Bug 4214

Could be a documentation related issue.

Can you also provide your configuration (local.conf). Can you confirm is you have the populate_volatiles script installed?  This would be done via initscripts.

Sorry for lack of info.

I do not use initrd.
In my case I write "ext3" filesystem image partition on disk partition, then make chroot and install grub on that partition.
After that I boot from that partition and see systemd error.

Problem disappears when I change "base-files" recipe to create /run directory.

Here is what I use in my local.conf:
===============================
==============================
MACHINE ?= "cedartrail"

LICENSE_FLAGS_WHITELIST = "commercial license_cdv-pvr-driver_1.0.3"
GLIBC_GENERATE_LOCALES = "en_US.UTF-8"

DISTRO_FEATURES_INITMAN = "systemd"
DISTRO_FEATURES = "alsa argp largefile usbgadget usbhost wifi xattr zeroconf 3g x11 ${DISTRO_FEATURES_LIBC} ${DISTRO_FEATURES_INITMAN} opengl"

IMAGE_FSTYPES = "tar.bz2 ext3.xz"
PACKAGE_CLASSES = "package_ipk"
BBMASK = "meta-systemd/meta-efl|meta-systemd/meta-gnome"

==============================
==============================

And the following is content of my "custom-image.bb" class.
========================================================
========================================================
inherit core-image

PR = "r15"

IMAGE_LINGUAS = " "

LICENSE = "MIT"

IMAGE_ROOTFS_SIZE = "8192"

IMAGE_INSTALL = " \
    base-files \
    base-passwd \
    busybox \
    tinylogin \
    kernel-modules \
    "

IMAGE_FEATURES = "ssh-server-openssh splash package-management"

IMAGE_INSTALL += "grub xinit xf86-input-keyboard xserver-xorg cdv-pvr-driver"
========================================================
========================================================

You're using meta-systemd from meta-oe? I'm guessing this from the BBMASK. Have you tried not using meta-oe? I suspect they might be conflicting with things in OE-Core...

I don't suppose you could replicate with current oe-core/poky master?  The way of using systemd has changed: add "systemd" to DISTRO_FEATURES and then assuming you're using packagegroup-core-boot set VIRTUAL-RUNTIME_init_manager="systemd" in local.conf.

No information so we can only assume the guesses about the cause here are correct. The builds we've run confirm that systemd is working in our tests too. I'm therefore reprioritising this since it is not a release blocker, it is something we need to release note however. Once release noted, I'd consider this bug resolved. The release note needs to be something like:

"""
If people are using meta-openembedded, it should be noted that the meta-systemd component of that layer has been superceeded by the systemd support merged into OE-Core so that layer component should not be enabled. Some compatibility issues have been observed if it is.
"""

Reassigning to Scott for the release note.

Obviously if we're incorrect about the guesses about meta-oe, please comment as such in the bug and Ross will take a further look.

OK, I'm using following layers:
===================
meta-openembedded/meta-oe 
meta-openembedded/meta-networking 
meta-openembedded/meta-systemd 
meta-openembedded/meta-multimedia 
meta-intel 
meta-intel/meta-cedartrail 
openembedded-core/meta 
===================

I don't use packagegroup-core-boot as I won't need busybox and support of SystemV init scripts.
I use systemd and it's enough for my needs.
Also I use connman that do not need ifupdown scripts.

Adding Beth because she usually creates the rough release note document that I then clean up.  

Beth - can you be sure to add this note into the raw release notes text file?

Scott

Duplicated stuff in meta-systemd was removed, now there are only .bbappends for components which don't have systemd support in oe-core and .bbappends to fix upgrade path

http://git.openembedded.org/meta-openembedded/log/meta-systemd

According to Martin fixed, and documented for 1.4 release notes