bitbake: codeparser cache improvements

It turns out the codeparser cache is the bottleneck I've been observing when running
bitbake commands, particularly as it grows. There are some things we can do about
this:

* We were processing the cache with "intern()" at save time. Its actually much
  more memory efficient to do this at creation time.

* Use hashable objects such as frozenset rather than set so that we can
  compare objects

* De-duplicate the cache objects, link duplicates to the same object saving
  memory and disk usage and improving speed

* Using custom setstate/getstate to avoid the overhead of object attribute names
  in the cache file

To make this work, a global cache was needed for the list of set objects as
this was the only way I could find to get the data in at setstate object creation
time :(.

Parsing shows a modest improvement with these changes, cache load time is
significantly better, cache save time is reduced since there is now no need
to reprocess the data and cache is much smaller.

We can drop the compress_keys() code and internSet code from the shared cache
core since its no longer used and replaced by codeparser specific pieces.

(Bitbake rev: 4aaf56bfbad4aa626be8a2f7a5f70834c3311dd3)

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

bitbake: cache: Optimise invalid cache file handling

If there is a corrupt/invalid cache file, we'd keep trying to reopen
it. This is pointless, simplify the code paths and delete the dead
file.

(Bitbake rev: c22441f7025be012ad2e62a51ccb993c3a0e16c9)

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

bitbake: cache: Don't reload the cache file since we already have this data in memory

If we're writing out merged data to disk, its safe to assume that either
we loaded the data or couldn't. Loading it again is relatively pointless
and time consuming.

(Bitbake rev: 14c31e18e4a4a52ee54056aeefb09ff8c295b475)

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

insane: Use a warning for FILEEXTRAPATHS issues for now

(From OE-Core rev: f907b3a5b414aa53ce8271763f04532d7127006f)

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

lib/oeqa: Fix accidental revert of code

(From OE-Core rev: 188545ba82119d75f80dde322a73712ce1f0f762)

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

wic: squashfs partition support

It is possible to instruct wic to create a squashfs partition by setting
--fstype=squashfs in *.wks. For now this is only useable for rootfs
partitions (note that you must have squashfs support in the kernel). An
attempt to create an empty partition will produce a warning.

(From OE-Core rev: 30266a0be946bd0ce76d6920e7afe840c6c3bf80)

Signed-off-by: Maciej Borzecki <maciej.borzecki@open-rnd.pl>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

wic: --fsoptions handling

Add handling of --fsoptions in parition definition. If no options are
specified, 'defaults' is used.

(From OE-Core rev: e91c0db15db74237606fae96c9b7d21936519a86)

Signed-off-by: Maciej Borzecki <maciej.borzecki@open-rnd.pl>
Signed-off-by: Maciek Borzecki <maciek.borzecki@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

insane: add checking to standardize how .bbappend files do FILESEXTRAPATHS

When adding patches or config files from bbappend files, it requires
the use of FILESEXTRAPATHS, which has been an issue and failure point
for people starting to work with bitbake and oe-core.

We add checking to standardize how to use FILESEXTRAPATHS. Only the
format of:
  FILESEXTRAPATHS_append := ":${THISDIR}/Your_Files_Path" or
  FILESEXTRAPATHS_prepend := "${THISDIR}/Your_Files_Path:"
is acceptable.

[YOCTO #5412]

(From OE-Core rev: 69e083237e632f7d84a7b218dd12d1a5ad95a229)

Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

oeqa/rutime: Added testcase decorators for automated runtime tests. Also added LogResults decorator for oeTest class in oetest.py

(From OE-Core rev: 95b83084487d0712362ade8ac487999c3274bb96)

Signed-off-by: Lucian Musat <georgex.l.musat@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

alsa-lib: remove non PN based -dev packages

All dev related items should be packaged in the core
PN-dev package not in seperate packages.

(From OE-Core rev: 9aa75f06e9e78602ad8e9a5ffa312f71057a90d7)

Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

cogl-1.0: Add libcogl-path package

The new library needs to be put in a specific package, or rootfs
generation fails as:

,----[ Error in rootfs generation ]
| Updating cache...               ########################### [100%]
| Computing transaction...error: Can't install
|   libclutter-1.0-0-1.18.2-r0@cortexa9hf_vfp_neon_mx6: unable to
|   install provider for libcogl-path.so.20:
|      error: libcogl-path20-1.18.2-r0@cortexa9hf_vfp_neon_mx6 is
|        conflicted by libcogl20-1.18.2-r0@cortexa9hf_vfp_neon_mx6
`----

Upstream split this library out from libcogl20.so, sothe libcogl package
no longer gets renamed to libcogl20 as it contains multiple libraries.

(From OE-Core rev: 9d63188950ecd1877adf1e40de1ca1928729d866)

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

autotools: Improve configure dependency code for finding m4 files

We have an open bug about the warnings issues in builds from an sstate
cache when something like glib-2.0 gets rebuilt. The issue is that
sstate is "clever" and prunes unneeded dependencies out the tree.

For example is X depends on pkgconfig-native but we've already build X
and installed it from sstate, it will not get installed when you build Y
which depends on X.

This patch changes the logic to match the sstate behaviour and prune out
unnecessary dependencies from the scope of aclocal. This in turn removes
the warning about missing manifest files.

The issue is that this patch exposes holes in our DEPENDS in recipes,
specifically that some native tools are not listed, specifically, and
problematically, pkgconfig, gtk-doc and intltool-native in particular.

I've sent out patches against OE-Core that address the bulk of the
issues there however I'm conscious this is probably going to a bug issue
in other layers and may be too annoying to consider at this point. The
other alternative is simply to turn the warning into a debug statement.

I appreciate the code below has commented blocks, this is simply debug
I've left around for now. It will be cleaned from any final version.

(From OE-Core rev: 6d2dc279faa8b28a00895dc6a620d80e2dbac685)

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

wget: fix build error with gnutls

This fixes '/usr/lib/libgnutls.so: error adding symbols: File in wrong format' in do_configured.

Reported-by: Jack Mitchell <ml@communistcode.co.uk>
(From OE-Core rev: 4e1649fe2fa2d512384324efcb1f9e04efeadd81)

Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

shadow-securetty: add freescale lpuart

Add Freescale lpuart tty's (ttyLPx) to securetty. Freescale Vybrid
devices running upstream kernel use this driver.

(From OE-Core rev: 1732ea461f1d0afe254e36a8bfe84a6675a42e66)

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

python: fix _json module arbitrary process memory read vulnerability

http://bugs.python.org/issue21529

Python 2 and 3 are susceptible to arbitrary process memory reading by
a user or adversary due to a bug in the _json module caused by
insufficient bounds checking.

The sole prerequisites of this attack are that the attacker is able to
control or influence the two parameters of the default scanstring
function: the string to be decoded and the index.

The bug is caused by allowing the user to supply a negative index
value. The index value is then used directly as an index to an array
in the C code; internally the address of the array and its index are
added to each other in order to yield the address of the value that is
desired. However, by supplying a negative index value and adding this
to the address of the array, the processor's register value wraps
around and the calculated value will point to a position in memory
which isn't within the bounds of the supplied string, causing the
function to access other parts of the process memory.

(From OE-Core rev: 9ec213bf67afbdfdbe25802ec86487bb22aeb2e4)

Signed-off-by: Benjamin Peterson <benjamin@python.org>
Applied to python-native recipe in order to fix the above mentioned
vulnerability.

Upstream-Status: Submitted

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

nspr: Fix for CVE-2014-1545

Mozilla Netscape Portable Runtime (NSPR) before 4.10.6 allows remote
attackers to execute arbitrary code or cause a denial of service
(out-of-bounds write) via vectors involving the sprintf and console
functions.Per: http://cwe.mitre.org/data/definitions/787.html

(From OE-Core rev: 191cab2f679491c2b6ddba49c5cf4886dcd22f57)

Signed-off-by: Xufeng Zhang <xufeng.zhang@windriver.com>
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gcc-4.9.inc: fix parallel building failure

In subdir 'gcc', Most C source files included config.h which was
generated by a rule. But no related prerequisites was added to
the C source compiling rule. There was potential building failure
while makefile enabled parallel.

The C source compiling rule used suffix rule '.c.o', but the suffix
rule doesn't support prerequisites.
https://www.gnu.org/software/make/manual/html_node/Suffix-Rules.html

We used the pattern rule '%.o : %.c' to instead, and add the config.h
as its prerequisite

We also moved the '%.o : %.c' rule down to the 'build/%.o :' rule, which
makes '%.o : %.c' rule doesn't override 'build/%.o :'.

[YOCTO #6568]

(From OE-Core rev: 86c2483f0fe05fb763d280ae22d70e54cb4bb0bc)

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

perl: fix rebuid failed while ${CC} changed

Reproduce steps:

1) bitbake perl
2) vim local.conf to tweak CC, just add redundant option.
...
CC_append = " ${HOST_CC_ARCH}"
...

3) bitbake perl
...
./miniperl -Ilib make_ext.pl lib/auto/Time/HiRes/HiRes.so MAKE=make LIBPERL_A=libperl.so LINKTYPE=dynamic
    Making Time::HiRes (all)my $filename=
Deleting non-Cross makefile
Running Makefile.PL in cpan/Time-HiRes
Makefile.PL: The "xdefine" exists, skipping the configure step.
("tmp/sysroots/x86_64-linux/usr/bin/perl-native/perl5.20.0.real Makefile.PL --configure" to force the configure step)
Warning: No Makefile!
make[2]: Entering directory `tmp/work/armv5te-poky-linux-gnueabi/perl/5.20.0-r1/perl-5.20.0/cpan/Time-HiRes'
make[2]: *** No rule to make target `config'.  Stop.
...

While ${CC} changed, the existance of 'xdefine' caused makefile
regeneration failed.

[YOCTO #6569]

(From OE-Core rev: fa43d4f268bc4a6fafcf14029049f2997bc72d6c)

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

autotools.bbclass: Enhance sed regexp to avoid extra subshell

head -n1 can be done using sed.

(From OE-Core rev: 38d50331acbc1e279449ae6b313ff0116b6c44e8)

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

oeqa/runtime: Added skipModule import for test modules that use it.

The modules that use skipModule should import it themselves and not rely on somebody else to import it.

(From OE-Core rev: 6a14db407d471e717f41342ac0700e6a383c32c3)

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

oeqa: Refactor test skipping decorators to use the unittest result object

In order to make the test skipping decorators independent of the oeTest object we rely on the unittest result object to construct skip, fail and error lists used by these decorators.
Created a new object getResults that analyses upper frames and retrieves the unittest result object instance, then return a list of failed, skipped and error tests.
Also removed the oetest import from decorators.py because it was no longer required.

(From OE-Core rev: 4d2d201158236bd4c72546cf8db88681ff921b11)

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

gstreamer1.0-plugins-good: add missing dependency on libcap

(From OE-Core rev: e1ac22b340a66e3da53182faf0d624db9ccbdee1)

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

gst-plugins-good: add missing dependency on libcap

(From OE-Core rev: ad68fd0e73a549356741bb050c4343ebd4a1e2fb)

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

pulseaudio: add missing dependency on libcap

(From OE-Core rev: 4f0a1fc6c324840459919b80a9c096efdd2bf5bd)

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

debian.bbclass: inherit package

This class defines a package_name_hook implementation but as EXPORT_FUNCTIONS is
order-dependent it needs to inherit package.bbclass first to ensure that the
"base" definition in there is defined first, otherwise with a suitable inherit
order of debian and then a packaging class can result in the stub
package_name_hook being incorrectly used.

(From OE-Core rev: 3aa52f85e774bf82e8470a8fd849bd6953930720)

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>

base-files: Don't export TZ if /etc/TZ exists

While glibc/eglibc looks for the time zone in /etc/localtime, other libc
alternatives (e.g. uclibc) may look for it in /etc/TZ.  If /etc/TZ
exists, don't fall back to setting TZ to "UTC" in /etc/profile.

(From OE-Core rev: a69c3ba99be1502c50d3d1eb9ff66f135884ed76)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

glib-2.0: use ptest-gnome

(From OE-Core rev: fb32cd7abac268747347ee5cff98a4d07dc4b780)

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>

ptest-gnome: add ptest helper for GNOME packages

Many GNOME packages are using their InstalledTests pattern, where the test suite
can be trivially installed.  To avoid repeating the same logic over and over,
add a class to encapsulate this.

(From OE-Core rev: 6a66fadbd7d7675c61c342f9c338edbdedf8b92c)

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>

shared-mime-info: upgrade to 1.3

(From OE-Core rev: 8abd1d0d605cec77faca1f21663feb4ad9e27d1d)

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-evdev: upgrade to 2.9.0

Build-dependency on libevdev added.

(From OE-Core rev: 527d84799e4a7fb194002ea4db5deee9a6a3086c)

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>

libevdev: add new recipe (needed by xf86-input-evdev)

(From OE-Core rev: ad8e7f6d379a99649a1431a4bae7ad0bb0a78d9d)

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>

libxi: upgrade to 1.7.4

Also remove outdated (~1.2.99.5) git recipe that was never used.

(From OE-Core rev: de36faa2a6a7c7a379807d69fa07a04a4c878ff8)

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>

curl: Upgrade to 7.37.1

(From OE-Core rev: 2cd9141c8a8b9639c95cb68496f0392ba26595dc)

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

cups: Upgrade to 1.7.4

Remove patch that was backported

(From OE-Core rev: 984c8d621a2c2315a6c18b0f0b2fc0b380e7ca4d)

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

binutils-cross-canadian: Explicitly DEPEND on nativesdk-flex, we require it anyway

(From OE-Core rev: 1ed42b9248ba1ba393c812e4c485d25db464e683)

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

qemu: Use PACKAGECONFIG for libusb to avoid floating dependency

(From OE-Core rev: 24cc1af031244ad7c152b98312012b6344e57d48)

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

gcc-multilib: Simply/fix MULTILIB_OPTIONS handling

MULTILIB_OPTIONS takes the parameters which trigger a given multilib to be
selected. It supports *one* option per multilib, '/' separated. Spaces
separate options used to generate additional multilib combinations.

Adding in all of CFLAGS to this is therefore clearly a really bad idea
but how do we fix things?

The best option I've come up with so far is a list of whitelist variables
to use to trigger the multilibs. Its populated with the standard multilibs
we support, anyone setting up an advanced multilib can populate the variable
with the correct trigger parameters.

This has the advantage of simplifying the code and allowing us to remove
the code filtering blocks since there is no longer option duplication. Testing
after this change shows a much improved sdk toolchain functionality.

(From OE-Core rev: 29202cd1b9d2e5d56e5b9f7a596e44e229c90492)

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

lib/oe/classextend: Avoid early expansion of PR values

Variables like RDEPENDS can contain EXTENDPKGV which in turn uses AUTOPR
based values. This gets set during do_package execution so we want to
defer expansion until then. The only way we can do this in the RDEPENDS
(and friends) mapping code is to subsitute a dummy value, then change it
back again. Horrible but I can't see any other way.

This resolves multilib build failures with inconsistent PR values.

(From OE-Core rev: 5aea553e6eaa3b9647f26944976d2a9da79cba42)

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

cross-canadian: Copy target_ definitions from cross.bbclass

A while back we fixed the cross definitions to work better in multilib
configurations, apply the same fixes to cross-candian.bbclass

(From OE-Core rev: 4544b7f1d0abd1b1efd74da430f1ddedf3fdbd1d)

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

populate_sdk_base: Extend TOOLCHAIN_TARGET_TASK to include multilib variants

Most people expect the toolchain from a multilib build to contain multilib
components. This change makes that happen and is easy for users to override
should they want something different.

(From OE-Core rev: 396371588c7fd2d691ca9c39cd02287e43cb665b)

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

libomxil-0.9.3: Remove versioning for bellagio .so files.

The so files installed under ${libdir}/bellagio are not versioned and should
be installed without version-based symlinks so that omxregister-bellagio
can properly find and register them.

(From OE-Core rev: 3846b1ec782ba0cc64d389d03116743208e2383f)

Signed-off-by: Drew Moseley <drew_moseley@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

Revert "libomxil-0.9.3: Remove versioning for .so files."

The previous version of this fix was too aggressive and removed
versioning from too many of the .so files in the libomxil package.

This reverts commit 0ef3734c2f279bf463ba4d1aef5241cd4882d483.

(From OE-Core rev: b76d254db92d9c08e8d5d41becb2e60178cebb33)

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

allarch: Generate same package for MIPS and non-MIPS targets

LINKER_HASH_STYLE differs between MIPS and non-MIPS targets. This means
that LDFLAGS differs too. LDFLAGS is exported so it influences all task
hashes. Unfortunately this means that packages with architecture "all"
differ depending on whether they are built for a MIPS or non-MIPS target.
This causes a lot of unnecessary churn in the ipk/all directory when
switching build targets.

The simplest way to fix this is to ensure that LDFLAGS stays the same for
architecture "all" packages by clearing it. It shouldn't being used by such
packages anyway.

(From OE-Core rev: 14f4d016fef9d660da1e7e91aec4a0e807de59ab)

Signed-off-by: Mike Crowe <mac@mcrowe.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libice: fix non-deterministic libbsd dependency

libice 1.0.9 added automatic detection of arc4random(), which is in libbsd on
Linux.  As this is automatic and leads to failing builds when ssstate is reused,
seed the autoconf cache as relevant to implement a PACKAGECONFIG for the
functionality.

Default to not using arc4random() as the fallback has been in use for many
years, but people interested in security may wish to turn this on to increase
the security of the X authentication cookies.

(From OE-Core rev: 71d5eb34678eafbcb0f8ff4c31d52a276949bb0c)

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

kern-tools: adjust to full history meta-data

In order to generate and support kernel trees with full history, we need
to modify the kernel tools

 e914d570232a kgit-checkpoint: ensure that full meta-data artifacts are maintained
 192be836d318 kgit-scc: allow meta-data history to be maintained

(From OE-Core rev: f2015ead17c875ae37a9ad496fdafef2b931f771)

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

linux-yocto/3.14: libata and generic CPU modalias handling

Updating the 3.14 yocto kernel to incorporate the following fix
and feature of interest.

   5724bf17acbf x86: align x86 arch with generic CPU modalias handling
   6b9a52451a78 cpu: add generic support for CPU feature based module
  38367de316bb libata: support the ata host which implements a queue depth less than 32

[YOCTO: #6489]

(From OE-Core rev: 1fcb9843d663a5746690c81fe1cd6cf84dac04b9)

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

kernel: don't copy .so.dbg files into kernel source install

In 3.16+ x86-64 kernel builds produce a vdso64.so.dbg file. If this file is
copied into the kernel source install multiple QA failures are triggered.
Specifically, this file triggers a debug package split that results in
files installed but not shipped, and invalid .debug file errors.

By ensuring that .so files are not copied, we avoid this incorrect split
with no impact on future build phases.

(From OE-Core rev: f1f6d153de9dc675b4bf4020799b2365b1b62ec7)

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

linux-yocto-dev: bump to v3.16+

(From OE-Core rev: f90af1a78c2b616469ab49e3d426878602ad3e7b)

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

linux-yocto: x86_64: expand kernel stack to 16K

Updating to backport the following mainline commit:

[
    x86_64: expand kernel stack to 16K

    commit 6538b8ea886e472f4431db8ca1d60478f838d14b upstream

    While I play inhouse patches with much memory pressure on qemu-kvm,
    3.14 kernel was randomly crashed. The reason was kernel stack overflow.

    When I investigated the problem, the callstack was a little bit deeper
    by involve with reclaim functions but not direct reclaim path.

   ....
]

(From OE-Core rev: b162bcf26a5b97fddd2a64789f97ec705bb354f9)

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

lttng-modules: re-enable ARM builds

With lttng 2.4.2 and gcc 4.9, we can now enable lttng-modules for ARM.

(From OE-Core rev: f6587be6cd3dc864143b1c0be0bb8179a61dc835)

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

lttng-modules: update to 2.5.0

During the uprev of the yocto kernel to 3.16, lttng-modules failed to build.
To grab the latest stable content, we update to 2.5.0, and add two patches
to also make it build against 3.16+.

We also drop the older 2.3.3 lttng-modules, since it is no longer required
to support ARM builds.

(From OE-Core rev: 10680c8f06c52b25ed63d315596707701402646d)

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

linux-yocto/3.14: vexpress and MVM firmware support

Updating the 3.14 SRCREVs to integrate the following changes:

 meta: iwlwifi: Add MVM firmware support
 vexpress: Pass LOADADDR to Makefile

(From OE-Core rev: da1eca40b5ff608df6f4041a92b2d4c71ee6784e)

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

insane: Add build depends check

Now that we can get the task dependency tree from bitbake, we can start
to use this to strengthen our QA checks. If a dependency is added on
something which isn't in our dependency tree, that is obviously a bad
thing for example.

This patch therefore checks the RDEPENDS against the list of tasks and
ensures we do have a dependency present, if not a QA warning or error
can be issued through the usual mechanism.

The implementation is complicated by needing to resolve the RDEPENDS to
a PN using pkgdata. Its possible that can be an RPROVIDES of another
package so we need to check that too if it isn't a direct RDEPENDS.

To allow this test to work, we need to extend the do_package_qa
dependencies to include all RDEPENDS. In practise the do_package_write_*
tasks already do this so there should be no new circular dependencies or
any issues like that.

For now the issues are warnings as there are issues this finds in
OE-Core which need to be resolved and certainly will be in other layers
too. This change should simplify and assist some of Martin's dependency
scripts, the idea for this came from a discussion with Martin. It has
changed in that it doesn't just cover shlibs dependencies but checks all
dependencies.

(From OE-Core rev: f6cb24cf2255297308ef57399a6be407129d9b8f)

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

bitbake: bitbake: remove choices for dump-signatures

The SIGNATURE_HANDLER can be defined by the user, so we can't use choices.

(Bitbake rev: 0036b36e333a68c6d30a6bf5c21bfd4fd3e8441b)

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

package_ipk.bbclass: Support hierarchical feed

This patch allows for an optional new layout for ipk feed directories which I've
called a 'hierarchical feed' and is based on how Debian pools package files. It
is disabled by default and is enabled by setting IPK_HIERARCHICAL_FEED to "1".

In the traditional feed layout, package files are placed in <outdir>/<arch>/.
This can lead to several thousand files existing in a single directory which is
often a problem if developers want to upload a package feed to a shared web
hosting provider. For example, in my case, listing files via FTP only shows the
first 2000 files, breaking my scripts which attempt to upload only new and
changed files via FTP.

In the hierarchical feed, package files are written to
<outdir>/<arch>/<pkg_prefix>/<pkg_subdir>, where pkg_prefix is the first letter
of the package file name for non-lib packages or "lib" plus the 4th letter of
the package file name for lib packages (eg, 'l' for less, 'libc' for libc6).
pkg_subdir is the root of the package file name, discarding the version and
architecture parts and the common suffixes '-dbg', '-dev', '-doc', '-staticdev',
'-locale' and '-locale-*' which are listed in meta/conf/bitbake.conf.

This change relies on recent patches to opkg-utils which support hierarchical
package feeds.

Signed-off-by: Paul Barker <paul@paulbarker.me.uk>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake.conf: move BB_NUMBER_THREADS and PARALLEL_MAKE to bitbake.conf

Currently, BB_NUMBER_THREADS and PARALLEL_MAKE default to unset and
are set in local.conf. Now that we have the automatic probing,
the default values can be set in bitbake.conf and an example of
explicitly defining how many tasks to run can be moved to
local.conf.sample.extended.

[YOCTO #6217]

Signed-off-by: Roxana Ciobanu <roxana.ciobanu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gst-ffmpeg: Add CVE patches

Security Advisory - ffmpeg - CVE-2013-0866

The aac_decode_init function in libavcodec/aacdec.c in FFmpeg before
1.0.4 and 1.1.x before 1.1.2 allows remote attackers to have an
unspecified impact via a large number of channels in an AAC file, which
triggers an out-of-bounds array access.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0866

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-0875

The ff_add_png_paeth_prediction function in libavcodec/pngdec.c in
FFmpeg before 1.1.3 allows remote attackers to have an unspecified
impact via a crafted PNG image, related to an out-of-bounds array
access.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0875

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-0860

The ff_er_frame_end function in libavcodec/error_resilience.c in FFmpeg
before 1.0.4 and 1.1.x before 1.1.1 does not properly verify that a
frame is fully initialized, which allows remote attackers to trigger a
NULL pointer dereference via crafted picture data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0860

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2011-3934

Double free vulnerability in the vp3_update_thread_context function in
libavcodec/vp3.c in FFmpeg before 0.10 allows remote attackers to have
an unspecified impact via crafted vp3 data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-3934

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2011-3946

The ff_h264_decode_sei function in libavcodec/h264_sei.c in FFmpeg
before 0.10 allows remote attackers to have an unspecified impact via
crafted Supplemental enhancement information (SEI) data, which triggers
an infinite loop.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-3946

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-7023

The ff_combine_frame function in libavcodec/parser.c in FFmpeg before
2.1 does not properly handle certain memory-allocation errors, which
allows remote attackers to cause a denial of service (out-of-bounds
array access) or possibly have unspecified other impact via crafted
data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-7023

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-7009

The rpza_decode_stream function in libavcodec/rpza.c in FFmpeg before
2.1 does not properly maintain a pointer to pixel data, which allows
remote attackers to cause a denial of service (out-of-bounds array
access) or possibly have unspecified other impact via crafted Apple RPZA
data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-7009

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-0855

Integer overflow in the alac_decode_close function in libavcodec/alac.c
in FFmpeg before 1.1 allows remote attackers to have an unspecified
impact via a large number of samples per frame in Apple Lossless Audio
Codec (ALAC) data, which triggers an out-of-bounds array access.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0855

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2011-4351

Buffer overflow in FFmpeg before 0.5.6, 0.6.x before 0.6.4, 0.7.x before
0.7.8, and 0.8.x before 0.8.8 allows remote attackers to execute
arbitrary code via unspecified vectors.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-4351

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-0848

The decode_init function in libavcodec/huffyuv.c in FFmpeg before 1.1
allows remote attackers to have an unspecified impact via a crafted
width in huffyuv data with the predictor set to median and the
colorspace set to YUV422P, which triggers an out-of-bounds array access.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0848

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2011-3944

The smacker_decode_header_tree function in libavcodec/smacker.c in
FFmpeg before 0.10 allows remote attackers to have an unspecified impact
via crafted Smacker data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-3944

           file://0001-huffyuvdec-check-width-more-completely-avoid-out-of-.patch \

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-7010

Multiple integer signedness errors in libavcodec/dsputil.c in FFmpeg
before 2.1 allow remote attackers to cause a denial of service
(out-of-bounds array access) or possibly have unspecified other impact
via crafted data.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-7010

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2011-3941

The decode_mb function in libavcodec/error_resilience.c in FFmpeg before
0.10 allows remote attackers to have an unspecified impact via vectors
related to an uninitialized block index, which triggers an out-of-bound
write.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-3941

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2013-0846

Array index error in the qdm2_decode_super_block function in
libavcodec/qdm2.c in FFmpeg before 1.1 allows remote attackers to have
an unspecified impact via crafted QDM2 data, which triggers an
out-of-bounds array access.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2013-0846

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2012-6618

The av_probe_input_buffer function in libavformat/utils.c in FFmpeg
before 1.0.2, when running with certain -probesize values, allows remote
attackers to cause a denial of service (crash) via a crafted MP3 file,
possibly related to frame size or lack of sufficient frames to estimate
rate.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2012-6618

gst-ffmpeg: Security Advisory - ffmpeg - CVE-2012-6617

The prepare_sdp_description function in ffserver.c in FFmpeg before
1.0.2 allows remote attackers to cause a denial of service (crash) via
vectors related to the rtp format.

http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2012-6617

(From OE-Core rev: 58f08a96764094189b5aaf3cc8b4cc0c95e23409)

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

serf: specify LIBDIR to avoid install issue

LIBDIR defaults to ${prefix}/lib, use ${libdir} to avoid issue on
64bit platforms.

(From OE-Core rev: e5c023abd5265293623ac993a048d7abb9b71a59)

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

perf: remove CFLAGS override

definition of __SANE_USERSPACE_TYPES__ was added in kernel source by
commit e3541ec75219819d3235f80125a1a75d798ff6e1
perf tools, powerpc: Fix compile warnings in tests/attr.c

Overriding the CFLAGS makes it impossible for kernel build system to
append to it, thus making the build fail in various ways as:

|     CC /.../perf/1.0-r9/perf-1.0/perf.o
| In file included from builtin.h:4:0,
|                  from perf.c:9:
| util/util.h:74:24: fatal error: lk/debugfs.h: No such file or directory
|  #include <lk/debugfs.h>
|                         ^
| compilation terminated.

(From OE-Core rev: 7e7063e495d6451b46e06d08d16f92d0597a7d11)

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

net-tools: ifconfig interface:0 del <IP> will remove the aliased IP on IA64

(From OE-Core rev: d96b5072d53f134c5038601e30368db6ccfefe78)

Signed-off-by: Yong Zhang <yong.zhang@windriver.com>
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

toolchain-script: Really fix CANADIANEXTRAOE issues

There was an error in the previous fix for the powerpc toolchain issue, this
should correct it (and simplify the code too).

[YOCTO #6490]

(From OE-Core rev: 54ddf3fe060715534cd9e23d6e89cf6b06ed3d9f)

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

sanity: Check for setgid/setuid TMPDIR

Building in a TMPDIR which has setgid or setuid is a bad idea. We could try and reset
the permissions but since these can also invade into other directories like the cache
or sstate, lets tell the user to fix it instead.

[YOCTO #6519]

(From OE-Core rev: 8e44fc36018fda9b1f9ca8aebde3e744afc07eaa)

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

initscripts: fix bashism in bootmisc.sh

In the commit 'initscripts: save /etc/timestamp with seconds accuracy',
a bashism was introduced in the bootmisc.sh script in the code to set
the current date from the stored value in /etc/timestamp.  This causes
that operation to fail with the following message when /bin/sh is not
bash:

/etc/init.d/rc: /etc/rcS.d/S55bootmisc.sh: line 73: syntax error: bad substitution

Fixed by using pattern matching removal rather than bash-specific
substring expansion.

[YOCTO #6566]

(From OE-Core rev: c8a451adc71e4ce9ab963f61d7830c2d75aaffca)

Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

squashfs-tools: Add nativesdk in BBCLASSEXTEND

Necessary for including it in meta-toolchain.

(From OE-Core rev: f4d288881d7b1b2390a0f1234724fdd5d7a84b57)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

sysvinit: bootlogd: ensure /var/log/boot is created in volatiles

bootlogd does not write to /var/log/boot if it does not exist,
so if using the volatiles facility (presumed to mount /var/log under a
tmpfs), ensure that /var/log/boot gets created.

(From OE-Core rev: 283d109098358c06ff10daaf612e5a244691d7ac)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

sysvinit: bootlogd: Don't run savelog if it's not installed

bootlogd's default log rotation code on stop requires `savelog`, which
is in debianutils, which may not be installed.  If it's not installed,
don't try to perform the log rotation.

That is: in the affected code block, `savelog` is what is responsible
for creating "boot.0".  When `savelog` doesn't exist, an error message
gets printed on bootup to the effect of "mv: can't find boot.0".

(From OE-Core rev: 5c22973e1bf76615bcf57984ac7a30cf7d0766df)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

sysvinit: bootlogd: Honor VERBOSE

The messages echoed when starting and stopping bootlogd are currently
printed regardless of the setting of VERBOSE.  Adjust the initscript so
they're only printed when VERBOSE is enabled.

(From OE-Core rev: 42f107513148aa6e9fd69d51e63a183c613114c0)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

wic: Remove fstype from mkefidisk canned wks

The bootimg-efi plugin specifies the fstype internally, so remove it
from the partition definition.

(From OE-Core rev: e9365e9d9e8a8696849fda182ec260398059089a)

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

wic: Add help entry for wic kickstart

Previous patches added help for 'wic kickstart' but forgot to add an
entry in the main help page showing its availability.  Add an entry
for it.

(From OE-Core rev: 5aa212f990c3d9b90a4f06cea346b82fc5d376ba)

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

wic: Fix bad directory name in bootimg-efi

The original move of the mkefidisk code to the bootimg-efi plugin
resulted in a bad hdddir - fix it.

(From OE-Core rev: ab337705521d8734f12be612709e70a163111fe3)

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

pango: upgrade to 1.36.5

Update pango-ptest to reflect moved test suite.

(From OE-Core rev: 360b4e750d4eb7d4cbf832a61d8b1a20280b1bed)

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

libxft: upgrade to 2.3.2

Remove patch that has been merged upstream.

(From OE-Core rev: c2a4fa3669ca8607845347c0c196fa96478156cb)

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

xcb-util-wm: upgrade to 0.4.1

Licence checksums changes as the files previously checksummed were generated and
subsequently removed from the tarball.  Change the checksums to use the source
files instead.

(From OE-Core rev: 9204ee439720528d1c013bc8bb9560fbbcf174e3)

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

xf86-video-vmware: update to 13.0.2

Rebase the patch to add an option to disable vmwgfx, and add a patch from
upstream git to fix builds without xatracker.

(From OE-Core rev: cd7096c614a1f39043d02ba7c3f62599bf89f19e)

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

gdk-pixbuf: upgrade to 2.30.8

(From OE-Core rev: 5de0d039f89efb0140cedcc0e7e9dd90e64eeed7)

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

libice: upgrade to 1.0.9

(From OE-Core rev: 54297905f71a70359713d5a99a1440b4578613d5)

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

gnome-common: upgrade to 3.12.0

(From OE-Core rev: f89ad2e4bfb32848a938f7a816c049c3e35dad2c)

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

at-spi2-atk: upgrade to 2.12.1

(From OE-Core rev: e5a07935565392aa202cafac029ae5756fa5f22b)

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

inputproto: upgrade to 2.3.1

(From OE-Core rev: 74c098c21f7adae4ddf109eed3d85828805e5790)

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

libxfont: upgrade to 1.5.0

(From OE-Core rev: f90101ea7963663cd1c3b4eb9ef6f96840e49ad6)

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

fontsproto: upgrade to 2.1.3

(From OE-Core rev: a30decc6d3ac2250c1af19054230ed02fc40d5e3)

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

kernel-module-split.bbclass: Allow autoloading multiple modules or modules where basename != module name

* new KERNEL_MODULE_AUTOLOAD syntax doesn't support modules where basename and
  module name don't match (usually - and _), e.g.:

  module_autoload_bq27x00_battery = "bq27x00-battery"

* sometimes it's useful to load modules in particular order and
  module_autoload allowed to just list multiple modules, e.g.:

  module_autoload_snd-soc-neo1973-wm8753 = "snd-soc-s3c24xx snd_soc_s3c24xx_i2s snd-soc-dfbmcs320 snd-soc-wm8753 snd-soc-neo1973-wm8753"
  or
  module_autoload_g_ether = "s3c2410_udc g_ether"

  restore this possibility which is useful for incorrect dependencies
  between modules

(From OE-Core rev: e9cd8ba3dda624615b68c601eac04427d9483f14)

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

gstreamer1.0-plugins-base: get caps from src pad when query caps in gsttagdemux

(From OE-Core rev: b9a2aacfc0343d522cce9183dae15f1ae042b36d)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gstreamer1.0-plugins-base: do not change eos event to gap event if no data has passed to streamsynchronizer.

(From OE-Core rev: 8b85075e592e867e7ffba919b10fca150f6a62b0)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gstreamer1.0: pass rate of input segment to output segment in gstbaseparse.

(From OE-Core rev: e25c5b228178f13b2f9e25b5bd423d5ef7b40765)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gstreamer1.0-libav: avoid using non-growable pool for videodec

Using growable pool with at least 32 buffers for decoders.

(From OE-Core rev: 81ae7794ddbc7e2d97118092e0613249793214ef)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gstreamer: fix a bug for gstbaseparse

self-comparison will always evaluates to true.

(From OE-Core rev: ecd6764843f1bb3a8eeb8580db9e22231ec3c94d)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gst-plugins-good: fix integer overrun for mulawdec

There might be more than 65535 samples in a chunk of data, so use
32 bit integer instead of 16 bit.

(From OE-Core rev: 60de6e5d9d3dec482256ea4db8837204ca3ba628)

Signed-off-by: Wang Zidan <b50113@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

adt-installer: fix sed input file error

When use default install directory, we can't get the environment setup
script path. The reason is that opkg-cl list incorrect files paths.
This patch sets env_script variable to make us get correct environment
setup script path.

[YOCTO #6443]

(From OE-Core rev: e0080f279d5ebb320c2ba285765048fcca523fe7)

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

systemd: change dependency relationships

When building out a systemd based image, the udev-hwdb postinstall
script always fails at rootfs time. This is because that the 'udevadm'
command used in udev-hwdb postinstall script is now in the udev
package and udev recommends udev-hwdb.

Instead of letting udev recommends udev-hwdb, we let systemd do it
and make udev-hwdb rdepend on udev.

(From OE-Core rev: f14ee284da19665f157453021e3fa0e3a0dbf75d)

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

shadow: fix the behavior of su

In systems where bash is not installed and /bin/sh is provided by
busybox. Commands like `su -l -c '/home/root/test' xuser' would fail
complaining the the 'su' applet could not be found.

This patch references the old version of shadow to keep the behaviour
the way it was in old version so that we would avoid the problem mentioned
above.

(From OE-Core rev: ab0115d1b8a0cb0b25bdb14fd2a3e6c6bb9a44f8)

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

mkefidisk.sh: Add signal handling

Add basic signal handling to unmount and remove any temporary files.
Correct a quoting issue with the die() function caught testing signal
handling.
Fix a minor typo in "formatting" output.

(From OE-Core rev: a085cdf0e2dcf6543f61a8bdf68f83bcf8006373)

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

mkefidisk.sh: Fix redirection to 1

The current script intends to redirect stderr to stdout, but instead
redirects to a file named 1. No doubt a regex replace error.

Replace all instances of 2>1 with 2>&1.

(From OE-Core rev: 1864ca9751c28cca248cfba77a3d23fc58ff43bb)

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

busybox: mdev: Ensure /dev/initctl exists after tmpfs mount

During boot, there is a brief window during which /dev/initctl is
missing, which breaks initscripts that would need to access it. This
occurs because /etc/init.d/mountall.sh (rcS.d/S02...) attempts to ensure
/dev/initctl is present, but /etc/init.d/mdev (rcS.d/S06...) mounts over
/dev and clobbers the work done by mountall, and then does not wait
synchronously until initctl is ready before continuing.

To close this window, in /etc/init.d/mdev, we check whether /dev/initctl
is present, and if not, we remove it and recreate it.  This is the same
thing that is done by /etc/init.d/mountall.sh, and we have verified that
any writers of /dev/initctl will wait synchronously until sysvinit
notices the change in fd and does the read, so no race exists.

(From OE-Core rev: 53543363a8ab4424c17ed7aec0e8aefc4df86b3d)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

busybox: mdev: Use mdev.seq

/dev/mdev.seq exists to synchronize concurrently running instances of
mdev and to ensure that they execute in the proper order.  Without this
synchronization, it is possible to have inconsistent mount points, to
leak device nodes, or to have a node erroneously removed in rapid
hotplug scenarios.

Enable the use of mdev.seq by creating an empty /dev/mdev.seq at boot.

(From OE-Core rev: 3b256653ae9ab2cf56b21cb82a8bd2a562397b27)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

busybox: mdev: Make /dev/shm world-writable

Otherwise, users cannot use named semaphores or shared memory.

(From OE-Core rev: a8f35be467836d8fa0bc8f8e85eba6f89a13f087)

Signed-off-by: Richard Tollerton <rich.tollerton@ni.com>
Signed-off-by: Ben Shelton <ben.shelton@ni.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

mkefidisk.sh: Remove initrd entry for gummiboot

My previous patch adding gummiboot support was missing the line to
remove the initrd line from the boot config. This was an oversight in
copying over the grub setup to gummiboot. Add the necessary logic to
remove it.

(From OE-Core rev: c7355550dc21a1ef3c2e828ed5f51e94e12fac5f)

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

connman: Fix PPTP VPN plugin QA issue

This error was seen when building ConnMan. Fixed by not
building ConnMan pptp plugin with module versioning info.

ERROR: QA Issue: non -dev/-dbg/-nativesdk package contains symlink .so: connman-plugin-vpn-pptp path '/work/core2-32-oe-linux/connman/1.24+gitAUTOINC+42779cd63c-r20/packages-split/connman-plugin-vpn-pptp/usr/lib/connman/scripts/libppp-plugin.so' [dev-so]
ERROR: QA run found fatal errors. Please consider fixing them.
ERROR: Function failed: do_package_qa

(From OE-Core rev: d2328d6aa52fd4f0b47ae3292ae6ae07301bf9d0)

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

image.bbclass/rootfs.py: add variables to rootfs[vardeps]

Added base variables and package backend specific variables to
rootfs[vardeps] in order for rootfs to rebuild when changes are made.

Set some variables as [func] to inform bitbake that they are shell
scripts, so that it invokes its shell dependency parsing. Without
marking them as functions, changes in the actual function body would
not trigger rootfs rebuilds.

[YOCTO #6502]

(From OE-Core rev: b8b6214b885a0757f0e628937f8fe21c92c45155)

Signed-off-by: Roxana Ciobanu <roxana.ciobanu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: toasterui: fixing event.data clash

This patch fixes a name collision on the event.data
in the MetadataEvent class. event.data is used in the
event system as a copy of "d" in certain situations,
and this collision triggered a bug leading to data loss.

[YOCTO #6332]

(Bitbake rev: 3f191b7cfe95aea4d4e96babf001d62d45dd3aaa)

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