE_MIRRORS pointing to the sstate-cache tree produced by the first build. The preload error will occur as early as step 2. Meta-template to reproduce: % cd <<POKY-SRC-TREE>> % mkdir <<BUILD-TREE1>> % . poky-init-build-env <<BUILD-TREE1>> -- Adjust conf/local.conf, etc, as necessary % bitbake poky-image-minimal When this finishes % cd <<POKY-SRC-TREE>> % mkdir <<BUILD-TREE2>> % . poky-init-build-env <<BUILD-TREE2>> -- Adjust conf/local.conf, etc, as necessary -- Set SSTATE_MIRRORS to point to <<BUILD-TREE1>>/sstate-cache/ % bitbake poky-image-minimal
This is still happening, and on my build machine I'm fairly confident it is what's causing the build to fail at a later stage (e.g. building the kernel). I have found if you bitbake pseudo-native explicitly so it is finished as early as possible, then build the image it works fine. I'm guessing when building from sstate with many threads running, a lot more packages get built before pseudo-native because individually they take less time.
With the planned switch back from exec to fork in bitbake and the need to handle pseudo as a special case, this problem should thankfully go away when that happens.
This should be fixed in master with the exec() to fork() changes and the new way we call pseudo.
With commit id 269172... for poky tree, this issue reproduce when building nightly-x86/x86-64/arm/mips/ppc on autobuilder, it cause sanity test failed. Part of log as below: ######## | NOTE: Run ssh test in scenario sanity | /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/scripts/runqemu | Copying rootfs /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/build/tmp/deploy/images/core-image-sato-qemux86-20111027213425.rootfs.ext3 to /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/build/tmp/work/qemux86-poky-linux/core-image-sato-1.0-r0/qemuimagetest/tmp/core-image-sato-qemux86-test.ext3 | Running xterm -display localhost:1 -e 'OE_TMPDIR=/srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/build/tmp /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/scripts/runqemu /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/build/tmp/deploy/images/bzImage-qemux86.bin /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-x86/build/build/tmp/work/qemux86-poky-linux/core-image-sato-1.0-r0/qemuimagetest/tmp/core-image-sato-qemux86-test.ext3 && /bin/sleep 60' & | Test_Info: There is no Qemu process | Test_Info: Wait for qemu up... | ERROR: ld.so: object 'libpseudo.so' from LD_PRELOAD cannot be preloaded: ignored. | Test_Info: There is at least one Qemu process running | Test_Info: Begin to check if qemu network is up | ERROR: ld.so: object 'libpseudo.so' from LD_PRELOAD cannot be preloaded: ignored. ######## Detail information could be found at: http://autobuilder.yoctoproject.org:8010/builders/nightly-x86/builds/202/steps/shell_20/logs/stdio or http://autobuilder.yoctoproject.org:8010/builders/nightly-mips/builds/123/steps/shell_20/logs/stdio
This message occurs whenever sudo is used since it would be a security issue to honor LD_PRELOAD when raising privileges. Are you 100% sure this is the reason for the sanity test failures?
Sorry for this operation, I didn't know this issue was caused by security issue. Then I don't very sure this is the reason for the sanity test failures.
(In reply to comment #5) > This message occurs whenever sudo is used since it would be a security issue to > honor LD_PRELOAD when raising privileges. Are you 100% sure this is the reason > for the sanity test failures? Sorry, Richard. The sanity failure is caused by bug 1699 - connman could not start due to dbus file missing. Mark the bug to RESOLVED FIXED again due to wrong operation.