bitbake: Revert "fetch2: Adapt encode/decode url to use URI class"

This reverts commit 21fe2683aefde10e847e66c11c26d4f4c1e07cfd
since bitbake-selftest doesn't pass when this is applied and
we're seeing multiple build failures from this change.

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

rootfs_ipkg: fix BAD_RECOMMENDATIONS handling

If multiple versions of the same package are in the package feed then the
generate status file would only contains a "deinstall" status for the last one,
which meant that BAD_RECOMMENDATIONS wouldn't actually work.

Use awk instead of grep and stop reading when we reach a newline, so we only
ever output a single stanza.

(From OE-Core rev: a5362de60c0051f16b88a40bd9cb41915bee0b0f)

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

toolchain-scripts.bbclass: add PYTHONHOME variable to environment-setup

When relocating the SDK, applications using python will search for
python modules in the default location and will fail to start.
The below errors are thrown by gdb, for example:

Could not find platform independent libraries <prefix>
Could not find platform dependent libraries <exec_prefix>
Consider setting $PYTHONHOME to <prefix>[:<exec_prefix>]
ImportError: No module named site

In order to overcome this, add the PYTHONHOME variable to the
environment-setup script for both standalone toolchain and
adt-installer. No need to do that for meta-ide-support environment
script since this toolchain does not get relocated.

[YOCTO #3839]

(From OE-Core rev: e7a21cd69a326ab7e2d0e410db28f24956f61208)

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

bitbake: utils: Improve remove funciton to handle whitespace

Improve the remove function to better handle cases where path contains
special shell chars like whitespaces, '$', ';' or'\'.

Thanks to Enrico Scholz for the fix.

(Bitbake rev: 617511c9f86cc4ef52457653c8adff582d94bce3)

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

bitbake: bitbake/fetch2: workaround urlparse in older python not support git

(Bitbake rev: 7e479dc6a574a8f3bd9f24d2ed1c3ceef91f3828)

Signed-off-by: Matthew McClintock <msm@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: wget: Improve mkdir handling

If there there isn't a subdirectory to the downloadfilename, this was
failing. This patch avoids that issue.

(Bitbake rev: 58bfd8d88495d4cae808e23b7af40e65ad05450f)

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

bitbake: cooker: parse using bb.compat.Pool

(Bitbake rev: 8af519a49a3374bd9004864ef31ca8aa328e9f34)

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

libsdl: Use 'virtual/nativesdk-libx11' instead of hardcoding nativesdk-libx11

(From OE-Core rev: bb7a913e4a0da450cc85edfe0d8f80bdf4065fad)

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

dbus: Drop x11 support from nativesdk

This avoid a libx11 dependency for nativesdk as it is not intended to
be run in the toolchain. We also remove the dbus-launch as without x11
support it might case trouble for user

(From OE-Core rev: 2dcb6bf9205a5b2144ccb7f2912bf3e13211ae88)

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

devshell.bbclass: Allow running devshell and fakeroot together

Currently its hard to run a devshell complete with fakeroot context. This
patch allows the fakeroot flag on the task to do this, as with any other
task. Since we may need to start X terminal applications, we need to
only start the fakeroot session on the final command, hence the hoops
this code jumps through.

As always with fakeroot, you can break out and run a command without
the fake permissions with syntax like "PSEUDO_UNLOAD=1 <command>"

[YOCTO #3374]

(From OE-Core rev: 0a2662a48eaf0487db043c348e2834bb9cdd0466)

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

populate_sdk_base.bbclass: fix SDKTARGETSYSROOT value

Currently, SDKTARGETSYSROOT points to PACKAGE_ARCH which, sometimes, can
be set to MACHINE_ARCH. When this happens, the default target sysroot
passed to the cross-canadian toolchain, which points to TUNE_PKGARCH,
will be different from the directory where the target sysroot has been
deployed.

In order to fix this, use REAL_MULTIMACH_TARGET_SYS variable instead of
MULTIMACH_TARGET_SYS.

[YOCTO #3784]

(From OE-Core rev: 41437aaac0cfc6f931d3b2974d380f20ec01f6e8)

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

sanity: check_path_length: replace tab with 8 spaces

(From OE-Core rev: de3f2367c46a213d07d834c7e874d2a4a0697026)

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>

pseudo.inc: pseudo 1.5 uprev, support extra config flags

The pseudo 1.5 update is a moderately experimental set of changes
which ought to improve performance. With these changes, pseudo
uses an in-memory sqlite database which is lushed on exit,
the protocol is changed to reduce waiting for server responses,
and pseudo can suppress any and all fsync/fdatasync type operations.

This last feature is optional, and not on by default, so we need
to pass in an extra configure argument, but that argument wouldn't
be known to an older configure, so... Enter PSEUDO_EXTRA_OPTS which
is passed to configure, and which pseudo_1.5.bb sets by default to
"--enable-force-async". (I haven't added it in pseudo_git.bb, but
maybe it should be changed; I'm not quite as sure there.)

The justification for these changes is that, for most of the real-world
build cases I deal with, they produce a 25% or more reduction in the
build time of a project. This increases when a system is heavily
loaded.

(From OE-Core rev: 79ddb0c33401da442dbaa8e0d73ebacf297d9185)

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

meta-toolchain-qt.inc: Use CXX when linking

We should use CXX compiler when linking objects as done in regular
builds inside of OE-Core. This fixes issues we've been seeing in field
where we needed to change it byhand so the toolchain could work.

(From OE-Core rev: 4ade9c60acea0ac92d0a08f3078361b6f54c85e2)

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

bitbake: fetch2: Add SFTP fetcher

This fetcher differs from the SSH fetcher in that it adheres more
strictly to the SECSH URI internet draft --- it uses the sftp://
instead of the ssh:// scheme, and it uses sftp instead of scp.

(Bitbake rev: d240baeb7a4107d2eba3f08c411c0f086674d8e2)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: Add editor modelines for bb.tests.*

(Bitbake rev: 230e00948ba093958e8e89e9ee380444b3dad307)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: Adapt encode/decode url to use URI class

(Bitbake rev: 21fe2683aefde10e847e66c11c26d4f4c1e07cfd)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: unittests for bb.fetch2.URI class

(Bitbake rev: a71e7d4137d2c5268b406f69992d09c00772a040)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: Add a class representing a generic URI

A class representing a generic URI, with methods for accessing the URI
components, and stringifies to the URI. This class should be a bit more
flexible than the existing {encode,decode}_url functions in that it
supports more components (e.g. port) and that it does not rely on a
specific order on the return values. This makes it easy to add new
properties without affecting the API.

(Bitbake rev: bd824da8a7eafe27310e410807319628378caeca)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: Remove unused code in wget fetcher

(Bitbake rev: 6d88fcee16ced3a8c1ab1daf8e88d36f70f13346)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

package_tar: Fix so it actually works

This tells us how long the code hasn't been used for :/

(From OE-Core rev: cd503c7f7fec6d177209832f73cec9c5d490be6b)

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

package_tar: Remove completely pointless code

Remove/simply overcomplicated code. No functionality change.

(From OE-Core rev: e3b21455e4ecc5b4961cf52e035a2d6d541faf69)

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

package_tar: bb.mkdirhier -> bb.utils.mkdirhier

Fix deprecated function reference.

(From OE-Core rev: 9e7a6492010aa2de65310d6fa13290068e7cadb1)

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

package_tar: Fix whitespace

(From OE-Core rev: fead6005b56f53535967476cc752691316e02183)

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

package_rpm/dev/ipk/tar: Drop unused functions

The package_xxx_install functions date from a different era and are not used by
anything. In the rpm case, they're simply unimplemented, in the tar case they're
using broken whitespace and deprecated functions. We might as well clean
out the old broken unused code.

(From OE-Core rev: 3684036213c9b1c27389260b7a1e3441c6bd659d)

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

package_rpm/deb/ipk: Error if we don't find packages when creating the package index

If for whatever reason the package directory is empty of packages, it
makes sense to error early rather than later in what become much
more obtuse errors. This adds in a sanity check to each of the packaging
backends. It also removes the duplicate createrepo call since the
core index creation function now uses this directly after the switch
to smart.

(From OE-Core rev: 721ef058b37604e100021ec7a90ad2f745d83916)

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

classes: Drop none package specific packaging variable accesses

(From OE-Core rev: 24cbe88429ba021d7c461e0271b67106d11d5eb2)

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

conf: Remove unused ROOT_FLASH_SIZE variable from the config

This variable is set but never used in OE-core and meta-oe. It
was historically used for the Opie collection but seems to be
unused now.

(From OE-Core rev: 323ef78e377525e2214f4700c30305c493137853)

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

perl: Add auto/XS/Typemap in perl-doc package

(From OE-Core rev: 8b8e471b4e6d90e36320fc608b6eec908fd43fe6)

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

bitbake: fetch2: Ensure directory for stampfile exists before trying to create it

(Bitbake rev: cf510e3da36e53f98ca86501747364bf4699ecc1)

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

bitbake: utils: Use rm -rf in remove()

Whilst shutils.rmtree() is pythonic, its also slow. Its faster to
use rm -rf which makes optimial use of the right syscalls.

(Bitbake rev: 96088ebdec08e49ba9e8dbcac437bfcdc21f5983)

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

dropbear: use pidfile for daemon start/stop/restart

Old init script killed all dropbear processes when doing stop/restart
including open SSH sessions which is very annoying.

(From OE-Core rev: 97aa5ac2df7593e343d82f5e64a422bb951eacf9)

Signed-off-by: Roman I Khimov <khimov@altell.ru>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

tzcode-native: Fix SRC_URI checksum typo

(From OE-Core rev: dac9290b20819bb7f74fa27b87b8077d8856c551)

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

systemd.bbclass: Dont use libdir and base_libdir for units

systemd always uses /lib and /usr/lib to store unit files
so using libdir and base_libdir is incorrect. It will work
where libdir is usr/lib and base_libdir is /lib but wont work
when say its /lib64

Add a check to make sure that SYSTEMD_PACKAGES are part of PACKAGES
too, otherwise error out

(From OE-Core rev: 22e16f44b58ae6bbf719b9074d39aac065a402f1)

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

gnupg: fix CVE-2012-6085

Code taken from Redhat

[YOCTO #3813]

(From OE-Core rev: 17f8e03ace51df6b94fdcb7e410acf79c7b046ce)

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

sanity.bbclass: when bblayers.conf is updated, it invokes a reparse

When bblayers.conf is updated(when sanity check is running), it should
tell to bitbake to reparse configuration files. I will send a patch to
bitbake-devel, with the actions needed in bitbake and hob.

[YOCTO #3213]
(From OE-Core rev: 5db1ff93f7204b43b7242fc7ef415216eb632ed8)

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

package_ipk: check CONFFILES exist before adding them to metadata

opkg-build verifies that conffiles exist, so verify that the specified files
actually exist before writing them to conffiles.

This mirrors the behaviour of FILES and package_rpm's CONFFILES handling.

(From OE-Core rev: fb87edc881009cf4d582cd95c248884551fe07fe)

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>

scripts/bitbake: Remove all instances of paths to a layer's scripts directory.

* Currently the assumption is made that only oe-core can include a scripts
  directory.

* However, when other layers create a scripts directory the bitbake script
  freaks out causing a infinite recursive loop until it crashes.

* Simply changing the regular expression to remove all instances of scripts path
  instead of just the first one fixes this problem.

 [Yocto Bug 3872]

(From OE-Core rev: 9b445cc39604223b0cfb21d28f748a86ff4cdf68)

Signed-off-by: Franklin S. Cooper Jr <fcooper@ti.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

alsa-lib: upgrade to 1.0.26

(From OE-Core rev: 759191ee0b11231e1719bd5c66ac0afbd02e2a80)

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

systemd: add udev-extraconf and USB/PCI IDs to RRECOMMENDS

The standalone udev has these dependencies and extraconf is required to have
working automount amongst other things.

(From OE-Core rev: 3d72ba9d25a02cac2c0a63ab32c5483be5c6ea3f)

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>

libpng: update to upstream version 1.5.14

obsolete_automake_macros.patch removed as it's not part of upstream.
The only changes to license texts are latest libpng version number
and release dates.

(From OE-Core rev: d68e9660327db2dd4f2ae7bcea9925146b94eb77)

Signed-off-by: Marko Lindqvist <cazfi74@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

harfbuzz: update to upstream version 0.9.12

Fixed license declaration to have AND between all the licenses
instead of OR.

(From OE-Core rev: 0a90fb2faebbb7848dbd5a23b37490866b397f5f)

Signed-off-by: Marko Lindqvist <cazfi74@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gtk+: update to upstream version 2.24.15

- configure-nm.patch updated to apply
- hardcoded_libtool.patch updated to apply
- obsolete_automake_macros.patch removes as it's now part of upstream

(From OE-Core rev: a4ed7da18d8622fc0b6e3d32ac8ce9456b912322)

Signed-off-by: Marko Lindqvist <cazfi74@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

webkit-gtk: update to upstream version 1.8.3

(From OE-Core rev: 570e5e20623031ad03cce470fc3d94d9a5b05827)

Signed-off-by: Marko Lindqvist <cazfi74@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

xev: upgrade to 1.2.1

Licence didn't change, the end line for computing the licence md5sum was
inside the actual code which, eventually changed. Adjust it to the right
line.

(From OE-Core rev: 96d9e102067471018acfa9651b810f0d33a363fa)

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

xf86-video-vmware: upgrade to 13.0.0

Removed backported patch.

(From OE-Core rev: f264bf699a42dfe2616386d03524642890308e80)

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

xf86-input-synaptics: upgrade to 1.6.3

(From OE-Core rev: a1eef756921d66138ef233c451a99b2ee1356cfb)

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

xkeyboard-config: upgrade to 2.8

(From OE-Core rev: 40083ca6decd05f480b0a8cd5923e68884c3b307)

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

libdrm: upgrade to 2.4.42

(From OE-Core rev: dd53452b1cf2ccec44f465ac420d134a17332c7b)

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

xf86-input-vmmouse: upgrade to 13.0.0

(From OE-Core rev: e8f5f341256430cf45ee186435bccf593793b661)

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

xf86-video-intel: upgrade to 2.21.0

(From OE-Core rev: 7bdc27511c9129e35caf1bdc0df37b382f1b5b98)

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

qemu: upgrade to 1.3.1

(From OE-Core rev: ca3ce0886df01936d5ad309de1d39520b5278a80)

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

bitbake: hob: remove save/load template functionality

The Hob templates functionality, in its current incarnation,
is confusing and has no clear utility.

[YOCTO #3696]
(Bitbake rev: 4eb3b6bb9f936808ddf085624078f6479c522c48)

Signed-off-by: Cristiana Voicu <cristiana.voicu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: knotty.py: fix unknown event bb.event.DiskFull

There is an error:

ERROR: No new tasks can be excuted since the disk space monitor action
is "STOPTASKS"!
ERROR: Unknown event: <bb.event.DiskFull object at 0x2ab6310>

This is because we don't handle the event bb.event.DiskFull in
knotty.py, we don't want the knotty to do anything here since we have
done everything in monitordisk.py, so just ignore this event would fix
the problem.

[YOCTO #3523]

(Bitbake rev: 0fa874dbe0532bbe88ade405f514a237e6cd2488)

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

bitbake: bitbake-layers: make show-cross-depends avoid long path

The "bitbake-layers show-cross-depends" print the absolute path in the
past, now it will print the relative path (relative to the layer dir) by
default, and add the "-f" option to make it print the absolute path.

For example:

$ bitbake-layers show-cross-depends
[snip]
meta-intel/meta-jasperforest/conf/machine/jasperforest.conf requires meta/conf/machine/include/ia32-base.inc
[snip]

$ bitbake-layers show-cross-depends -f
[snip]
/path/to/poky/meta-intel/meta-jasperforest/conf/machine/jasperforest.conf requires /path/to/poky/meta/conf/machine/include/ia32-base.inc
[snip]

[YOCTO #3824]

(Bitbake rev: df217701318d60559be0b10214b883b8ce4f5d2a)

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

bitbake: hob: remove parsing warnings when machine is changed

When the machine is changed, the warnings list is emptied and the warnings
bar is hidden.

[YOCTO #3830]
(Bitbake rev: 8e43da7798a98988f7da3f7c424e8f9e5eac2010)

Signed-off-by: Cristiana Voicu <cristiana.voicu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: hob: no sanity checks re-run

Hob was hanging when the user tried a rebuild a new image, because
of sanity checks. It should not re-run sanity checks.

[YOCTO #3829]
(Bitbake rev: f7ff550f53cf3e5b166940db83f27ec733f3ef72)

Signed-off-by: Cristiana Voicu <cristiana.voicu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: monitordisk.py: disable the inode checking for some fs

There is an error when use disk monitor on btrfs:
WARNING: The free inode of rootfs is running low (0.000K left)
ERROR: Immediately abort since the disk space monitor action is "ABORT"!

This is beucase some fs formats' statvfs.f_files (inodes) is zero, thus
the statvfs.f_favail (free inodes) is zero, too, this a feature of the
fs, we disable the inode checking for such a fs.

[YOCTO #3609]

(Bitbake rev: b2ada3ad5f7aefff107f013e0c9decea481c7ef6)

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

bitbake: fetch2: Improve lock/done stamp file paths for local files

Currently stamps end up at the top level of DL_DIR even if the files
themselves are in a directory structure. This patch preserves path
components allowing the top level directory to be less populated
which is an advantage for large sstate caches.

(Bitbake rev: 59921ce3ed7a4c0b7f8ef1a101ad9127469bf1fd)

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

bitbake: bitbake & hob: reparse config files, when sanity check updates bblayers.conf

I've sent a patch to poky, with some changes in sanity.bbclass. When bblayers.conf
is updated, it updates also a variable to know when to reparse configuration files.
This patch contains the changes to reparse conf files, after a sanity check which
changes bblayers.conf.

[YOCTO #3213]
(Bitbake rev: 966e3e59ceb1c8b33b5881f291c148ea6513eda0)

Signed-off-by: Cristiana Voicu <cristiana.voicu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: fetch2: Ensure expansions happen in URL parameters in uri_replace

For example, this allows us to specify downloadfilename=PATH in sstate
mirror urls to improve directory structures in DL_DIR when using sstate.

(Bitbake rev: 7850a1364b6b37c58664d84f9c14806b4479b45c)

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

bitbake: ConfHandler: Use re.X to make long regexp more readable

The __config_regexp__ in ConfHandler is quite long, and using re.X to
break the expression onto several lines make it a bit easier to read.

(Bitbake rev: 54dce9e14ab0657d76f0d0ae22eef7fab8e8950d)

Signed-off-by: Olof Johansson <olof.johansson@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: tests/codeparser: Hack around circular inclusion problem

There is a circular dependency issue where bb.data can't directly depend on
bb.parse. The tests were failing due to this issue which is misleading. This
patch hacks around it for now. I'd rather that than not running tests at all.

(Bitbake rev: a206ae0d7769a41ff3666d0f53ff9cf422dfa518)

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

bitbake: wget: Only use -O option when not spidering

If we pass the -O option whilst spidering, empty files are created which is not
desired. We also need to ensure any subdirectories are created when using the
downloadfilename parameter.

(Bitbake rev: d5f78e98d5aba36c95288fbaac267c2d54537b02)

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

bitbake: codeparser: Track bb.utils.contains usage

The bb.utils.contains function usage is getting increasingly used in the metadata
but isn't handled automatically by the python dependency tracking code. This patch
changes that and also adds the "OE" names for the functions.

Whilst there are reasons this is a bad idea, its likely outweighed by the
shear number of these references and the current holes in dependency information
which we're now relying heavily upon.

(Bitbake rev: 0b9d117631ce909312d53b93289e61defc6be01c)

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

bitbake: codeparser: Track appendVar and prependVar calls as we do for getVar

We need to track appendVar and prependVar calls just as we do for getVar in order
to ensure we're not missing variable dependencies.

(Bitbake rev: 767b4751232f4ee3979deb4d3f733fcf9ee2bd44)

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

tzcode/tzdata: Update to 2012j

Readme update to indicate status of project and authorship

(From OE-Core rev: 6c6568e47d6b3b0edaba569bfa4e38f9d72b5122)

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

tzdata: Simplify code removing not used cases

We shouldn't have an use-case where we'd use 'FUBAR' timezone so
instead of adding postinst handling for this use case we handle it at
install time and keep the Universal as fallback if user did something
wrong.

This also ensure the /etc/localtime file is kept as a symbolic link.
This will make timezone not available when /usr is in separated
partition (and not mounted) however the applications ought to fallback
to GMT timezone in this case and when /usr is made availble timezone
will work fine.

Change-Id: I9a4f05db7a0bdc06511deb5693d1d16569d2fc63
(From OE-Core rev: 4b5e1757224085deac54b2583853fc779dc70614)

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

tzdata: We shouldn't override the localtime if it is valid

The code where mistakenly replacing the localtime file setting so we
end with a copy of file instead of a symbolic link. This fixes it so
now, we'll only do that in case the link is pointing to invalid data.

Change-Id: I16dfa5ea4f293c48bb396f4e23a2ea53e6c9e745
(From OE-Core rev: a9b42c8e85332a65788b1434f926186d4887b287)

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

Solved package namespace errors in PRS

For some packages PRS reported incorrect upstream version
as it was either the raw string or it mismatched some
alternative groups.

(From OE-Core rev: 6f3ace8ba75eed891aa4844a6830bedca3d8b70e)

Signed-off-by: Emilia Ciobanu <emilia.maria.silvia.ciobanu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

insane: remove unused variable

(From OE-Core rev: 1165a0b161ffc32258b0726e6ef3082d97dfb32a)

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

remake: do not create po files

The 'make update' was using wget to get the gmo and other gnu files from
upstream, since need to work cleanly in a non-networked or proxy environment
this does not so well.  Remove the list of languages from the LINGUAS file.

[YOCTO #3745]

(From OE-Core rev: 9987f210e3faf31bfeab35ae56606c8a577b3aa0)

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

mulitilib_header: Avoid sstate checksum issues for -native recipes

The use of arch specific variables like MIPSPKGSFX_ABI was creeping
into the -native sstate checksums of package like ncurses-native.

This is pointless and undesireable. We could add specific variable
exclusions but we might as well just brute force the code to be disabled
in the -native case since we don't use multilibs in the native case.

[YOCTO #3827]

(From OE-Core rev: cc7352ed990b1bdd26203a4d3b21286280ac26c5)

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

kconfig-frontends: Depend on pkgconfig-native

We were seeing errors like:

| autoreconf: running: /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly/build/build/tmp/sysroots/x86_64-linux/usr/bin/autoconf --force
| configure.ac:27: error: possibly undefined macro: AS_IF
|       If this token and others are legitimate, please use m4_pattern_allow.
|       See the Autoconf documentation.
| configure.ac:100: error: possibly undefined macro: AC_MSG_ERROR
| autoreconf: /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly/build/build/tmp/sysroots/x86_64-linux/usr/bin/autoconf failed with exit status: 1
| ERROR: autoreconf execution failed.

which turns out to mean the pkgconfig macros were unavailable (thanks for clear
error messages autoconf).

This patch adds in the missing dependency.

(From OE-Core rev: 26431ffda8886412147ff347c000a0ecc2671db5)

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

bitbake: Remove whitelisted vars from non-task deps

Though the value of variables in the BB_BASEHASH_WHITELIST is kept out of the
checksums, dependency on them is not, at least for variables and non-task
functions. In the code, the whitelist is removed from the overall task dep
list, but not the individual variable deps. The result of this is that
functions like sysroot_stage_all and oe_runmake end up with whitelisted
variables like TERM listed in their dependencies, which means that doing
a 'unset TERM' before building will result in all checksums for tasks that
depend on those changing, and shared state reuse not behaving correctly.

This is only really a potential issue for variables from the environment, as
it's the existance/removal of the variable that's an issue, not its value, and
the other whitelisted variables are set in our metadata. This which means in
practical terms the only cases where this is likely to be an issue are in
environments where one of the following are unset: TERM, LOGNAME, HOME, USER,
PWD, SHELL. This may seem like an unlikely circumstance, but is in fact a real
issue for those of us using autobuilders. Jenkins does not set TERM when
executing shell, which means shared state archives produced by your jenkins
server would not be fully reused by an actual user.

Fixed by removing the whitelisted elements from the individual variable deps,
not just the accumulated result.

(Bitbake rev: dac12560ac8431ee24609f8de25cb1645572d350)

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

bitbake: ConfHandler: Improve regexp to fix mis-parsing of += and no whitespace

If you have:

FOO = "a"
FOO += "b"
FOO+= "c"

The expected result is "a b c" however we were seeing "a b" with the FOO+
variable being assigned the value "c". This isn't the expected result.

We need to make the name part of the variale non-greedy so that any + character
becomes part of the operator. This patch does that. I compared the configuration
in OE-Core before and after the change and only the test case changed.

[YOCTO #3834]

(Bitbake rev: 2cd8d7fd12a646e6516e2c985e6a54121d19eb59)

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

util-linux: Remove -systemd package

unit files are added to util-linux itself when selected

(From OE-Core rev: dbabe19d4ee5cc291b5d996e9b828eed3e4484d8)

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

Python: Add missing dependency "textutils" to "io" package

Modify the include file and script to generate a missing RDEPENDS.

Install python on target with python-io. Import ssl:

Python 2.7.3 (default, Feb  9 2013, 16:04:35)
[GCC 4.7.2] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import ssl
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python2.7/ssl.py", line 58, in <module>
ImportError: No module named textwrap

Installing python-textutils solves the issue.

(From OE-Core rev: 900ae881c3483eea36aa0be456b93f92980f4924)

Signed-off-by: MiLo <milo-software@users.sourceforge.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

multilib: Fix an OVERRIDES expansion order issue

There were problems where a SRC_URI with:

SRC_URI_append_powerpc = " xxx"
SRC_URI_append_powerpc64 = " xxx2"

would end up with *both* xxx and xxx2 being added when using a multilib
which is clearly incorrect and undesirable.

The issue is that OVERRIDES has virtclass-multilib-xxxx added to it,
this eventually changed DEFAULTTUNE which then changes
TRANSLATED_TARGET_ARCH which is in OVERRIDES meaning we then need to
re-evaluate the overides and the TRANSLATED_TARGET_ARCH gets applied
twice since once you apply an override, it doesn't get undone.

Expanding DEFAULTTUNE to the correct value in advance avoids the issue
and means only the correct overrides get applied.

[YOCTO #3874]

(From OE-Core rev: 920c9024f5a47ad14670067f910450983bae2aa7)

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

pseudo_1.4.5.bb: Finish fixing linkat()

The 1.4.4 fix replaced possible double-prepending of chroot paths
with possible non-prepending of chroot paths. After significant
evaluation, have settled on a single prepending of the chroot
path as a workable compromise.

(From OE-Core rev: a79597994e3f680e34a1a45fb37d76977903ded5)

Signed-off-by: Peter Seebach <peter.seebach@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

busybox: add config fragments

Add config fragments to busybox.

The implementation makes use of merge_config.sh script in kern-tools-native.
The use case is similar to the yocto kernel's configuration fragments.

We also add kern-tools-native to busybox's DEPENDS variable to ensure
that merge_config.sh is available when required.

[YOCTO #3379]

(From OE-Core rev: bbcd8b344598850ea2c8d3ad375c519713581fde)

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

gtkhtml2: remove, nothing depends on it

(From OE-Core rev: 1c3ed8c109f46e7e38649d6914693ed4c18f45a5)

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

web: remove gtkhtml2 version

The gtkhtml2 version of Web is even older than the webkitgtk port, remove it.

(From OE-Core rev: 7ee3c35ca51e358e1d5a710922acb52c0724086e)

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

pseudo_git.bb: Bump to pseudo 1.4.4.

The pseudo 1.4.2 linkat() implementation had a broken edge case
in which you could end up with chroot paths being doubled when
using plain link() calls instead of linkat() calls.

(From OE-Core rev: c70443ef21713d805012ef839e3fac04de8eadd2)

Signed-off-by: Peter Seebach <peter.seebach@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libtool-native_2.4.2.bb: Always use /bin/sed for SED

If you never use sstate and always build everything from scratch you
will never see this problem.  However, if you use sstate and build
directories that last a long time eventually you can end up with the
scenario where libtool gets a hard coded path in it for sed, and sed
may not exist.  The reason you don't see this problem to often if you
generally build from scratch is that libtool builds before sed and
will pickup the host's /bin/sed.

The way to reproduce the issue is:

bitbake some_image
bitbake -c cleansstate libtool-native
bitbake sed-native
bitbake libtool-native
bitbake -c clean sed-native
bitbake ANY_PACKAGE_THAT_USES_LIBTOOL_NATIVE

In my case I used modphp, which doesn't exist in the oe-core. You will
end up with a strange looking error like:

| make[1]: *** [buckets/apr_buckets_alloc.lo] Error 1
| /opt/build/bitbake_build/tmp/sysroots/x86_64-linux/usr/bin/x86_64-linux-libtool: line 981: /opt/build/bitbake_build/tmp/sysroots/x86_64-linux//bin/sed: No such file or directory

The solution is to always use /bin/sed for libtool-native.

(From OE-Core rev: 605e4484840e70c64acddb4aa1a3c9fec4078d9d)

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

site/x32-linux: Specify double alignment

Double alignment is 8 bytes on x32 but it is defaulting to 4 currently.
This leads to various issues and fontconfig fails to build due to the
mismatch triggering assert failures.

(From OE-Core rev: f2a0784f368fa8a766aae4242a0c187759b35393)

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

scripts/qemuimage-testlib: Use wide option to ps calls

Forcing ps to display unlimited column width allows the qemu IP address to
be discovered during sanity testing when the command line is extremely long.

This seems to fix the sanity testing problem on AB05 which was recently updated
to OpenSUSE 12.2. I'm not sure what about qemu or process listing is different
on that distribution but this simpile fix seems to work and my help on other
distro's as well.

(From OE-Core rev: 4cea35cc4e4ed8e68cd117825b1dd4ef1be768c2)

Signed-off-by: Michael Halstead <michael@yoctoproject.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

gtk+: use gtk-immodules-cache class

In order to have the proper postinst/postrm scriptlets generated for
gtk+ immodules packages, use the already existing class.

[YOCTO #3853]

(From OE-Core rev: 1c5646dde09008662f064ce7e7400c4d68775278)

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

gtk-immodules-cache: add weak asignment for GTKIMMODULES_PACKAGES

This is needed if the GTKIMMODULES_PACKAGES is changed later, in
do_populate_packages for example. This way, we don't have to add another
dumb asignment in the recipe inheriting this.

[YOCTO #3853]

(From OE-Core rev: e9e80eac6ab4982cb42fa2c5403630926351efed)

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

gnome-keyring: compile schemas on host

gsettings.bbclass offers just that.

[YOCTO #3854]

(From OE-Core rev: 2d5a935b70218335af4859f409b81769c9238db3)

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

librsvg: use the new pixbufcache class

Also, fix the GDK_PIXBUF_QUERYLOADERS path.

[YOCTO #3582]

(From OE-Core rev: 801ca93cbd3e8fd8ce08f72432b6d4c80ed51261)

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

gdk-pixbuf: use the new pixbufcache class

[YOCTO #3582]

(From OE-Core rev: f6dce5376894d4ac91f01a85cffc828678695b04)

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

Add pixbufcache class

All packages exporting pixbuf loaders should inherit this class in order
to generate the correct postinst/postrm scriptlets.

[YOCTO #3852]

(From OE-Core rev: 61afa98f96f5c62473cb2db383b48d3d23c5d7ac)

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

fontcache.bbclass: use the postinst_intercept script

"Link" the package to the postinstall hook by running the
postinst_intercept script.

(From OE-Core rev: a14faa3b9c55574a096d517431393e4ac3c86823)

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

gtk-icon-cache.bbclass: use postinst_intercept script

Since the hook has been made a standalone script, use postinst_intercept
script in order to "link" the package to the hook.

(From OE-Core rev: d7ddae84165b5b84dc6ac640fd492ade891ddd4e)

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

rootfs_(ipk|deb|rpm).bbclass: check package installation status after ROOTFS_POSTPROCESS_COMMAND

Since the intercept fall-back procedure will change the package
installation status, do the checking after ROOTFS_POSTPROCESS_COMMAND
ends.

(From OE-Core rev: 414c3918baccc2166bba536c956eebaeb49aace3)

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

image.bbclass: add fall-back functionality when running intercepts

If an intercept script fails, it would be helpful to fall-back to
running the postinstall on target's first boot. In order to achieve
that, the postinstalls that install a host intercept hook will have to
return 1, so that the postinstall is marked as unpacked only. If the
intercept hook fails, then we're ok, the postinstalls will be run on
target anyway. If it succeeds, then mark the packages as installed.

This logic was chosen mainly because of rpm backend which saves the
failed postinstalls in /etc/rpm-postinsts. Hence, in order to mark the
packages as installed, all we have to do is delete the scriptlets from
there.

(From OE-Core rev: ed8ac4ee43132ae974794038821f7ca5465ae556)

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

Add separate directory for postinstall intercepts

The scripts/postinst-intercepts will contain all postinstall hooks that
we need to run after all packages have been installed.

If one wants to install such a postinst hook, all it needs to do is put
the hook in this directory and, from the package postinstall scriptlet,
call:
postinst_intercept <hook_name> <package_name> <var1=...> ...

This will, practically, add the package_name in the list of packages
that need the hook to run and, also, set any variables that would be
needed in the hook. For example, variables like ${libdir}, ${bindir},
etc. that might depend on distribution can be passed on to the script in
this way.

(From OE-Core rev: 0ef538d75c2f3921a2fcbe6ca1deed5525b276cc)

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

distro_check: Remove creation of empty Meego filelist.

This fixes permissions errors when running distrodata.

(From OE-Core rev: c58e6cf352774e147038e6543ac95ab0060f2327)

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

mesa: upgrade to 9.0.2

(From OE-Core rev: 7a0eca4379b72981a378ffe62f4e8e3428ee791a)

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