systemd: use update-alternatives.bbclass

switch from using plain update-alternatives command to
update-alternatives.bbclass style

(From OE-Core rev: 6e86da976d296b926b462e976d1f79f524f061b3)

Signed-off-by: Radu Moisan <radu.moisan@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: data.py: Add a warning when expandKeys overwrites an existing key

When two variables are defined as:

${var} = "bar"
foo = "foobar"

The value of 'foo' when ${var} == foo becomes indeterminate.  We
want to warn a user when this situation has been encountered so they
can take corrective actions.

In the above example usually foo == bar, unless multilibs are enabled.
Then ml-foo = "ml-foobar".

(Bitbake rev: 7c568132c54a21161de28907159f902462f1e2bb)

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

README.hardware: bring up-to-date

* Fix Yocto Project documentation URL
* Indicate physical reference hardware support comes from meta-yocto-bsp
* Remove/replace references to Poky where appropriate

(From meta-yocto rev: e2d620445993d56008e78a7e8463080315828e4c)

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

yocto-layer / yocto-bsp: tweak layer.conf comment

We have recipes-* directories not a recipes directory; this is left over
from the old old layout (2010).

(From meta-yocto rev: 8adbbb4b688e60113f68d3974310774686551eff)

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

qemu script: explicitly set 32 bit depth for x86-64

This patch is the same as 6c22c591374d258228f74814cded34a24b4bf2d3,
but for x86-64 targets which exhibit the same problem.

Qemu update from 1.2 to 1.4 now allows for 16bit depth in guests,
whereby previously only 32bit depth was supported. However,
the new support is broken, so we force 32bit depth in all cases.

(From OE-Core rev: 6719400533453d0df482ef6e7bb347491e8a3e2b)

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

busybox: fail on no media

The current behaviour of busybox is to try all fstype when automounting
even when no media exists.  The util-linux mount command bails when no
media exists, so change the behaviour of busybox to do the same.

It could also be argued that the KERN_INFO message from btrfs could be
removed, but that would be harder to accomplish.

(From OE-Core rev: e5403f55a1e9b1747535450fd95f499c85211771)

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

postinst-intercepts, qemu.bbclass: fix segfaults in postinstalls

Postinstalls that use qemu are throwing a segmentation fault when
building for qemux86-64 on a 64bit host (it might also happen for
qemux86 if building on a 32bit host but I didn't test). It looks like
qemu looks for ld.so.cache which is not found because it is generated
after rootfs_(rpm|ipk|deb)_do_rootfs is called and then it tries to load
libraries from the default paths (which are the host's). In order to
avoid this, pass the LD_LIBRARY_PATH explicitly to the target's dynamic
loader.

(From OE-Core rev: 48e8b613b3f5c7b1d917bf3147606d44072ce49e)

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

update-rc.d: correctly look up the initscript params with overrides

The creation of a clone of d with extra OVERRIDES was removed in
72c1fd72d3b479c728e249eaa763116d352e945b but some of the lookups are essential
so that variables such as ${INITSCRIPT_PARAMS} get overriden and resolved
correctly on a per-package basis.

[ YOCTO #3960 ]

(From OE-Core rev: b016bc9aaabc90fe4dc98af8c5e73dfcb4526ef4)

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

initrdscripts: Add udev sbin based libexec path

For better or worse we need to use base_sbindir for udev's libexec dir. This
updates the initrdscripts to also cover the new location. I'd prevously assumed
that it was already covered but its not. udev internal binaries shouldn't be in
PATH so we have to do this to deal with the issue.

(From OE-Core rev: 7e17cba75c20ad820d30128d9b4b0132e7b924a8)

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

bitbake: hob:Only display scrolled list of images if needed

- do not set the Images window to always display a scroll bar if it is not needed

[Yocto #4171]
(Bitbake rev: 970e2e6f079fa9a49646f86364eae9a4ee241f90)

Signed-off-by: Ioana Grigoropol <ioanax.grigoropol@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: bitbake:hob: use a socks proxy mechanism for git

Instead of a custom git proxy mechanism, Hob now
uses a SOCKS proxy in order to work with external
repos via the oe-git-proxy helper script.

Fixes [YOCTO #4187]

(Bitbake rev: 0b81a2c4a5611b64dbdd40131730a82c149b94a2)

Signed-off-by: Cristian Iorga <cristian.iorga@intel.com>
Signed-off-by: Cristian Iorga <ubik3000@gmail.com>
Signed-off-by: Cristian Iorga <cristian.iorga@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

classes/license: remove outdated comment

Package listing was implemented in the deb backend some time ago.

(From OE-Core rev: e2915b6e1d2088d3a791bf629dabc58f38940961)

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

meta-*/conf/layer.conf: tweak BBFILES comment

"packages" was the old name (pre-2010) under which the recipe files were
stored.

(From OE-Core rev: c71fa87bc2e7155e69ea5ff7a284a05073602eed)

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

meta/recipes.txt: add recipes-lsb4

(From OE-Core rev: 79c2845b12a53d9cdcf4f7beacd09db7ee1ae2bc)

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

udev: fix init script for the location of udevd

Ensure we can update the script base don the location of the udevd installation

(From OE-Core rev: 25ff5960e41b9d7c62b05a08dd77cf11390962a1)

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

systemd.bbclass: restart service in postinst, not start

When upgrading packages it's possible that the service is already running
because opkg doesn't actually execute the prerm hooks on upgrades, which is
where the service should be stopped.

Handle this case by restarting in postinst instead of starting.  If the service
isn't already running then this doesn't make a difference, but if it is running
then the service will be restarted.

[ YOCTO #4213 ]

(From OE-Core rev: 319ef0df4ae7ed0372eff90e11244123eccb023c)

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

udev: Move udevd back to /sbin

Along with v182 upgrade udevd was moved to ${base_libdir}
making scripts like init-live.sh to fail in finding udevd.

We have some problems here since the placing binaries into either
libdir breaks the way our multilib handling works. That code and its
associated sanity tests assume that libdir contains binaries of a
particular architecture and that these are not allowed to overlap.

This is in contrast to the bindirs where conflicts are expected
and handled appropriately.

So whilst upstream may desire this directory layout, it won't work
for OE's usage of it and we need to configure udev differently. The
scripts already have fallback code to handle udev in the two locations
so there is no issue is going back to our previous layout.

[Yocto #4046]

(From OE-Core rev: a866e1e298dab5c52e7b8ba9ab68104604511713)

Signed-off-by: Radu Moisan <radu.moisan@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemuimagetest/scenario: Move dmesg to end of test run

The dmesg test detects segfaults. This is useful information to have and if one
occurs in one of the earlier tests, this can aid debugging. Move the dmesg test to
the end of the list of tests so we gain the extra debug info in those cases.

(From OE-Core rev: 472dc52974f12c255d9e98e63e82736c7ca2c223)

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

base.bbclass: Fix matching of MACHINEOVERRIDES in COMPATIBLE_MACHINE

The use of SOC_FAMILY here is old code and SOC_FAMILY is now implemented by
MACHINEOVERRIDES behind the scenes. It therefore makes more sense to use
the replacement value in this code. Just like SOC_FAMILY, this is a ":"
delimited variable so we should iterate over the components, not use
the value directly.

Finally, MACHINEOVERRIDES contains MACHINE so we don't need to check that
directly.

This makes the functionality match what most users would expect it to do
and is also compatible with the way things previously worked.

(From OE-Core rev: 8ceef74dd4f662b4c7e3c170ce486e966ebebeff)

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

oe-buildenv-internal: Only add to $PATH if needed

First strip $PATH of any existence of the paths needed by Open Embedded
and BitBake. Then add the needed paths at the beginning. This makes sure
the needed paths are searched first, without growing $PATH unnecessarily
if oe-init-build-env is rerun for a directory for which it has
previously been run.

(From OE-Core rev: 7429db6f38e405774ba66b3fa1bc3ac4b74ae6b9)

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

libpng12: remove prefer version and add it to lsb packagegroup

Because rename libpng_1.2.50 to libpng, remove the perfer verion from
default-versions.inc and add libpng12 to lsb packagegroup.

(From OE-Core rev: 01fa98083df0931e07e8715616dafe600258adba)

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

libpng12: rename libpng_1.2.50 to libpng12

As Mark's suggestion, rename libpng_1.2.50 to libpng12 that
multi-versions libpng could coexist.

We want to make sure we have both the old and new versions to meet LSB
compliance (for people who have that enabled) as well as the new version
for newer applications.

And drop link files that conflict with higher version.

[YOCTO #4221]

(From OE-Core rev: fc626e6861e491b0144b813a5b48b0f5f57664e6)

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

openssl: Upgrade to v1.0.1e

Dropped obolete patches and pulled updates for debian patches.

Addresses CVEs:

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2012-2686
http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0166
http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0169

[YOCTO #3965]

(From OE-Core rev: 0470edd01c0aebaa78db137e365a7e22bfb199e9)

Signed-off-by: Radu Moisan <radu.moisan@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

classes/buildhistory: fix interaction with rm_work

Change do_write_srcrevs to a postfunc of do_fetch, avoiding a dependency
being created that causes large numbers of setscene tasks being executed
on every build with both buildhistory and rm_work being enabled.

(From OE-Core rev: a751e9042dfffcc5c4701634a1f1f598012d609c)

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

rpm: fix RDEPENDS

The rpm-postinsts runtime dependency was overwritten.

(From OE-Core rev: 834ea4ed891c874e0336abb8f0b96664250208c9)

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

qemu: Add backported patch to address random segfaults

We've been seeing random sefgaults on a variety of architectures which appear
to be from an issue in qemu. The attached backport from upstream appears
to fix these.

[YOCTO #4216]

(From OE-Core rev: 55a22b7341571179d5e026d102953a6d9f2045bf)

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

oe-buildenv-internal: Only add to $PATH if needed

If $PATH already has the needed paths at the beginning, there is no need
to add them again. This allows rerunning oe-init-build-env for the same
directory without having $PATH increase unnecessarily every time.

(From OE-Core rev: 161abcd3672f83990ede03d67b7388678c07150e)

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

alsa-tools: Fix sys/io.h patch

I blew my #if expression!

(From OE-Core rev: b458309845185a3cd473daa0969ce17e2ff5c602)

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

kernel.bbclass: do_sizecheck: update path to build image and do not delete

do_sizecheck has a few issues especially with vmlinux image type.

It breaks because KERNEL_OUTPUT is a path relative to ${B}.  When
do_sizecheck runs it does not find the file (because the working
directory is elsewhere) and does not fail.

Also, the image file referenced by KERNEL_OUTPUT may be a link.

Finally, when do_sizecheck deletes the oversized kernel image it leaves
the previously run do_compile task with inaccurate status.

So, do the following:
 - specify that the working directory should be ${B}
 - use ls -L to reference to the real file, and ensure that the link
   file is created
 - keep the oversized image file so the status of do_compile is valid

[YOCTO #3514]

(From OE-Core rev: f0b19ddce3c92c5d06976cf73d4c4c480e053dff)

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

layer.conf: add systemd-serialgetty to SIGGEN_EXCLUDERECIPES_ABISAFE

* it was imported from meta-systemd without SIGGEN_EXCLUDERECIPES_ABISAFE
  change

(From OE-Core rev: 40c6090e67fe4def94223954e4ada01115f267dd)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

scripts/oe-pkgdata-util: find complementary packages for split packages

Check after getting the original package name (e.g. undoing Debian
renaming) if there is a complementary package for that name, e.g. if
the glob is *-dev, then libudev0 -> libudev -> libudev-dev.

Fixes [YOCTO #4136].

(From OE-Core rev: 84a1c6922934a99e8afee0185e58dc4789b54a22)

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

bluez4: add readline dependency

bluez4 uses readline to be build, but the dependency is not listed
This is listed in the configuration log.
So we add it.

(From OE-Core rev: 99194be0332ac35da729ec53a2cc423cc520db28)

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

systemd: Set the default firmware path to enable firmware loading in udev

After some breakage in udev the kernel gained direct firmware loading.
For older kernels (e.g. 3.2 in my case) udev still needs to load the
firmware. Firmware loading is enabled once a default firmware path is
set. Apply a compile fix from the upstream project.

(From OE-Core rev: 2009c6899d7d4ddd71350b1026a27336dc3a94b8)

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

tcf-agent: Use kill instead of killproc to stop agent

When shutting down a core-image-lsb-sdk image, there is a lot of time spend stopping tcf-agent,
which slows down the whole process. The reason for this slowdown is the fact that it tries in a
loop to kill tcf-agent service by using killproc with the path of the executable and killproc
does not seem to available in lsb images. This patch fixes the issue by using "kill" instead of
"killproc".

[Yocto #3928]

(From OE-Core rev: 251361eb78176a04e3da00e0f77b7f3ff459d571)

Signed-off-by: Ioana Grigoropol <ioanax.grigoropol@intel.com>
Signed-off-by: Bogdan Marinescu <bogdan.a.marinescu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

smart: disable CHANNELSDIR

Make CHANNELSDIR in smart empty, since this causes host contamination issues
on some RPM-based hosts on which smart is already installed.

[YOCTO #3881]

(From OE-Core rev: 94e76a98b6cdafe9547630be159401ac1d8c5edd)

Signed-off-by: Bogdan Marinescu <bogdan.a.marinescu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gtk-update-icon-cache-native: create wrapper script

When using the sstate from another build machine, the path to the pixbuf
loader's cache points to a path on the remote machine. Hence, the update
of the icon cache fails on host.

(From OE-Core rev: f2cb906bdce08441a20eab927ca9e2a2a9735ed0)

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

image.bbclass: fix postinstall intercepts fallback

The wrong type of paranthesis was used so 'continue' did nothing (was in
another context) and the packages were marked as installed.

(From OE-Core rev: 0bdde53e885aae3506c7b070b6e21f64a7cd4115)

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

gst-ffmpeg: fix --disable-yasm

The gst-ffmpeg build shows the following warning:

configure: WARNING: unrecognized options: --disable-yasm

which means that the following test in configure always fails and
--disable-yasm never gets passed to the embedded ffmpeg build:

'if test "x$disable_yasm" = "xyes"; then'
  embffmpeg_configure_args="$embffmpeg_configure_args --disable-yasm"

commit 4d309730 ['gst-ffmpeg: configure-fix patch used wrong test']
actually fixed the obviously backwards syntax by reversing the test -
prior to that, --disable-yasm would always unconditionally be passed
into the embedded ffmpeg config.

This fixes things so that the variable actually exists and makes the
test meaningful.

(From OE-Core rev: da9515621134c26e54f43b96cdad0c6e6c5876bf)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

insane.bbclass: Updated MicroBlaze machine definitions

* Removed existing definition with machine 47787, this
  definition is outdated, a sanity error should occur if an ELF uses this
  value.
* Added new definition with machine 189. This value replaces the existing
  value since August 2009. See binutils thread for more information.
  (http://sourceware.org/ml/binutils/2009-08/msg00127.html)

(From OE-Core rev: 0c60d3b04eb77629abc3bbc2a6d8a2b8f0a44309)

Signed-off-by: Nathan Rossi <nathan.rossi@xilinx.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemuimage-tests/sanity/boot: Increase timeout

As we've increased the parallelisation on the build servers, we've started to see
core-image-minimal sanity test boot failures where the network never comes up. We
don't see those failures for core-image-sato, its always minimal.

Looking at the results, it can take ~100 seconds for the network to come up,
even on the sato images if the machine has a high load. The timeout for the boot
test is only 120 seconds compared to 400 on every other test.

This change makes the timeout equal for all the tests at 400 seconds in the hope
that the load on the autobuilder is causing the sanity tests to run slowly and
hence triggering the false negatives.

(From OE-Core rev: 331118a253e26821011a31ca9087611ea58a18b8)

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

oe-init-build-env: Make it use the correct $OEROOT with zsh

(From OE-Core rev: f0aa69296f4c1d4214f9dbea236b0ed330b8154b)

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

oe-setup-builddir: Allow $OECORENOTESCONF to not exist

(From OE-Core rev: 6fc14169ac0c3001e3a69eda8d07fc0ac93a15ee)

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

curl: backport patch to fix segfaults

* e.g. ecore, efreet segfault a lot without this patch

(From OE-Core rev: b93011d3e719c46089ccdb39c60d3a9e9cfa5a14)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: change qemu-based mips BSP default branch

The default branch for the qemu-based mips BSP template no longer
exists, so change to one that does.

(From meta-yocto rev: 5af614322269ee7c79928d1ff343f2e3bcf35509)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: set SRCREV for arm-based qemu machines

arm-based qemu machines won't boot with the default 3.8 machine SRCREV
because it's missing the commit 'arm: add dummy swizzle for versatile
with qemu', so we need to use a SRCREV that has it merged.

(From meta-yocto rev: 176ec06589032b0b589da8345adfc87dddcb74f0)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: qemu machine template updates

A few small changes to the machine.conf from the previous version that
should be incorporated.

(From meta-yocto rev: 05a86a2e8d69b32243ab1915b279411d3d82235f)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: use specific bsp metadata for qemu machines

For the qemu-based BSPs, use bsp metadata that's guaranteed to boot in
qemu.

(From meta-yocto rev: e274a2e66c26489a4da895194eb6e7a9c1476a73)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: pass in file object to replace_file()

Pass the file object instead of the filename to replace_file for the
custom template, as now required by replace_file().

(From meta-yocto rev: 56091c019000cfe3d22ec464c596d97ae78fc619)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: have replace_file() close file before copying

replace_file needs to make sure the file it's replacing is closed
before replacing it, otherwise unexpected results may ensue.

Fixes [YOCTO #4145].

(From meta-yocto rev: 1339dbb690d51456b4474356992e430638469e47)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: add linux-yocto-3.8-rt to templates

RT support is now available in the linux-yocto-3.8 kernel, so we can
also add that as kernel option for users.

(From meta-yocto rev: 2e425b5c6c7e685e8a0e0c8cb2cf64040e454cad)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

yocto-bsp: add KBRANCH for existing kbranch cases

For the cases where a BSP reuses an existing branch, we still need the
KBRANCH in order to be able to specify an existing branch.

(From meta-yocto rev: 5a3167c4fa6cb53ec501e9de185b93748973ec18)

Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-yocto-tiny: Add 3.8.4 recipe

Bring linux-yocto-tiny up to the latest linux-yocto 3.8.4 version.

(From OE-Core rev: f2a2256f08287baac9cf1f9fd58b5b8244c09610)

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-firmware: make the main package depend on all sub-packages

Whilst splitting out specific large firmware blobs is a good move for space
saving, it makes installing "all the firmware" tricky.

Make linux-firmware depend on all of the separated packages so that installing
that pulls in all of the sub-packages.

(From OE-Core rev: 644dfe0b13f68a04bdde67b5f1bf210bbe8ab918)

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

shadow: add patch to fix crypt: Invalid Argument

This patch came from Slackware and address a change in crypt()'s handling
of an invalid seed, which in the past returned an encrypted string and now
returns a NULL.

[YOCTO #4097] related to tinylogin segfault

(From OE-Core rev: a7f7e6da8383b4bde6d8ce951e5c3c955073c0bd)

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

tinylogin: fix segfault from crypt()

In glibc 2.17, crypt() now expects 2 valid chars for the seed or
it will error out and return a NULL. The tinylogin code took the
result from crypt directly into a strcmp() which caused a segfault

Tinylogin has been deperacted, busybox now has login support, I will
investigate using busybox login support for 1.5.

[YOCTO #4097]

(From OE-Core rev: 03034e0f5dff426ee7adaa2364082dd47c23260a)

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

sanity/connman: when connman test fails, dump syslog

(From OE-Core rev: a51041db57666c60f39c4effa4aceb53cae815dc)

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

qemuimage-testlib: add function to fetch the remote syslog

Add a new function to scp from the target, and another to fetch
/var/log/messages and dump it to the console.

(From OE-Core rev: f94cb0d175309ad6b29598c57ba74cf1c3646661)

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

qemuimage-testlib: silence some key warnings

Set StrictHostKeyChecking to no to silence the fingerprint warnings, and instead
of creating a temporary file for the known hosts and then deleting it just use
/dev/null.

(From OE-Core rev: 24e4a570eb527cff017386976296d5747c1adf57)

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

connman_test.sh: show all processes when dumping ps

We know the grep failed because the error case is being executed, so don't do
the grep again when attempting to help diagnose the problem, as seeing the full
process list might be useful.

(From OE-Core rev: 6ee4a2ba6ee9633c1fa08d3b162d6d00da307798)

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

lib/oe/classextend.py: avoid extending any kernel package

For multilib and other uses of classextend, we don't want any
dependencies on kernel packages to be extended since there should only
be one kernel variant.

Fixes [YOCTO #2918] (where kernel-dev was being extended.)

(From OE-Core rev: b684c0f0d5d93f5147dee79951647eb3ddf4c840)

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

postinst-intercepts, qemu.bbclass: fix issue on 32 bit hosts

The intercept scripts fail to run on 32 bit hosts. Apparently, the
current approach worked on 64 bit hosts due to the larger virtual address
space (probably). On 32 bit hosts, however, calling the target binary like:

qemu-arm ld-linux.so --library-path /lib:/usr/lib arm_binary

fails with:

arm_binary: error while loading shared libraries: arm_binary: failed to
map segment from shared object: Operation not permitted

When run like this, qemu-arm fails to map the arm_binary executable in
memory because it's hitting the lower limit of
/proc/sys/vm/mmap_min_addr. That's because it loads the
ld-linux.so binary successfully, taking into account mmap_min_addr, runs
it, and then ld-linux.so will map the arm_binary at a fixed address but this
will fail because it is below mmap_min_addr. The qemu's guest base probing,
apparently, doesn't work fine when a program runs inside other.

One way around this would be to set mmap_min_addr to 0 (on recent
distributions is set to 65536 to avoid "kernel NULL pointer dereference"
defects) but this approach is not safe.

The other way is to call the binary directly but providing qemu with a
prefix (-L option) in order to find the elf interpreter correctly. This
way, both the target binary and dynamic loader are mapped into memory
under qemu's control and, only after, the dynamic loader is started.

[YOCTO #4179]

(From OE-Core rev: 78f91e08c8a7b0f0c831a087f7c89e2c76047e7a)

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

site/common-uclibc: add predefined configure vars for coreutils

* this sets some configure vars which will be guessed
  false in cross-compile case for uclibc

(From OE-Core rev: c5337326005c975425b1eb2b62796e9b33f72ac3)

Signed-off-by: Henning Heinold <heinold@inf.fu-berlin.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemuimage-testlib-pythonhelper: Fix process mixups

runqemu-internal runs "ldd qemu-system xxx" and the detection code was returning this
as the PID of qemu. This patch improves the detection code to avoid this problem,
fixing certain race type failures on the autobuilder.

(From OE-Core rev: fc914a6fb3204f8b5bdfc0f56364606673d5356a)

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

scripts/sstate-cache-management.sh: fix return value by adding exit 0

* usefull for jenkins jobs, which will otherwise fail
  because 1 was returned

(From OE-Core rev: a864de0f2a326f857125229fc986845044931196)

Signed-off-by: Henning Heinold <heinold@inf.fu-berlin.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

site: add endianness information for libmemcached

(From OE-Core rev: 2587a33134fde80dd1367629d9def45ac70256ee)

Signed-off-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

dbus: set INHIBIT_UPDATERCD_BBCLASS without sysvinit in features

* fixes udev configure in run-postinsts failing with:
  update-rc.d: /etc/init.d/dbus-1: file does not exist
  because dbus-udev is installed only with sysvinit in features
  but update-rc.d was always called from PN postinst

(From OE-Core rev: f15192a65e02026308253e6723b990b24780be5b)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

systemd: set INHIBIT_UPDATERCD_BBCLASS without sysvinit in features

* fixes udev configure in run-postinsts failing with:
  update-rc.d: /etc/init.d/systemd-udev: file does not exist
  because systemd-udev is installed only with sysvinit in features
  but update-rc.d was always called from PN postinst

(From OE-Core rev: b1dca3a693bb439181a155c5248a2c6a900f729d)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

utils: add helper to get all non-system packages

For example if PACKAGES is "foo foo-data foo-dev foo-doc", this will return
"foo-data".

(From OE-Core rev: 3115187e468398a8c1edaf3e5369a2d10fb112f4)

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

oe.terminal: add tmux classes

This adds two new Terminal classes. It's separated into two, so that opening
a split inside a tmux window is preferred to the other terminal types, but
opening a tmux session is prioritized only slightly higher than screen.

- tmuxrunning: Open a new pane in the current running tmux window. Requires
  that the TMUX variable be added to the env whitelist to use it.
- tmux: Open a new tmux session

(From OE-Core rev: 31c58d584f838738a6b6258b87b1c7e6ca173086)

Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

liberation-fonts: remove 1.06

1.06 requires fontforge-native to build, which as we don't have this version has
never been used.

(From OE-Core rev: 035e074cb7ff943defe3a10dc2a73b3cb2fd7e96)

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

libxcb: remove obsolete version 1.1.91

We have 1.9 and git snapshot recipes, we don't also need this ancient version.

(From OE-Core rev: b037ac6f6e319c14895b2d3d7dd1b4a72a143670)

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

scripts/qemuimage-testlib: Dump extra info if the network doesn't come up

(From OE-Core rev: db4a4cc8ba8082a27224a3e55fb5e8eb7de2bbe7)

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

bitbake: propertydialog adjustments for package.bbclass

After moving the code from packageinfo.bbclass to
package.bbclass, minor adjustments were made to the
parsing of the package items.

(Bitbake rev: 414fe98fe367ba44ea0fd20d8fe1296bcef58ea6)

Signed-off-by: Andrei Dinu <andrei.adrianx.dinu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

Add file information to package information window

Removed the package files parsing routine from the
packageinfo.bbclass file and added it to the
package.bbclass file.

(From OE-Core rev: 225e7826b0d082f43db82201e826b98b3a95cd57)

Signed-off-by: Andrei Dinu <andrei.adrianx.dinu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-dtb: fix whitespace in bash functions

* Fix the whitespace in the base functions to use TAB instead
  of spaces.  This is to address feedback from:
        Darren Hart <dvhart@linux.intel.com>

(From OE-Core rev: ab6630df4d100ff501b33a1c7ec9d1e6a2d4f0ee)

Signed-off-by: Chase Maupin <Chase.Maupin@ti.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

linux-dtb: Add simple DTB symlinks for devicetree

* This is similar to the symlinks provided for the kernel image
  in the /boot directory of a file system.  The goal is to have
  simply named symlinks in /boot that mirror the device tree
  name in the kernel sources.  This is so that programs like
  U-Boot can easily find the default device tree binary in the
  /boot directory and use that when booting the kernel.
* Use update-alternatives to handle proper creation and removal
  of the symlinks.

(From OE-Core rev: 750a9554e1b85d9bd23d18e0630723c3c193c604)

Signed-off-by: Chase Maupin <Chase.Maupin@ti.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

pointercal: Move override files from 'files' to 'pointercal' directory

(From OE-Core rev: b9bb3ce5b92cde6bff08e9cb2fd27c576c92125d)

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

gcc: add patch to disable texinfo when texinfo is 5.0 or newer

* this is needed only for 4.7 series, newer works fine with texinfo-5*

[YOCTO #3947]

(From OE-Core rev: d85d15972d78b5dda7a03dd273a64305f115282b)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemu: Add 2 patches to fix build with texinfo-5

(From OE-Core rev: af65260dbf17fcd47b6630db473d95f2f3225d68)

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

qemu: Enabled SDL when compiling for target architecture

Enables qemu to run images with video output without the need for vncviewer.

(From OE-Core rev: 30d5c1d5bc9a3931a09425962d980a3571dc56f3)

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

python: adding missing runtime dependency python-io to python-pprint

When trying to import python-pprint on a minimal image, it reports that
the cStringIO python module is missing.
This is provided with python-io, so we add python-io as runtime
dependency.

The complete observed trace was:

Python 2.7.3 (default, Apr  4 2013, 07:45:36)
[GCC 4.7.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import pprint
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python2.7/pprint.py", line 40, in <module>
    from cStringIO import StringIO as _StringIO
ImportError: No module named cStringIO

(From OE-Core rev: abe7bf9992e298f1b53e790eee7b064a9e4e8589)

Signed-off-by: Lukas Bulwahn <lukas.bulwahn@oss.bmw-carit.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libproxy: add dependency on glib-2.0

libproxy uses glib-2.0, but the depends is missing

Fixes intermittent build errors.

(From OE-Core rev: 9d8543d98f40d5260ee3830305d83e27412351c3)

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

rpm-postinsts: Split out run-postinsts

This patch allows for the run-postinsts script to be provided outside
of the rpm package itself and not pull in all the associated build
dependencies.

[YOCTO 4175]

(From OE-Core rev: 7841ee7041d04f11a3d879fb5bc60bb37de0a5c0)

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

gst-ffmpeg: configure-fix patch used wrong test

(From OE-Core rev: d07bf78a11c5aee37da653404f8aaf413cf14e8f)

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

icon-naming-utils: import version 0.8.90 from meta-oe

* Use newer version 0.8.90
* Updates to a BBCLASSEXTENDed recipe instead of just a native recipe
* Use PV in SRC_URI instead of hardcoded version number

In copying over the recipe from meta-oe, some minor changes were made:
* Preserve the existing OE-Core nativeperl wrapper usage
* Drop setting of S which is effectively the default value

(From OE-Core rev: ef24f8e7e6b91ad8e83942bd956e0d6ab0fc077b)

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

xserver-nodm-init: Add xuser to input group

Fixes [YOCTO 4164](3/3)

Input devices come and go, so a single chmod in this init script is not
adequate to ensure rootless X servers can use input devices.

The o+rw method also introduces a security hole.

The newly added input group and input udev rule address this in a secure
way. Ensure the xuser is added to the input group.

(From OE-Core rev: 150b7ac8e1c0f029b90f63424867ee5347821cf7)

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: Saul Wold <sgw@linux.intel.com>
Cc: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

udev-extraconf: Add rule adding input devices to input group

Fixes [YOCTO 4164](2/3)

Add all /dev/input/* devices to the input group with g+rw.  This is
needed for rootless X without adding a security hole by making the
device o+rw.

(From OE-Core rev: 66c9b46f987f3e4f1f9b7b11d1ae157897454f07)

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: Saul Wold <sgw@linux.intel.com>
Cc: Radu Moisan <radu.moisan@intel.com>
Cc: Ross Burton <ross.burton@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

base-passwd: Add input group

Fixes [YOCTO 4164](1/3)

Add input group for the /dev/input/* devices. This is needed for
rootless X without adding a security hole by making the device o+rw.

(From OE-Core rev: 262234ab50636463f03fd4daccecc1232682ff59)

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

buildhistory: record tag names and show warning when the same tag corresponds to different revision

* persistent cache records tag-srcrev mappings, but is not shared between builders
* when tag is moved in remote repo, all builders should rebuild the component to
  use the same source, show warning when revision is different than what was used
  in last build

(From OE-Core rev: 0bc22ed6bd67031749e8f2cb5415dabf933eef56)

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

classes/sstate: avoid traceback when no files have been staged

If no files have been staged we want to continue without error instead
of showing a traceback.

Fixes [YOCTO #4056].

(From OE-Core rev: ca36be708e54c0c86535bc8512295c76c48f6cf5)

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

qemu script: explicitly set 32 bit depth

Qemu update from 1.2 to 1.4 now allows for 16bit depth in guests,
whereby previously only 32bit depth was supported. However,
the new support is broken, so we force 32bit depth in all cases.

MUST_REVERT: on qemu update, if 16bit depth support is working ok

Fixes [YOCTO #3828]

(From OE-Core rev: 354377789628d96fa589cb5721134f631815cfeb)

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

gst-plugins-bad: disable librsvg when x11 is disabled

librsvg depends on gtk+, which in turn does not support framebuffer
as backend in current version (2.15.24). This patch disables librsvg
when x11 is not in the distro.

(From OE-Core rev: 022cc0d3f0f7468428d708c27dbc561f619ee841)

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

alsa-tools: fix build when x11 and gtk+ not available

Current verion of gtk+ (2.15.24) does not accept pure framebuffer as
backend and some alsa-tools sub-modules depend on gtk+.

This patch removes those sub-modules from the build only when x11 is not
set in DISTRO_FEATURES.

(From OE-Core rev: e611bba7bba02ba167b2ae3671b00cc99e4fb29c)

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

runqemu-internal: use MACHINE_SUBTYPE variable for qemuarm*

(From OE-Core rev: ed07bb4214abb472da6aa7e164a20fd4be127e54)

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

package_deb.bbclass: fix 'armel' override

The 'armel' override for DKPG_ARCH was causing the meta-toolchain
build to fail. The assignment was moved to an anonymous fragment
of Python code, so it doesn't affect the assignments in
cross-canadian.bbclass anymore, thus fixing the issue.

[YOCTO #4080]

(From OE-Core rev: 6f86fe5d66e401377bccd9f635270033b99a9f4b)

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

qt-mobility: added list of modules to be compiled

This is useful for users that want to .bbappend this recipe to select specific
modules to be compiled.

(From OE-Core rev: 2ddb7afd15e53ef75b5084d691115e0f58ff24ab)

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

util-linux: Use u-a for getopt

* when enable busybox installs getopt to ${base_bindir} and
  util-linux to ${bindir}, so there is no file conflict, but
  because busybox implementation does not support --long used
  by lsb_release (which RDEPENDS on util-linux) we need to use
  util-linux getopt even when busybox defconfig has it enabled

(From OE-Core rev: 2dcc867247b402bb4223cc7b9861088958599866)

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

archiver: fix srpm archiving build errors

srpm archiving doesn't need to be handled as a different case
when deciding what archive tasks to add.

When srpm is selected as the archiving type, the scripts and logs
archive staging directory ${WORKDIR}/script-logs is cleaned, and
its contents moved out to ${WORKDIR}.

Now that we are including ${WORKDIR}/script-logs in sstate-inputdirs,
the directory must be preserved.

[YOCTO #4032]

(From OE-Core rev: 0c80286a3383b436a0a63a0b00eb357dd9dea4fb)

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

base.bbclass: Update the preferred_ml_updates

When processing the blacklists, we should avoid cross-canadian packages, as they
will not have any multilib prefixes to expand.

Similarly look for "virtual/nativesdk-" in addition to the existing "nativesdk-".
These items should also be ignored.

Finally, in order to avoid undeterministic variable key expansion, such as:

MYVAR = "foo"

PREFERRED_PROVIDER_${MYVAR} = "bar"
PREFERRED_PROVIDER_foo = "foobar"

during the multilib processing of PREFERRED_VERSION and PREFERRED_PROVIDER,
the code was changed to rename the variable key, to the final key.  This along
with the existing code avoids the problems.

(From OE-Core rev: 1416613e94af46c6e74532bca0f026d1540becbb)

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

update-rc.d, systemd: redirect also stderr from type

* different shells different behavior?
  bash prints 'type: update-rc.d: not found' on stderr
  busybox's sh on stdout

(From OE-Core rev: 45e22312c48b23480bd6dff98702b0691a48f7d1)

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