bitbake.conf/package: Collapse PKGDATA_DIR into a single machine specific directory

Currently we have a hierarchy of pkgdata directories and the code has to put together
a search path and look through each in turn until it finds the data it needs.

This has lead to a number of hardcoded paths and file globing which
is unpredictable and undesirable. Worse, certain tricks that should be
easy like a GL specific package architecture become problematic with the
curretn search paths.

With the modern sstate code, we can do better and construct a single pkgdata
directory for each machine in just the same way as we do for the sysroot. This
is already tried and well tested. With such a single directory, all the code that
iterated through multiple pkgdata directories and simply be removed and give
a significant simplification of the code. Even existing build directories adapt
to the change well since the package contents doesn't change, just the location
they're installed to and the stamp for them.

The only complication is the we need a different shlibs directory for each
multilib. These are only used by package.bbclass and the simple fix is to
add MLPREFIX to the shlib directory name. This means the multilib packages will
repackage and the sstate checksum will change but an existing build directory
will adapt to the changes safely.

It is close to release however I believe the benefits this patch give us
are worth consideration for inclusion and give us more options for dealing
with problems like the GL one. It also sets the ground work well for
shlibs improvements in 1.6.

(From OE-Core rev: 1b8e4abd2d9c0901d38d89d0f944fe1ffd019379)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

layer.conf: Version bump for DEPLOY_DIR layout change

Increase the version to signify the layout change of the images
in the deploy directory.

(From OE-Core rev: 4246e7dd59800a1d6c6d02c00f4e86eeac020767)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake.conf: include machine name in DEPLOY_DIR_IMAGE

This allows a clean seperation between image outputs from different
machines, and makes it possible to have convenience symlinks to make
the output ready to deploy.

This did require some surgery in runqemu; if explicit paths to the image
and kernel are not supplied then DEPLOY_DIR_IMAGE needs to be determined
from bitbake or set in the environment. However the script does try to
avoid requiring it unless it really is needed. Corresponding changes
were made in the automated testing code as well.

Based on an RFC patch by Koen Kooi <koen@dominion.thruhere.net>

(From OE-Core rev: 7e90261aec61f79680b5eaeaf5b18c7b795412a4)

Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

icu-native: do_install: Segmentation fault

There was a "Segmentation fault" error when build icu-native when the
TMPDIR is in a deep directory (for example, when len(readlink -f $TMPDIR
== 410)), use LARGE_BUFFER_MAX_SIZE for cmd rather than
SMALL_BUFFER_MAX_SIZE would fix the problem, this should be a misplay
because other cmd uses LARGE_BUFFER_MAX_SIZE.

[YOCTO #5171]

(From OE-Core rev: ebfdddc23ff78231a819c62c8ffcced9633aa08b)

Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

perf: source should be ready after do_unpack

In perf.bb:

S = "${STAGING_KERNEL_DIR}"

So the source should be ready after the do_unpack, and we need this:

do_unpack[depends] += "virtual/kernel:do_populate_sysroot"

Otherwise, maybe no source after do_unpack.

[YOCTO #5168]

(From OE-Core rev: 01d3b15518b981199120b3b9c6923678244aefdc)

Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bb-matrix-plot: Use interpolation for sparse data

If not every combination of BB_NUMBER_THREADS and PARALLEL_MAKE have
been tested by bb-matrix.sh, e.g., by using BB_RANGE="04 08 10 12 16"
and PM_RANGE="04 08 10 12 16", then the graph that gnuplot generates by
default looks very jagged due to the missing data points. By using
splines to interpolate the missing data the graph looks a lot better.

This should not change graphs where all data points are available in any
way, only improve sparse graphs.

(From OE-Core rev: 9642c1314da64c70254f6b012aa73ef37bbaa33f)

Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake.conf: Stop providing ${P} and ${PF} by default

For a long time we've provided PN-PV and PN-PV-PR by tweaking PROVIDES. This looks
nice at first glance however it turns out to be a bit problematic. Taking make as an
example where there are two versions, 3.81 and 3.82, what should "bitbake make-3.81" do?

Currently it builds make-3.81 and make-3.82 and breaks in interesting ways. Is that
a bitbake bug? Well, it certainly shouldn't try and run the build. Why is it building
3.82 though? Its due to finding a dependency on "make-dev" and then trying to figure
out what provides it? The answer is "make" and the default version of "make" is 3.82.

So arguably, finding "make-3.81" should infer PREFERRED_VERSION_make = "3.81". Doing
so resolved the above problem since now "make" resolves to "make-3.81".

So what about if we have Recipe A:
DEPENDS = "make-3.81"
and Recipe B:
DEPENDS = "make-3.82"

That is clearly an error, easy. So finally what about if we have Recipe A:
DEPENDS = "make-3.81"
and Recipe B:
DEPENDS = "make"

The first recipe infers the PREFERRED_VERSION_make = "3.81" and then forces that
version on everything else. Is that desired? Probably not in most cases, at least not
silently.

As mitigation, we could print a WARNING about this happening. The final part of the problem
is that we can ony figure this out within bitbake itself. That means we'd have to teach bitbake
about the PN-PV format of PROVIDES which is breaking the separation between bitbake and the
metadata. We can't win :(.

Nobody that I know of is using or relying on this functionality so perhaps we should
just remove it instead which is what this patch does. Opinions?

(From OE-Core rev: a87c205bb6cefd5e1a41b8e7ef02b5bfa380e3b6)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bb-matrix: Clean before, rather than after, building

This makes sure the the first build starts from a clean state. Otherwise
one could have the first build affected by any leftover state from
a previous build.

This also leaves a working state behind after the final build.

(From OE-Core rev: f8f86ac88aa1bba99ba28762cfbd97d3721da7d9)

Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

sstate.bbclass: fix parallel building issue

sstate_package creates hardlink from sysroot to SSTATE_BUILDDIR, then
sstate_create_package will store SSTATE_BUILDDIR into a archive file by
tar, but once other packages install the same file into sysroot, the
creating the archive file will fail with below error:

    DEBUG: Executing shell function sstate_create_package
    tar: x86_64-linux/usr/share/aclocal/xorg-macros.m4: file changed as we read it

This kind of error is harmless, use --ignore-failed-read to ignore it.
The error in tar occurs when the timestamp of the file changes and this
can happen when the number of symlinks change. The file will be included
in the archive.

[YOCTO #5122]

(From OE-Core rev: 4b3e353a532c7b68b0bb86df4a2fcc44f8bb3ef2)

Signed-off-by: Roy.Li <rongqing.li@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: cooker: updateCache should rause exceptions, not sys.exit

Exiting from the server is antisocial, instead we should raise an exception. This
will correctly fail the current command and reset the server state. We use
the handled exception since for these conditions to occur, something was
already displayed to the user.

(Bitbake rev: dacc94bcace85a2e95aee2dccd8e680c59e4545f)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: knotty: Cleanup error/interruption handling

Only display a CommandFailed ERROR: message if there is an error to display.

Only display an errors summary if we actually displayed errors.

(Bitbake rev: 568ea00acd226d48e725bb01d4f8c410ed1eaa61)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: command: Treat empty messages as failures, not CommandCompleted

Empty messages should trigger CommandFailed, not CommandCompleted as
otherwise the exit code will be incorrect.

(Bitbake rev: 70a8ead31f9ffc987d9c6db61a926f7a9af8f8b1)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: cooker/command: Add finishcommand to reset cooker state

After running a command on the server, it needs to reset to the initial
state. This ensures that subsequent clients start from a known state
and notice any configuration changes.

Ultimately we may want to do more than this buts a good start and better
than nothing.

(Bitbake rev: dd15648fc2654b8d7c3e00ea7ab3dbf04f24f24b)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: cooker: Rename confusing 'stop' state to 'forceshutdown'

The shutdown state causes the server to finish what its doing, stop was
them meant to completely stop it. It doesn't mean the server is stopped
though. Renaming the current stop event for forceshutdown gives more
meaning to what it actually does. The stopped namespace then becomes
available to indicate a completely stopped server.

(Bitbake rev: 12e9d33bfae5294e3870dfd1202f63383ad05e92)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: cooker: Clean up init/reset configuration code

Currently the cooker event data isn't rebuilt upon reset and the cache
configuration cannot be changed after init. These are both bad things
and this patch refactors the init/reset code so that it is possible
to reconfigure the server.

(Bitbake rev: 1193b8d76fcb6cb87e9ec135a2514370d7dd90ac)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: cooker: Drop obsolete worker test

This call only ever happens in cooker context now so we can drop the
nasty worker check from here.

(Bitbake rev: bc0b30199a8e3624c5b9914430adbcc7c6bd4497)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: bitbake: xmlrpc: delete unused code BitBakeUIEventServer

BitBakeUIEventServer is an unused class that pushes UI
events over a separate thread.

The current version of XMLRPC server works just fine with
the classic UI event handlers, so this class is not needed.

(Bitbake rev: 8e8e17631d790271b1be747c4b45059ec38ab606)

Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: data_smart: Account for changes in append/prepend/remove in the config hash

bitbake wasn't reparsing when _remove items were added to its configuration
and equally, appends/prepends were also being badly tracked. This
change enrures these variables are accounted for in the configuration
hash.

[YOCTO #5172]

(Bitbake rev: 62914f9208ef2427a34daa523af857f4027900eb)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libnewt-python: Don't write a whiptail package

The whiptail package is written by the libnewt recipe so for the python version
we need to ensure its not in PACKAGES.

(From OE-Core rev: b26157a8b575aa7b5c96ffe782fc23ec0920311c)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

populate_sdk_rpm: Only remove trailing whitespace, not all whitespace

(From OE-Core rev: 06cb4deba6e4df4ffe3015602d5997b26d28778b)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

weston: as weston-launch depends on PAM, control it with a PACKAGECONFIG

(From OE-Core rev: fb430271da67d7ec98c285a3420b00db737cae55)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

clutter: fix compilation when building without X11 backend

If Clutter is building the Wayland backend but not the X11 backend, the Cogl
support doesn't get enabled so the Wayland backend (which uses it) fails to
compile.

Backport a fix from upstream to fix this situation.

(From OE-Core rev: 6df36a004210901acceff4a201d0f910d99eccfe)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

x11vnc: add dependencies for xrandr/xfixes/xdamage/xtst

Inspired by a patch by Martin Jansa but with these libraries required, as a VNC
server without them is suboptimal.  Don't add an option for Xinerama as our X
server always disables it.

(From OE-Core rev: 4bca0b8adcac6eebc4466a2eea03de3493bf1cc4)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

populate_sdk_rpm: Ensure empty strings aren't passed to attemponly rpm code

If empty strings are passed to the rpm attemptonly code, it breaks. This
ensures we don't do that.

(From OE-Core rev: 6a0d5a3158608364f6baa11fe9ab7fa1f8e251ab)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

xf86-video-vmware: respect opengl DISTRO_FEATURE

Conditionalise the GL dependencies on the opengl DISTRO_FEATURE so this driver
can build without the opengl feature active, as the configure script will
enable/disable the GL sub-driver depending on the presence of these libraries.

This is an interim patch to fix the autobuilder, a more complete patch heading
upstream to make the GL sub-driver deterministic will follow.

(From OE-Core rev: 66b7c59442cc494179418995292c57bab56fada3)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bluez4/5: Add EXCLUDE_FROM_WORLD = 1

We can't build both recipes in the world build as there is a collision of package name
and PR values. Specificly the libasound-module-bluez which is the same in both goes
backwards from r5 (bluez4) -> r0 (bluez5) and the subpackage_metadata check fails:

ERROR: Recipe lib32-bluez5 is trying to change PR from 'r0' to 'r5'. This will cause do_package_write_* failures since the incorrect data will be used and they will be unable to find the right workdir.

[YOCTO #5165]

(From OE-Core rev: 3653500c7e5d365ddc2868d985c6001d40123672)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

xf86-video-intel: disable DRI2 tests if no DRI2 is available

Somehow this slipped past the initial testing and can cause build errors in
non-GL distros.

(From OE-Core rev: 975412f446066215a94be78a7f7e1db9e28ebbbf)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

xf86-video-vesa: change depends to refect reality

This recipe appears to have been a copy-and-paste from -intel, because it has
dependencies and configure options that xf86-video-vesa just doesn't have, such
as virtual/libgl and XVMC.

(From OE-Core rev: 4995acafc139107c2323da69eca718009384efe4)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

ia32-base: only depend on GL if opengl DISTRO_FEATURE enabled

As Mesa refuses to compile if the "opengl" DISTRO_FEATURE isn't enabled,
mesa-driver-i9xx and the GLX X module have to be conditional in the ia32 machine
defintion too.

(From OE-Core rev: 8b5c07e6c3b492f56ce9c5f99a732793403d6b36)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lib/oeqa/runtime: smart: limit channel add to useful ones

Don't add inappropiate channels on the target.
This happens when building two different machines in the
same dir and then running the tests for each machine.

(From OE-Core rev: 9e4e475b598f6b9bd2b653c72a1c1d5bd5c0b8c9)

Signed-off-by: Stefan Stanacar <stefanx.stanacar@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

classes/package_rpm: remove unsed outdir variable

Not only outdir had the wrong value, it wasn't used actually used in that function.

(From OE-Core rev: 5db4b53e5d969a6da314904fa2335462947c97ea)

Signed-off-by: Stefan Stanacar <stefanx.stanacar@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: conditionalise mesa-swrast on opengl in templates

As per the similar commit in oe-core on the qemu machines, the templates should
respect the opengl DISTRO_FEATURE when adding mesa-driver-swrast to new qemu
machines.

(From meta-yocto rev: 8d42ea2b2566913642c759e48ffe1f5f5166c510)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

ptest-runner: trivial fixes and refine

1. ptest files may be installed under /usr/lib64/ for 64bit filesystem
or under /usr/lib/ for 64bit multilib filesystem, so we should check both
directories

2. If a soft link is linking to a directory under the same directory, we
only run once.

[YOCTO #5125]
[YOCTO #5126]

(From OE-Core rev: 51c43e08b388ed15520c66977bbb49df18e5f124)

Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libcroco: disable Bsymbolic if it is not supported on some hosts

When trying to build on my Centos 5.5 machine, got below error:
| checking for gtkdoc-mkpdf... no
| checking whether to build gtk-doc documentation... no
| checking for CROCO... yes
| checking if gcc supports "-Wl,-Bsymbolic-functions" flag...
| configure: error: -Bsymbolic requested but not supported by ld. Use --disable-Bsymbolic to disable
| Configure failed. The contents of all config.log files follows to aid debugging
|
| configure: exit 1
| ERROR: oe_runconf failed

Set --enable-Bsymbolic=auto to disable it when it is not suppported.

(From OE-Core rev: e7c691b33573b3309752d6eb397486c2c8620adb)

Signed-off-by: Ting Liu <b28495@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

python-smartpm: Add an attempt install mode

[ YOCTO #3723 ]

Add a mode to smart that will allow an installation to continue, instead of
failure in the case that one or more items is uninstallable.

Uninstallable packages are simply ignored, and no error is generated.

(From OE-Core rev: bdf07b1698d228dc7ff555199a269b1ff8ceca19)

Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: utils: Fix BB_PRESERVE_ENV

BB_PRESERVE_ENV wasn't working since data.inheritFromOS wasn't getting a
correct list of keys to import into the data store. This fixes
things so it does add all environment variables into the data store
when BB_PRESERVE_ENV is used.

(Bitbake rev: 843e9339c5ee3c99657a40a0e2c7dbd777b6ef06)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

uclibc: Add missing DEPENDS on kern-tools-native

This is needed for the kconfig handling and the merge-config.sh script.

(From OE-Core rev: 1c34b0e871e475a90e27d338469ba31f322d5300)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

image/populate_sdk: Ensure symlinks in target sysroot are relative

In the target sysroot of an SDK we can have target system absolute links
which don't make sense. This adds a script which fixes them up to become relative
paths instead.

[YOCTO #5020]

(From OE-Core rev: 57d6bdcad55c119e9ab8089d23d462436a0e4440)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemu: only depend on mesa-driver-swrast if opengl is enabled

As Mesa refuses to compile if the "opengl" DISTRO_FEATURE isn't enabled,
mesa-driver-swrast has to be conditional in the QEMU machine defintions too.

(From OE-Core rev: 9951e1da6a755f9a46d3a595aa4c2f975aee8f46)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

init-live.sh: distinguish between a read-only image and a read-write image

The iso and hddimg share a common concept of 'live image', and they
use the same initramfs and thus the same init. However, that init
script in initramfs made a wrong assumption that the rootfs image
was read-only by itself. This is apparently not true for hddimg.

To make things work as expected, this init script should at least
distinguish between a read-only rootfs image and a read-write one.

This patch adds this ability to the init script. After this change,
the init script would be able to check whether the rootfs image is
read-only or not. If the rootfs image is read-write, the image will
be mounted and then booted directly. No union mounts will be attempted
in this case.

[YOCTO #5164]

(From OE-Core rev: 29f869b68a9017502f75915784a924f0fe9d4be1)

Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

midori: excluded from mips64 temporarily

Exclude midori from mips64 temporarily. Becuase midori depends on
webkit-gtk which could not build for mips64 with 64bits userspace now.

[YOCTO #5141]

(From OE-Core rev: 1b0602b5933b3ee8e4ae36d447bf881308528a47)

Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd.bbclass: get the correct /lib directory

By using os.path.dirname(systemd_unitdir) we get the correct /lib directory instead of
dealing with possibly multilib directories.  This address a QA Error for shipped/not
installed /lib with multilib and x32.

(From OE-Core rev: 5b451a46550ba62e2fbfe5dbe50723b34a4fd527)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd: Remove base_libdir if installed only for systemd_unitdir

If a recipe installs systemd_unitdir and it is a non-systemd build than
systemd.bbclass deletes systemd_unitdir (/lib/systemd/) but not
base_libdir (/lib). In this case if base_libdir is empty than following
QA Issue is reported.

ERROR: QA Issue: openssh: Files/directories were installed but not shipped
  /lib

If base_libdir was installed due to systemd_unitdir installation than for
non-systemd build it should also be removed.

(From OE-Core rev: c700cb21c189e23a78f0efaaf763259c8cfefa4d)

Signed-off-by: Muhammad Shakeel <muhammad_shakeel@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

nfs-utils: Add systemd support

- Remove dependency on meta-systemd

(From OE-Core rev: 644394c074c359505fcd750356fb4db58b939729)

Signed-off-by: Muhammad Shakeel <muhammad_shakeel@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

rpcbind: Add systemd support

- Remove dependency on meta-systemd

(From OE-Core rev: 3ecb5e21b3b4f05f2d527264c3574fdc37984a77)

Signed-off-by: Muhammad Shakeel <muhammad_shakeel@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lighttpd: Add systemd support

- Remove dependency on meta-systemd

(From OE-Core rev: 36e3d63abaeafc5d2e671bad45b599c159ed10ab)

Signed-off-by: Muhammad Shakeel <muhammad_shakeel@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

openssh: Add systemd support

-Remove dependency on meta-systemd

(From OE-Core rev: 6e6445d487c033913a29763f8e3a7a339d5b612d)

Signed-off-by: Muhammad Shakeel <muhammad_shakeel@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

avahi: use correct names in SYSTEMD_SERVICE_* variables

Don't hard-code full package names in SYSTEMD_SERVICE_*, because in multilib
they'll be changed.

[ YOCTO #4803 ]

(From OE-Core rev: 013cc8b6397c29e8f0d7adf63d8e06caab778da2)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

dbus-ptest: fix populate_sdk error

 When we enable "ptest", populate_sdk fails with the following error:
 # bitbake core-image-minimal -c populate_sdk
 ..
 Collected errors:
  * satisfy_dependencies_for: Cannot satisfy the following dependencies
    for dbus-ptest-dev:
  *    dbus-ptest (= 1.6.10-r0) *
  * opkg_install_cmd: Cannot install package dbus-ptest-dev.

Disable that dependency.

(From OE-Core rev: a59b7341d136bf8ee4a22c2968ca95dc17be3947)

Signed-off-by: Maxin B. John <maxin.john@enea.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

initscripts: filter out empty lines from volatiles in check_requirements()

So far the rules of check_requirements() is too strict to lead mismatch
when empty lines exist in volatiles.

(From OE-Core rev: 71ab9ee58b0ba5e3f5cbf403d1b8fb79fc7f5ed1)

Signed-off-by: Ming Liu <ming.liu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

btrfs-tools: fix a parallel build issue

(From OE-Core rev: 637e0484ea4e7a39ac4d9c407c585201e6be1bc0)

Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

mount.sh: automount cdrom device even if no ID_FS_TYPE available

This script was modified to check whether $ID_FS_TYPE is empty before
automount, however, for cdrom devices on qemu, the ID_FS_TYPE is not
set, yet the device should be mounted. Otherwise, when booting an iso
image with runqemu, the boot process hangs at 'waiting for removable
media'.

This patch fixes this problem by first checking whether the block device
is a cdrom.

[YOCTO #4487]

(From OE-Core rev: 2df21a3b27543df912a3ce05bef2415fae13cb7c)

Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gstreamer-plugins: Change anonymous python structure

This change ensures that the SRC_URI is set before base.bbclass code is called. This will
also ensure that the xz-native depends is set correct.

[YOCTO #5127]

(From OE-Core rev: 787335494dc0a3f93d9bf5a6cf1b603c8675df66)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

cmake.bbclass: ensure CMAKE_SYSTEM_NAME is correct

Using TARGET_OS can add the ABIEXTENSION so ensure that is is removed for the Linux
TARGET_OS, we might have other TARGET_OSes so don't hard code CMAKE_SYSTEM_NAME

[YOCTO #5145]

(From OE-Core rev: 7d8b700242b1b32c6b6d0735b497701800f54fc4)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

adt-installer: allow installation of other machines

Currently, adt-installer allows only the installation of qemu target
sysroots.

The changes in this patch do the following:
 * add a new setting in adt-installer.conf (YOCTOADT_TARGET_MACHINE) for
   each target architecture. For example, for arm we can choose to use a
   qemuarm sysroot or a beagleboard sysroot. By default, only the qemu
   target sysroots are selected (current behavior);
 * change adt_installer scripts to allow installing the correct
   meta-environment package for the selected machine;
 * remove some left-over commented lines;
 * use packagegroup-cross-canadian-${MACHINE} when installing the cross
   canadian packages instead of doing it separately for each package;
 * change the opkg config files in order to be able to find the
   packagegroup package, which is allarch;

[YOCTO #4783]

(From OE-Core rev: 02085d410bf734e833d45293f4d5b06bb9536a60)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

meta-toolchain-qt: include the proper cross-canadian package

packagegroup-cross-canadian package name was changed to contain MACHINE.
Make the necessary changes here too.

[YOCTO #4783]

(From OE-Core rev: 8f8cd338faf63f665c2214f0110b712736f8ed5d)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

populate_sdk_base.bbclass: select the proper cross-canadian package

Since packagegroup-cross-canadian package name has MACHINE in it, make
the necessary changes here.

[YOCTO #4783]

(From OE-Core rev: 23531590d40acde2775b7c3b90682edc501d68f4)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

packagegroup-cross-canadian: change package name to include MACHINE

Since meta-environment package name was changed to contain MACHINE,
change packagegroup-cross-canadian package name too, in order to be able
to select the proper environment files for a certain machine.

Also, remove the RPROVIDES since adt-installer doesn't really need it.

[YOCTO #4783]

(From OE-Core rev: d713447dc2fab8900f47c542cfbbb76bad19a7e2)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

meta-environment: change package name to include MACHINE

Currently, the package name contains just the TRANSLATED_TARGET_ARCH.
When compiling the toolchain for two machines, belonging to the same
architecture (for example: qemuarm and beagleboard), this package gets
overwritten and adt-installer repo will contain just one
meta-environment package... This leads to situations like installing the
toolchain for qemuarm and end up with meta-environment package with
beagleboard cpu options.

[YOCTO #4783]

(From OE-Core rev: 09a2b158818e2ec3c3e3b53b6d14fd3527c32c99)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-yocto/3.10: bump kver to 3.10.11

Bumping to the 3.10.11 -stable release. Of note in this update is
the timer_list fix, which broke ssh access in the previous 3.10
series of updates:

    timer_list: correct the iterator for timer_list

    commit 84a78a6504f5c5394a8e558702e5b54131f01d14 upstream.

With this, we can revert the two temporary fixes for timer lists.
ssh access has been directly tested after this update.

(From OE-Core rev: 659528f0f99cec94991ec96fa124c4497237e9f8)

Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

u-boot, u-boot-fw-utils-cross, u-boot-mkimage: Upgrade to 2013.07

This upgrades the U-Boot based recipes for 2013.07 release.

This removes the 2011.03, 2011.06 and 2013.01.01 versions so we keep a
single one in core. The following recipes has been upgraded:

 - u-boot (remove old versions)
 - u-boot-mkimage (remove old versions)
 - u-boot-fw-utils -> u-boot-fw-utils-cross (renamed and reworked)

The u-boot-fw-utils-cross recipe has been reworked as it uses the
UBOOT_MACHINE to find default environment for use so it is indeed a
cross binary and not a native one.

(From OE-Core rev: c5fff5748e0aaf7e135fdd464c2104b1d3cbfd5a)

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd: Fix build with multilib

rpm macros should be independent of libdir
Fixes errors seen when multilib is turned on
it shows up since then libdir != usr/lib

(From OE-Core rev: 6126c6e4b9c40c222519914c8fe0a51f9eee8d41)

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: bitbake: runqueue: add sceneQueueTaskCompleted event

Adding an event to be fired when a scene task is completed.
It is analogous to the run task completed event, and has
been missing for some reason.

(Bitbake rev: 73b8f4d3fbeaf1b330a66d76012d0a5cef8dbe2d)

Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd: update patch header and comments

(From OE-Core rev: 856f74db0f56e4885af625665f8d85841f198e8d)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

xf86-video-intel: allow builds without opengl, add more options

Allow builds without the opengl DISTRO_FEATURE by disabling DRI.

Also add new options for UXA (not default upstream, requires DRI) and udev
(enabled by default)

Finally remove a few obsolete dependencies.

(From OE-Core rev: bb9c01bd627cae9c037811c943ad38eebb2f8b25)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

mesa: use distro_features_check to abort build without opengl feature

Stop mesa from building on distributions without the "opengl" DISTRO_FEATURE.

(From OE-Core rev: 05bacb69a3b0f3bbd9dd608855456c32b4007fde)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

kmod: avoid parallel-tests

Avoid parallel-tests for kmod as it remove
buildtest-TESTS and runtest-TESTS targets required by ptest.

(From OE-Core rev: e1a3ff7d455889e4c6b899227e8ca1919a443b22)

Signed-off-by: Tudor Florea <tudor.florea@enea.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lttng-modules: Update to version 2.3.0

(From OE-Core rev: df6943c29c0d31a78b0b2954dfe8ee6b97cefffe)

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lttng-tools: Update to version 2.3.0

(From OE-Core rev: 09dd0be9bb6d8a300700de71658abcdeb21713db)

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lttng-ust: Update to version 2.3.0

(From OE-Core rev: a7e60e7f3b8443106a767cee40954fa893905e07)

Signed-off-by: Paul Woegerer <paul_woegerer@mentor.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

ptest.bbclass: add cleandirs flag to do_install_ptest_base

The do_install_ptest_base task should have cleandirs flag, just like
the do_install task. The ${D}${PTEST_PATH} directory should be cleaned,
Otherwise, there would be similar errors like below if the
do_install_ptest_base task is rerun.

    ln: failed to create symbolic link `xxx': File exists

[YOCTO #5129]

(From OE-Core rev: 25885d3b4f679c33a514d858bc20b0e21aa63721)

Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

ptest.bbclass: fix error on ubuntu host

The do_install_ptest_base function uses 'type -t' command to check
whether do_install_ptest is a function and acts correspondingly.

However, the 'type' command is a shell builtin and its behavior is
not all the same across Linux distros. On ubuntu, if we use #!/bin/sh
as the interpreter for the scripts, as in the case of our intermediate
scripts, the '-t' option for the 'type' command is not supported. So
the check always fails and the do_install_ptest function, even if defined,
is not run.

The same problem also applies to the do_configure_ptest_base and the
do_compile_ptest_base functions.

This patch fixes this problem by avoiding using the 'type' builtin command.

[YOCTO #5128]

(From OE-Core rev: d5a4f031b460437e9501e4e65194ce94d3641130)

Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: remove apm as a default MACHINE_FEATURE

APM is not only obsolete, but also requires a kernel config enabled, which is not enabled by default

[YOCTO #5121]

(From meta-yocto rev: e25c43661f27b27e61aa7fae868237c1c60e3e25)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: runqueue.py: check whether multiple versions of the same PN are due to be built

There would be an race issue if we:

$ bitbake make-3.81 make-3.82

This because they are being built at the same time which would cause
unexpected problems, for example:

[snip]
ERROR: Package already staged (/path/to/tmp/sstate-control/manifest-qemux86-make.populate-sysroot)?!
ERROR: Function failed: sstate_task_postfunc
[snip]

Or there would be python's strack trace such as:

[snip]
 *** 0004:    mfile = open(manifest)
     0005:    entries = mfile.readlines()
     0006:    mfile.close()
     0007:
     0008:    for entry in entries:
Exception: IOError: [Errno 2] No such file or directory: xxx
[snip]

[YOCTO #5094]

We can quit earlier to avoid this kind of issue when two versions of the same PN
are going to be built since this isn't supported.

(Bitbake rev: ab377c00c33a2d296bfda1b0b6c2a62b29d1004f)

Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: providers.py: enhance the runtime debug degbug messgae

The runtime provider debug message is the same as the build time debug
message, make them different would be better.

[YOCTO #5067]

(Bitbake rev: 92b624cbc2711d3d859994099fb63918dfd0031a)

Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemu.inc: remove apm as a MACHINE_FEATURE

APM is not only obsolete, but requires a kernel config enabled and is meaningless for QEMU VM

[YOCTO #5121]

(From OE-Core rev: b0f8c47b1e808421f03308527beb8bde15644acd)

Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: prserv: Ensure data is committed

In exclusive mode, we need to complete the transaction for writes to make
it to the database. Therefore add sync calls to ensure this happens.

Autocommit mode is significantly (100 times) slower so caching the
data is of significant benefit.

(Bitbake rev: 4e55f7821786a59c2cd7dbd8bfa2a22f5f196e99)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

cmake: set system name correctly

For unknown reasons, the cmake class is using SDK_OS as the
target system OS. This makes no sense but only shows up as a problem
when you try a different SDK OS. Fix it to use TARGET_OS which is
the correct thing to do. For the vast majority of users this will
make no difference.

(From OE-Core rev: 57be84259f0885865c85d7bac350979430b956b5)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

package.bbclass: skip already-stripped QA test if asked for

Some packages like grub have already stripped binaries e.g.

ERROR: QA Issue: File '/boot/grub/kernel.img' from grub was already
stripped, this will prevent future debugging!
ERROR: QA run found fatal errors. Please consider fixing them.

We would like to have a possibility to skip it using something like

INSANE_SKIP_${PN} = "already-stripped"

This adds the logic to do so

it acts at PN level and not at package level. so something like

INSANE_SKIP_${PN}-misc = "already-stripped" wont work.

(From OE-Core rev: 765982f4c050d9cd3eb608d630312da482c737c7)

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

ghostscript: Add missing dependency to resolve parallel make issue

(From OE-Core rev: a24027321b99dffd79b1f0c009ce89f0be7cc384)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

uboot-config.bbclass: Allow choose of U-Boot config for machine

Some machines provide several possible configurations and until now
there was no easy way for user to override the default setting.

This class provides a system similar to PACKAGECONFIG but for
U-Boot. The format is:

UBOOT_CONFIG ??= <default>
UBOOT_CONFIG[foo] = "config,images"

There are two possible parameters:

 - config: it is used to set UBOOT_MACHINE
 - images: it is used to append onto IMAGE_FSTYPES

Below there's an usage example:

,----[ i.MX6Q SABRE AUTO based example ]
| UBOOT_CONFIG ??= "sd"
| UBOOT_CONFIG[sd] = "mx6qsabreauto_config,sdcard"
| UBOOT_CONFIG[eimnor] = "mx6qsabreauto_eimnor_config"
| UBOOT_CONFIG[nand] = "mx6qsabreauto_nand_config,ubifs"
| UBOOT_CONFIG[spinor] = "mx6qsabreauto_spinor_config"
`----

User can, from local.conf or environment, use UBOOT_CONFIG=nand and
override the default setting, as:

,----[ Override example from command line ]
| MACHINE=imx6qsabreauto UBOOT_CONFIG=nand bitbake core-image-base
`----

(From OE-Core rev: 5dba521611d644357cf0a98d2e30dcf41777c6ef)

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-firmware: Update SRCREV, pull in iwlwifi-7260 support

Fixes [YOCTO #5110]

Add support for the iwlwifi 7260 adapters. This creates a new package
and includes support in the default linux-firmware (everything) package.

Update the iwlwifi and radeon license checksums. Extensions to the
copyright date ranges were the only change to the LICENSE files.

(From OE-Core rev: bf4044ab8f9ef3be087996a559c54eeded1fb0c8)

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: yunguo.wei@windriver.com
Cc: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

kern-tools: support no author patches

Updating the kern-toosl SRCREV to pick up the following fix:

    previous versions of the kern-tools supported the ability to import a bare
    patch, with no From: Subject: or other identifying fields that are typically
    in a full commit.

    The same type of commit with kgit-s2q will prompt for a author ID, just
    as git-quilt-import does. In build system environment that leads to an
    infinite loop and the commit is never pushed.

    To fix this issue, we add an interactive flag (-i), that when passed the
    prompt based behaviour is used. When it isn't passed (the default), the following
    name and email will be used for the git author:

       GIT_AUTHOR_NAME="invalid_git config"
       GIT_AUTHOR_EMAIL="<unknown@unknown>"

    And a bare/incomplete header patch will be applied.

[YOCTO #5100]

(From OE-Core rev: cb0d8f8b9c59b351d11eef9c4951c4ce5601acb8)

Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

kernel-yocto: improve error checking on non-git repos

While non-git kernel repos are not the preferred format for a kernel upstream,
they are supported. Depending on the creator of the archive the expanded
source directory name varies. If the recipe for the kernel doesn't properly
set S to the right value, a cryptic git error message is produced. We can
detect the situation and offer some advice on how to fix the issue.

A second check is also added in this commit for archive based kernel repos
which won't have a SRCREV to validate. If we have no SRCREV or SRCREV is
INVALID, we can exit the branch validation step immediately. This saves yet
another cryptic git error message and simplifies a custom tgz based recipe.

(From OE-Core rev: 0ebf67e8b4f7aaf259d7abac4af645070d846ec8)

Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-yocto/3.10: update meta SRCREV for beagleboard config changes

Updating the beagleboard configuration to match the 3.10 kernel changes.
With this, the beagle* boards boot out of the box, with no additional
changes required.

Simarpreet Singh <simar@linux.com>
(From OE-Core rev: 410a4db2faec084a0e918dfce9fcb6c54d2aeaaf)

Signed-off-by: Liming Wang <liming.wang@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bluez5: Rename tests package to testtools

Bluez5 ships some tools that can be used to test its
functionality.

The installation can be tested using "make check" and
this should be included in a ptest package.

[YB #5028]

(From OE-Core rev: 8c3cbaf3fa5eafa55f209100211bd5c124b8cfaa)

Signed-off-by: Mihai Prica <mihai.prica@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bluez5: Install the bluetooth library

Bluez5 doesn't install libbluetooth by default. This is required
by connman, ofono or other packages.

(From OE-Core rev: bf7415366646db7661795620fa1ab2e78b12d947)

Signed-off-by: Mihai Prica <mihai.prica@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd: split out systemd-binfmt as a separate package

If binfmt_misc was compiled as a module but isn't installed then systemd-binfmt
will put an automounter on /proc/sys/fs/binfmt_misc and when it attempts to
automount (e.g. when df is called) it can't find support for the filesystem, and
throws an error.

As binfmt_misc isn't commonly used, split this helper into it's own package, add
a dependency on kernel-module-binfmt-misc, and ensure the service gets started
when it's installed.

[ YOCTO #4863 ]

(From OE-Core rev: d42f7fc333495dc35227a6d1027492ab70f29b23)

Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lib/oeqa/qemurunner: Use a timeout in select() call

A 0 value to select.select() immediately returns with no timeout. This was
pegging the cpu at 100% for the python process which was bad and may be
contributing to some of the timeout problems.

Profile from -P of a core-image-minimal before:

    97526792 function calls (97525652 primitive calls) in 45.189 seconds

and after:

       50204 function calls (49064 primitive calls) in 17.318 seconds

Saving 97.5 million function calls has to be good :)

(From OE-Core rev: c0551436974d179df23418567f18a082830380f6)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

core-image-weston: Ensure wayland distro feature is enabled

(From OE-Core rev: 753be24f0bd63520b6840d1a68500f5bb38f086d)

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

core-image-directfb: Use 'distro_features_check' class

(From OE-Core rev: 151c69a2f9f66bdd943f2db4d0b733c3d5091f4e)

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

resolvconf: specify configuration for systemd-tmpfiles --update

Running systemd-tmpfiles --update without specifying a configuration
file results in all tmpfiles.d configuration files being processed.

/usr/lib/tmpfiles.d/systemd.conf creates /run/nologin on boot to
prevent non-root users from logging in while the system is booting.
If systemd-tmpfiles --update is run after the system has started,
it will still create /run/nologin which would prevent non-root users
from logging in with the message "System is booting up.".

(From OE-Core rev: 24f9280c35001ff6c1d5a263fab41ae21a8056f3)

Signed-off-by: Jonathan Liu <net147@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lib/oeqa/runtime: add basic scanelf test

This uses scanelf from the pax-utils package and scans the binaries in PATH
for TEXTREL and RPATH information. For a sato image with pax-utils installed
it shows no output (which is good).

(From OE-Core rev: 629099ad66f5fa2814e5f7908b426149e8978e43)

Signed-off-by: Stefan Stanacar <stefanx.stanacar@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-firmware: package Marvell SD8787 firmware

Also split out Marvell license as separate package.

(From OE-Core rev: 64fbea5625488adc0dcccf2cf3c09880b9554a52)

Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-firmware: correct name of Marvell license file

Follow upstream in changing the name of the license file for Marvell
firmware:

  commit 2e79e60b7e4771427327ed508fa27b90d841afcb
  Author: Bing Zhao <bzhao@marvell.com>
  Date:   Fri Jan 7 16:06:56 2011 -0800

    linux-firmware: use single license file for Marvell firmwares

    Libertas and mwl8k firmware images are under the same license.

Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Brian Cavagnolo <brian@cozybit.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
  :000000 100644 0000000... 3224e1b... A  LICENCE.Marvell
  :100644 000000 1fd8766... 0000000... D  LICENCE.libertas
  :100644 000000 3224e1b... 0000000... D  LICENCE.mwl8k
  :100644 100644 d0740ce... 35b82c8... M  WHENCE

(From OE-Core rev: 0947e1fd559ce3dbce3705d1f3267860f04e3348)

Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

devshell: Don't corrupt the fakeroot variables

The devshell anonymous python fragment overwrites variables in the
datastore with their expanded versions. If this runs before the code
in allarch.bbclass which changes TARGET_OS, we can end up with different
directories in the fakeroot environment variables, some expanded with
the original TARGET_OS value.

The devshell code only needs to run before the task itself so we change
to trigger it to run at task execution time only using a flag.

[YOCTO #4795]

(From OE-Core rev: 56baf177cdf074929a090cc66a8b89d346a5d79c)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

run-postinsts: move script execution from S98 to S99 in rcS.d

Apparently, when opkg/dpkg and run-postinsts are installed in the same
time, opkg/dpkg postinstall overwrites the run-postinsts link in rcS.d.
This will make run-postinsts script useless and the delayed postinstalls
will not be run.

This issue happens only when 'package-management' is disabled and, in
the same time, dpkg/opkg ends up in the image: either pulled by some
dependency or manually installed.

With this patch, both opkg/dpkg and run-postinsts scripts will run but
the former will silently fail because the package metadata is removed from
the image since 'pacakge-management' is disabled.

[YOCTO #4484]

(From OE-Core rev: 882da38f226acc40c041155218549edad461b7d7)

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

connman: Ignore the NFS root network interface in init script

The connman init.d script tried to ignore all the network interfaces
if NFS root is configured. We should only ignore the interface
that is used by NFS root.

[YOCTO #4587]

(From OE-Core rev: 1838671b832015ae28c8c101e8b20afbbf4b3c98)

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

package_rpm.bbclass: Fix no_recommendations and package_exclude

When the code was refactored to address review comments, the wrong version
was sent to the community.  Replace the $1 with ${target_rootfs}

Fix identified by: Yue Tao <yue.tao@windriver.com>

(From OE-Core rev: a04f4fe8db425f0ea87a67b5c72d61816b8d53e2)

Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

useradd: Handle users from a package being used in others

If there is a package A (TUNE_PKGARCH) which is depended upon by B which
is MACHINE_ARCH and you build B for machine X, then Y, the user isn't
present in the sysroot for machine Y since the useradd code is never
triggered.

The change ensures the code does get triggered and the user is present.

[YOCTO 4739]

(From OE-Core rev: 5871337da49f8cd1eaf53f7cd0aacc026dc7bcdb)

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>