xserver-xorg: add runtime provides for the driver ABI version

The xserver driver ABIs can and do change in a way that is unrelated to the
version of xserver, so it's entirely possible to build an image that has a
mismatch between the server ABI version and the version that the drivers were
built against.  xserver detects this and refuses to load the modules.

By adding RPROVIDEs to the xserver package that describe the ABI versions it has
(such as xorg-abi-video-13, xorg-abi-input-11), drivers can RDEPEND on the
version that they were built against.  This means that when the ABIs change,
there will be package dependency errors at image time instead of images that
build fine but don't work.

(From OE-Core rev: 8ef5f205aec04140198d5ba0f5c405ae6e977dbe)

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

mtdev: New recipe for mtdev

(From OE-Core rev: 5ae7825ed7b8221d1c37e3d053404a3f2f7d27f2)

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

cpan-base: Add more debug paths to FILES

We seem to be mostly installing modules into vendor_perl nowadays.
Make sure that the .debug data from there is captured appropriately.

Also, expand ${PERLLIBDIRS} at the point of assignment so we don't
call the python again and again.

[RP: Fixup to whitespace]
(From OE-Core rev: ed7690bf5cc964b5cee55f5ef13c10c75d8e3463)

Signed-off-by: Phil Blundell <pb@pbcl.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

atk: Update to 2.6.0

Newer Clutter releases want 2.5.3+, so upgrade to the stable 2.6.0.

(From OE-Core rev: 01ecc9fbf59712d0f8e8a9b212171efc9d28ac57)

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

package: Hardlink debug source to improve performance

When copying the source files needed for the -dbg package, use hardlinks
where possible. This saves some disk space and hence helps performance
of the builds.

(From OE-Core rev: 6775feb9fe935ab01fd9cae2b2d3fce5824a9a72)

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

xorg: remove XF86 BigFont extension

This has been disabled by default upstream since 2007, nothing uses it.

(From OE-Core rev: 06d27cf0fbcc4004e6f456880eca49893c9290bf)

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

kernel.bbclass: Do not chdir to /boot before running update-alternatives

The symlink from uImage-3... to uImage is not created at image creation
time and not properly update on kernel upgrades. This is fixed by removing
the chdir. The other users of update-alternative do not change the directory
before calling it.

(From OE-Core rev: c77ca9ee901468c93570b5264b226f7d17a41c16)

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

eglibc-2.16: Use tar ball instead of svn SRC_URI

Adapt the recipes to fetch a tarball.
Tarball is generated from latest 2.16 branch
which has e500-math_private.patch already applied
hence we remove that patch.

(From OE-Core rev: 77ee4d7d88976c7bb2bb25b57e06b83edaacbd4c)

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

eglibc: Move perl- and bash-using scripts to separate recipes

This removes the dependency of eglibc.bb itself on perl and bash
which, in turn, eliminates the need to build those two recipes if the
scripts which need them are not going to be installed.

Also provide dummy do_evacuate_scripts() for all variants of eglibc-initial
otherwise the nativesdk and multilib variants might crash trying to
copy a non-existent mtrace script.

(From OE-Core rev: 74b5f8943b2b29c7b3b62be7d81fb2b3a86b9584)

Signed-off-by: Phil Blundell <pb@pbcl.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

package.bbclass: Exclude the PKGTRIPLETS variable

Without this, we'd rerun packaging for every machine since this variable contains
a machine specific component.

(From OE-Core rev: 61131828c59178c923b3d5b5fcacf0dbcba275a5)

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

multilib/clsextend: Improve handling of regexps in PACKAGES_DYNAMIC

Now that PACKAGES_DYNAMIC is more standardised, starting with ^ anchors,
the variable manipulations performed by clsextend for multilib don't work.

This patch at least improves it to hack around the problem and enable
mulitlib builds to work again. If this code doesn't do the right thing, the
recipe is free to override the variable with the correct multilib case.

(From OE-Core rev: 593faec6e0155bdd7a43ee84c24de8ee20287681)

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

yocto-bsp: set branches_base for list_property_values()

yocto_bsp_list_property_values() is missing the context it needs to
properly filter choicelists, so add it to the context object.

Fixes [YOCTO #3233]

(From meta-yocto rev: 064b15f76c5b52899f4c3fdef06412c3063062a5)

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

sstate: Improve handling of machine specific manifests

Now do_package isn't machine specific, we're only left with do_populate_sysroot as a
machine specific task. This change marks only the machine specific manifests as machine
specific, defaulting to PACKAGE_ARCH for everything else.

This means we do less work where there are multiple machines using the same
core package architecture and we can start to clean up the sstate duplicate files
whitelist.

(From OE-Core rev: febeaf3d1b8917b660c7279b008d8b03337568e9)

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

package.bbclass: Switch shlibs to pkgdata directory and make package non-machine specific

Currently, do_package is machine specific since the shlibs data is installed
into each machine specific sysroot. This change moves the shlibs data to the
pkgdata structure, at the expense of having to iterate over a set of shlibs
directories instead of a single one.

It turns out this isn't any particular hardship for the code and as a result,
do_package stops being machine specific leading to optimisations for builds
that use a common PACKAGE_ARCH.

(From OE-Core rev: cc088489d70fb27d460c3dbe35d6ea382123c134)

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

lib/oe/packagedata: Use the PKGMLTRIPLETS variable

(From OE-Core rev: 26e16a5e5ee1059fb8e55ab915ae9cb8e2b54dcd)

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

base.bbclass: Add PKGTRIPLETS and PKGMLTRIPLETS variables

These variables correspond to the PACKAGE_ARCH list combined with the TARGET_VENDOR
and TARGET_OS values. These can be used to traverse the pkgdata structure.

Setting these once in base.bbclass stops pkgdata needing to recalculate the values
and is also useful for the reworked shlibs code in a patch that will follow this.

(From OE-Core rev: f91322edc8b9f2a5906f3908bde2508ae97f2816)

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

perl: Fix perl module dependency issues

With the move of the strict/vars/config/warnings modules to the main perl
recipe, we need to RPROVIDE those modules to ensure that package dependencies
on those modules continue to work correctly.

(From OE-Core rev: fe88ae8605f22d9075e4200159aa66605ec36587)

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

kernel.bbclass: add kernel-modules to PACKAGES

* kernel-modules is always added to PACKAGES later in python code and
  needed to be defined as PACKAGES_DYNAMIC
* add it to PACKAGES directly and set
  ALLOW_EMPTY_kernel-modules
  FILES_kernel-modules
  DESCRIPTION_kernel-modules
  outside populate_packages_prepend like for other packages and set only
  RDEPENDS_kernel-modules from python code

(From OE-Core rev: 0884bdbbf39f2b3a8a342918812f29ddcd3b1e6f)

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

PACKAGES_DYNAMIC: use += instead of = in most cases

* to keep ${PN}-locale from
  bitbake.conf:PACKAGES_DYNAMIC = "^${PN}-locale-.*"

(From OE-Core rev: 73252b16b501c0986b0ca0895e4534895a9ba3db)

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

PACKAGES_DYNAMIC: use regexp not glob

* bitbake uses PACKAGES_DYNAMIC as regexp
  ^ could make matching faster (and it will be more clear that we're expecting regexp not glob)
  * made all those last '-' optional, use .* (or nothing)

(From OE-Core rev: 2f3ebdfa5f42dae51063b043cc4b0fbe20b40064)

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

console-tools: Fix build issues with make 3.82

The intl directory is part of older gettext and has macros which no
longer get expanded with recent gettext versions. This simply removes
the intl directory from the equation since we'd never need it.

(From OE-Core rev: 199298b6b114db09fd8ff100642ae101050a2e9a)

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

kernel.bbclass: remove explicit version.h target

The compilation routine for the kernel has an explicit call to
build version.h, which works fine for most kernels, but the
location of it has recently changes.

commit d183e6f5 [UAPI: Move linux/version.h]
commit 10b63956 [UAPI: Plumb the UAPI Kbuilds into the user
                 header installation and checking]

moves the file to include/generated/linux/version.h and then to
include/generated/uapi/linux/version.h.

As a result kernel builds of 3.7 or bisection builds of intermediate
kernel commits will fail with:

  make[2]: *** No rule to make target `include/linux/version.h'.  Stop.

Making the explicit version.h build conditional on the version, or
via a file test would fix the problem, but it introduces some complexity
to the build.

Even without an explicit call to build version.h, it is always produced
by the kernel build, so it can simply be removed.

This extra make line was originally so that the kernel version could be
determined, so that then different instructions could be executed depending
on whether it was a 2.4 or 2.6 kernel. Since we no longer support 2.4, this
code is no longer needed.

[YOCTO: #3293]

(From OE-Core rev: 4cb20fa89e571ffbc448c579a758db0b9074acf4)

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

perl: fix dependecies

This patch fixes 2 problems.

The first one is that when run "perl -V" on target, it fails with lack
of some .pm files. So add these perl module files to package perl itself
to fix this failure.

The second problem is that package nativesdk-perl-modules doesn't depends
on the single perl modules.

In the .bb file, dependencies of perl-modules are set by:

RRECOMMENDS_perl-modules = "${@d.getVar('PACKAGES', True)...}"

The PACKAGES would be reset by do_split_packages since:

PACKAGES_DYNAMIC = "perl-module-*"
PACKAGES_DYNAMIC_virtclass-nativesdk = "nativesdk-perl-module-*"

Then:
1) The target perl-modules RRECOMMENDS on perl-module-*, this is what
   we expect.

2) But the nativesdk-perl-modules doesn't RRECOMMENDS on
   nativesdk-perl-module-*, this is not what we expect.

The value of PACKAGES after do_split_packages has been set correctly (it
contains the nativesdk-perl-module-* packages)

But the:

RRECOMMENDS_perl-modules = "${@d.getVar('PACKAGES', True)...}"

doesn't work correctly for nativesdk, the

d.getVar('RRECOMMENDS_perl-modules', True)

doesn't get the new value of the PACKAGES, it gets the value of PACKAGES
before the do_split_packages.

This patch will fix the problem.

(From OE-Core rev: d50be1876f7a41822ef7e73207fdf8cccd39e400)

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

bison: Fix gplv2 version to work with recent gettext

(From OE-Core rev: 4b132d440ed97053dbef5a4deeb39e37e1167def)

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

diffutils: Fix gplv2 version to work with recent gettext

(From OE-Core rev: c3e086805649bbe782ac76670acf5858536d8801)

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

sed: Fix gplv2 version to work with recent gettext

(From OE-Core rev: 3c32d9c53f1789bdb072a5957ddd83e5c4e16914)

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

grep: Fix gplv2 version to work with recent gettext

(From OE-Core rev: 2433846255767d5a22fb2c7b2b723f290ac12fbf)

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

autotools: Use STAGING_DATADIR_NATIVE for config.rpath

For builds that don't use gettext, config.rpath may not exist in the target
datadir. This change uses the native directory where it will always
be present due to gettext-minimal-native (which allows us to autoreconf
recipes using gettext even if we don't have gettext built).

(From OE-Core rev: 0ea24447842e6b76ccfee0881f557e1a82e89ef1)

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

sstate: when warnings about sysroot overwrites, say what the recipe was

(From OE-Core rev: 936e2868bb9973213630477ab9c880dbdf4aac09)

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

gconf: Avoid error when trying to delete files that don't exist

Use "rm -f" in do_install_append() so we don't fail if the files we're
trying to delete have already been removed.  This can happen if the
distro policy suppresses both static libs and .la files.

(From OE-Core rev: 2ee9601ae590b2964d1c44b131188a80cc5198a9)

Signed-off-by: Phil Blundell <pb@pbcl.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

lib/oe/qa: Trap exceptions when running objdump

This avoids propagating a failure if we encounter an ELF file
that objdump can't parse for any reason.  Some versions and/or
configurations of objdump will refuse to read files for "the
wrong" architecture.

(From OE-Core rev: 11f5998e539f7b884ae1387252f8995b2dc7437f)

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

insane: Don't try to run objdump on symlinks

If the link is absolute then we might end up reading from a host binary
or a nonexistent path, neither of which will produce useful results and
may result in objdump failure and python backtrace spew.  If the link
does point to a binary within the installation root then we will scan the
pointed-to file at some point anyway so there is no need to do it again.

(From OE-Core rev: 91769af1c1175ac9bb43d16d05fb1c8736dd9287)

Signed-off-by: Phil Blundell <pb@pbcl.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

insane: Rationalise phdrs-based QA checks

Various different QA checks are based on essentially the same data from
the ELF program headers.  Calling objdump to extract it repeatedly is
inefficient, particularly if the shell is involved.  Instead, let's
cache the output from objdump inside the qa.elf object and allow it to
be reused by multiple tests.

Also, using objdump instead of scanelf to check for bad RPATHs (in the
same way that the useless-rpaths check was doing already) allows the
dependency on pax-utils-native to be dropped.

(From OE-Core rev: bf19eeb9f65e91bf2b5d89e7c0b099c55d7c15ff)

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

gnome-icon-theme: update mapping program location

When we changed the /usr/libexec default to be /usr/lib/<pn>,
the icon name mapping needed to be updated also.

(From OE-Core rev: d05e63c6c936a9ff209efd1341401ef7cd2ec0c1)

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

sato-icon-theme: update mapping program location

When we changed the /usr/libexec default to be
/usr/lib/<pn>, the icon name mapping needed to
be updated also.

(From OE-Core rev: 58b20de32f4b6ca684120d1b87a7aece2df6f0a6)

(From OE-Core rev: b3ff57c7a15c8509397d4b8e84410aab08ea4d1c)

(From OE-Core rev: 49d8f5d2257f00f02437b95e624960ba4aadd9d4)

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

qemu: don't ignore libexecdir in configure

This allow the relocation of libexecdir to be done correctly
for the qemu-brigde-helper.

(From OE-Core rev: 945c8f5c687ec61c312209e075edc402f6272186)

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

perf: set the perfexecdir

This allows the files installed into /usr/libexec to be
relocated to ${libexecdir}. removed unneded prefix=/usr,
which would prevent ${prefix} relocation.

(From OE-Core rev: 10d28438c1e7d793bc398a0ad484782e5baa4877)

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

ia32-base.inc: don't depend on mesa-dri

mesa-dri is an empty package, so depending on it doesn't achieve anything.

(From OE-Core rev: a41f8341971a958cf55c07f3c91e1742570053cd)

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

sysklogd: fix update-rc.d handling

The sysklogd recipe had a cut-n-paste version of the
update-rc.d.bbclass code which didn't work, but this was hidden
because all images contain the busybox version which does.  Building a
busybox-free image unmasked the issue and syslogd wouldn't start on
first boot.

The comments seem to be wrong/stale.  AFAICT update-rc.d and
update-alternatives work fine with each other, though there is an
ordering constraint (alternatives must be specified last, so it
"wraps" update-rc.d).  This version builds and works both with and
without busybox.

(From OE-Core rev: 644673631bf57bd8d0e152b5fe7621344b5ad24f)

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

pulseaudio: add missing switch-on-port-available dependency to the server

The PulseAudio server recently added a dependency by default on the
switch-if-port-available module, but this was not enforced by the package
dependencies so the server won't start.

(From OE-Core rev: b0fd984dd132e332056863dcc11b312141d443fd)

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>

lttng: support more compatible hosts

THis change extends COMAPTIBLE_HOST matchings, which allows to include more
hosts with TARGET_OS like linux-gnuspe or linux-gnueabi etc.

(From OE-Core rev: 76138d4b183eff28c678ab13cb1a6da358be2340)

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

qt4: use extra variable for more QT_CONFIG_FLAGS fragments

* qt4-embedded was forcing -DQT_KEYPAD_NAVIGATION which depends on feature-completer
* separate variable makes it easier to not enable QT_KEYPAD_NAVIGATION in some upper layer where we have disabled feature-completer

(From OE-Core rev: 0479242a18661cb7fc3d76d208c82fe6ae4378ce)

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>

gtk-doc.bbclass: Run gtkdocize in ${S} not ${B}

Otherwise it will fail if these two directories are not the same.

(From OE-Core rev: 491823fdc65d124093f1fed5a56173917443e1d6)

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

build-appliance: ensure zip file is linked locally

This makes the symbolic link portable with the dated zip file, otherwise
the link still points to the original deploy directory.

(From OE-Core rev: 0fc83102eeb48b85027c5b1202d8a584f51679a7)

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

buildhistory.bbclass: Fix hostname print for 'No changes' case

(From OE-Core rev: f9e3745d8eeef0df7e6dba3ba17d0b00645c92fa)

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>

autotools.bbclass: Fix gettext macro versions issues

gettext m4 macros don't use the usual versioning/serial mechanism used by
aclocal. It therefore won't update them over and above any local version of
the macro. Equally, we don't run gettextize due to it doing slightly crazy
things to the build.

When we put the aclocal directory as a -I option to aclocal, if this was
found first compared to any recipe provided macros, the correct version
of the gettext macro would still "win". With the switch so correctly override
the system directory, older recipe provided macros may get used.

This patch manually removes the problematic m4 macros in the case we're using
gettext and need to use the correct m4 macros.

This patch also always ensures the gettext manipulations happen, even in the
-native case since missing or stale gettext files could cause build failures.

(From OE-Core rev: e9645d2bbeabaa5251d49edd659ab320fd66d0ee)

(From OE-Core rev: 841ea3c1c18e50e77fccbd5f44d6a79a50913b67)

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

autotools: Fix race over aclocal macro directory

The previous steps taken to address races over the aclocal macro directory and the removal
of files hasn't been sufficient since aclocal still looks at that directory as part of its
default search path. This patch passes the aclocal-copy directory into aclocal as its system
directory, removing any chance of it accessing the original aclocal directory.

Hopefully this should therefore fix the race issues once and for all.

In order to do this, cp-noerror needs to not error if the directory already exists.

Its also been noticed that aclocal defaults to using STAGING_DATADIR_NATIVE even when
building for the target. Only using the target directory would cause errors such as
missing pkgconfig macros (since we only depend on pkgconfig-native, not pkgconfig).
This patch processes both sets of macros maintaining existing behaviour. At a future
date we could look into potentially optimsing this.

[YOCTO #3216]

(From OE-Core rev: ad29b331e0d61708e68ef772cdb19154956fa67e)

(From OE-Core rev: f362cc419e5a480acd16c71c802636dbedc932d9)

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

scripts/cp-noerror: Try and use hardlinks if possible

Since we generally have lots of copies of the directories created using this tool, use
hardlinks where possible. This should save a little disk space and improve performance
slightly.

(From OE-Core rev: bfa11c028c2da093f7b4e6b7b1d611da90ae052f)

(From OE-Core rev: 8c5544c2311b080bb212efb7f6b804db63e125f5)

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

scripts/cp-noerror: Copy the code from shutils.copytree, update not to error if the mkdir fails

(From OE-Core rev: 08542718504d2b53d140a9e6be73c84cc0e047e0)

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

texi2html: Add check for directory existence

Without this, if configure fails, it won't be able to run again as the directory
already exists.

(From OE-Core rev: 71a3ba536d022eea3a199cf4d6c5c791d91603a0)

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

diffutils: Remove rather bizzare gettext macros

diffutils has a rather confused set of getext macros with different names and
strange conflicting version requirements. This patch removes the problematic
macros allowing it to 'gettextize' to the latest standard gettext code without
issue.

(From OE-Core rev: a40b89333652ca22a6e6957ab8a2a4e41b87b4c0)

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

kexec-tools: Add dependency on xz

kexec-tools optionally looks for the lzma code provided by xz. Since this
is generally useful for lzma compressed kernels, add the dependency and
make builds determinstic.

(From OE-Core rev: accea64234124f25345a9288c0739c433de671f8)

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

gettext: Add config.rpath and Makefile.in.in to gettext-minimal-native

We need gettext-minimal-native to be able to install config.rpath and
Makefile.in.in so that we don't get version mismatch errors when subsequently
using the reconfigured software.

This patch moves the two files to be provided by minimal-native so
that we can better 'gettextize' software without needing the full
gettext-native when using --disable-nls.

(From OE-Core rev: 6b12d4cd39bacb087654b59e25f5052a4e839b26)

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

console-tools: Clean up recipe

This cleans up various bits of nastiness in this recipe:
  * Drop unneeded m4 macros
  * Update to a recent version of gettext (needs addition of Makevars file)
  * Drop split do_compile and SUBDIRS hacks, just patch out the docs
  * Remove some of the configure.in hacks since they seem unneeded now
    and break gettext (the AC_OUTPUT change).
  * Wipe out acinclude.m4 since it has corrisive contents

(From OE-Core rev: 87a9a3b3b2603516704a38fccc8c396e547ac101)

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

qemu: Explicitly disable bluez, its not in DEPENDS

(From OE-Core rev: d5745cdc01f7d21999c5983bae5983072e3e8f30)

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

bitbake.conf: change libexecdir to ${libdir}/${BPN}

In order to be more compliant with the Filesystem Hierarchs
Standard (FHS), this change removes the /usr/libexec default
in favor of ${libdir}/${BPN} (which is typically /usr/lib).

http://www.pathname.com/fhs/pub/fhs-2.3.html

This also address the native and STAGING variations

[YOCTO #2915]

(From OE-Core rev: 68c31b095a1cb20bd297df596024fc568614f5e8)

(From OE-Core rev: 406bd38b4232f9f399ef5ffe0b4fac72ed605a23)

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

xorg-proto: remove evieext

It was removed from xserver in 2008.

(From OE-Core rev: 574843864dcdb65d28bc2c3753339f123a9bc528)

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

xorg: remove xf86rushproto

The dependency in xserver is spurious and was removed in 2005.

(From OE-Core rev: aad06196254f1d08696ea0fcf50007ce3be933ac)

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

sanity.bbclass: trigger network tests explicitly

The network tests in sanity.bbclass can now be trigerred explicitly
by firing the NetworkTest event. This is part of the fix for bug #3026.

[YOCTO #3026]

(From OE-Core rev: f1f43d55dbb020a0145c58731d4259fd906d9d1e)

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

sstate: Use -m option to tar when unpacking sstate

We've noticed failures on the project autobuilders where a shared sstate
directory is used across multiple builders and the clocks become skewed.

Most of the time this causes harmless building but if this happens where
an environment is changed (make install vs make in qt4-x11-free for example),
the build can fail.

This avoids modification times in the future and should make builds safer
in shared environments sstate was designed for.

(From OE-Core rev: 8f1bdb4f4afd7f5f4c121be8ba82f4675f73e300)

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

libdrm: Add --disable-cairo-tests switch and update to upstream patch

Rather than implicitly relying on Cairo being disabled through not being
present, add a configure switch to forcibly disable it.

The updates the code to use a patch backported from upstream git
instead of our custom version.

(From OE-Core rev: fa9ccb23e5788f331cc868ce4bad4abd1eaeee9c)

Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libdrm: Bump git recipe to latest 2.4.39+ revision

(From OE-Core rev: 14c4c1de25b73c918a7ebb074359160290e9642f)

Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

scipts/combo-layer: Fix check_rev_branch() for cases where the revision is on more than one branch

If a revision is in more than one branch, the check_rev_branch() function can't
cope with it and the tool returns incorrect errror messages. This patch
ensures it copes with this situation.

(From OE-Core rev: 14bd101c6a86dd048da98817f47694fb21504209)

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

lttng-tools: skip new libexec insane test

Because lttng-tools installs files into /usr/lib/lttng/libexec, the test matches
and throws a false positive, so use INSANE_SKIP

(From OE-Core rev: 01300254d710c91b3dbcded9c42f6dcf21b75462)

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

qemu: add libexecdir to configure call

This address the following when libexecdir is not set to /usr/libexec

WARNING: QA Issue: qemu: Files/directories were installed but not shipped
  /usr/libexec
  /usr/libexec/qemu-bridge-helper
  /usr/libexec/.debug
  /usr/libexec/.debug/qemu-bridge-helper

(From OE-Core rev: e2fa821033785a44f80002eafac73b7e110023ce)

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

insane.conf: add new libexec test

This tests for /usr/libexec as we are moving things to /lib/.
the test is ignored if the distro defaults to /usr/libexec.

Currently this test will be disabled by default since the current
value of ${libexecdir} is "/usr/libexec".  Also this tests needs
to be enabled in the WARN_QA list.

[YOCTO #2915]

(From OE-Core rev: 4c60c2779dde6962f342f9c9b61713cf2d3a564c)

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

chkconfig: pass CFLAGS to Makefile

The environment CFLAGS is not used by the chkconfig
Makefile, so debug and optimization options are ignored.
So, we use RPM_OPT_FLAGS to pass CFLAGS into Makefile.

Upstream-Status:  Inappropriate [configuration]

(From OE-Core rev: ecdb24c21b7b90b83748cbe5891437b2183321d7)

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

packagegroup-self-hosted: recommend kernel modules instead of depending on them

Requested kernel modules may be integrated in kernel or totally disabled
as not needed on target.

(From OE-Core rev: 2129b793bc7875d929a91e22be72108d4d15e081)

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

Move 'tag=' to SRCREV in mtd-utils recipe

(From OE-Core rev: 1713cef886274b8992977900da1110390d7940d3)

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

Move 'tag=' to SRCREV in btrfs-tools recipe

(From OE-Core rev: acb942dbaf118b8021100df95a68fe49dcc77858)

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

populate_sdk_base.bbclass: Make it possible to override the create_shar method of populate_sdk_base.

If you wish to change the install/unpack method of the sdk, this can now be done by making
your own create_shar method, and setting a SDK_PACKAGING_FUNC variable to your
new create_shar function.

(From OE-Core rev: 3955c8eced352226bb4c9ceb710dbe02474b9024)

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

gypsy: removed gypsy from meta/recipes-connectivity

removed also entry from meta/recipes-gnome/packagegroups/packagegroup-sdk-gmae.inc

(From OE-Core rev: 817c4e661da1e9baa947a420d690248971301102)

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

recipe_sanity: Don't bother checking LICENSE

Since e3d7890cace71b0a57d2530bf615a58dcb46d96f or so, base.bbclass has
considered invalid LICENSE settings to be a fatal error.  This means we
will never see them so there is no point checking for that.

(From OE-Core rev: e2d71503847f72f55666143a2a320925838fd26f)

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

libnss-mdns: fix mDNS resolving speed

We need to fix the "hosts: files dns mdns4" nsswitch.conf line
because for a .local lookup it does a DNS lookup first which will fail.
The recommended solution is:
hosts: files mdns4_minimal [NOTFOUND=return] dns mdns4

[YOCTO #2502]

(From OE-Core rev: dbb350b90417962f2da4c1064ab0174badeb0f26)

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>

gnu-config: update to 2012.08.14 to get support for AArch64 architecture

(From OE-Core rev: fd082d328f1312847097794dea588ed670206cb4)

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

pcmanfm: mark AArch64 as compatible

(From OE-Core rev: 468c074ba15cd4b93600d5ba5a1fdc032718d7c3)

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

packagegroups: disable kexec, valgrind, lttng, systemtap on AArch64

(From OE-Core rev: c9514779de7fa6ea4cfa0c911cff25ea8c6a5152)

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

util-linux: add AArch64 support

(From OE-Core rev: 1dbf17d221ee9d50c5de8c04144c92fdc78d6d73)

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

openssl: add AArch64 support

(From OE-Core rev: 490b12126aff7e8e59569ebb471ce04ba4962b7c)

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

insane.bbclass: add AArch64 support

(From OE-Core rev: 39d71c8c41276091688decb64d418e1e3637e2b6)

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

kernel-arch.bblass: add AArch64 support

(From OE-Core rev: 9583a46221580ba46ebfd6d561d3a4d6b0d42eea)

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

siteinfo.bbclass: add AArch64 support

(From OE-Core rev: fde788cf5b0e480a675d2aa256af0915a120bc65)

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

xorg: remove fontcache support

This was removed from the Xorg server in 2008.

(From OE-Core rev: 02fc0a197ca16e983becb0aedeb9238a0aeb6661)

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

libxfont: remove spurious dependency on fontcacheproto

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

xorg: remove all traces of XPrint

The XPrint server was removed from Xorg in 2008.

(From OE-Core rev: 5b3748d463a6666c0d8e2624092619da8d8e6328)

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

xserver-xorg: use INC_PR in PR

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

xorg: remove XTrap

This functionality was broken upstream so it was removed.

(From OE-Core rev: 7661d15957525885e5e9b1129da7a99eef19f4be)

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

mesa: remove libegl-dbg, put all debugging into mesa-dri-dbg

(From OE-Core rev: c914495e1431ad56fcd81460fa4f675be3b4be3b)

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

mesa: remove mesa-dri dependency in mesa-dri-dev

mesa-dri is empty, so instead of allowing an empty package, remove the default
dependency on mesa-dri and let the system not generate mesa-dri.

(From OE-Core rev: 5d6596321a996278ffbaa111247367ec9e50d721)

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

xf86-video-intel: bump to latest release

(From OE-Core rev: fa248f86105ddd4a1b0be039ee8e85ab275c4e44)

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

xserver-xorg: upgrade to 1.11.4

gcc-47-warning.patch was integrated, so drop.

Add pkgconfig-deps.patch, a backport of a commit to fix bad exposed library
dependencies (which resulted in the keyboard driver depending on pixman).

(From OE-Core rev: 723e81af2a5b07024ab744c14cdccc12f554ef12)

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

strace: Don't package strace-graph

Commit 9c9ea24b115a9bb87df1323b5f185ce426262aec made strace depend on perl because the
strace-graph script needs it.  However, this cost of the dependency is large (building
all of perl) and the value of the script is marginal.  Let's just delete the script
instead and remove the dependency again.

If anybody wants strace-graph then it should be packaged in its own recipe and that one
can be made to depend on perl without disrupting the main strace package.

(From OE-Core rev: 2e887af1c81f9b373684180f61a7c25163ed0e2c)

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

module.bbclass: Move do_make_scripts() to module-base

It's sometimes useful to have this function available to recipes which
don't wish to use module.bbclass for whatever reason.

(From OE-Core rev: 7632b44e7f487180811d47fbe9c29aa8e58868a2)

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

xf86-input-keyboard: upgrade to 1.6.2

(From OE-Core rev: 1c2ffdf26bdb2e1443dddb143e0e96760e6572bf)

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>

kernel.bbclass, module.bbclass: Make update-modules optional

The update-modules mechanism is something of a historical relic and it isn't
entirely clear that it has a great deal of value nowadays.  Also, it causes a
problem when building a read-only rootfs since update-modules itself refuses
to configure offline.

Allow DISTROs to circumvent this whole thing by declaring (via DISTRO_FEATURES)
that they don't wish to use update-modules.  This is backfilled for existing
distributions and will have to be marked as CONSIDERED by those who actually
don't want it.

(From OE-Core rev: 14bf8ed115453077b4d4042b4b70ed6b3bca2a9f)

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

gettext: Remove spurious-looking dependencies on libxml2-native

These were added in aae5021101224344a2b1a3af5becf74291fbbfe5, ostensibly to resolve
some sort of "host contamination" issue.  However, that commit contains no information
about what contamination was being observed or how the dependencies help.

gettext-native is being configured --with-included-libxml so it shouldn't be using
libxml2 from either the host or the sysroot, in which case the dependency would seem
to be useless.  Using the included copy of libxml2 is in any case preferable to adding
a dependency on libxml2-native because the latter brings quite a large stack of other
dependencies with it.

(From OE-Core rev: 132d329638ae32b98c36b9498c470cf0ffdcedb3)

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

xf86-video-fbdev: bump to 0.4.3

(From OE-Core rev: 6157644e3ca06a97bee2294d1a55c5c17b0f52b1)

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-input-evdev: bump to latest release

(From OE-Core rev: 1f73a2ded611de659be05efe734753f538fd84a1)

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>

cpan_build: Unify directory layout for native and target builds

There seems to be no benefit in having them be different and this
appears to cause a certain amount of confusion about paths for
the native modules.

(From OE-Core rev: 3926a45a26c2d23d9c2fc9c3956f780f607ec7a4)

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

Add packagegroup-core-buildessential

[Not heavily tested, but sent for discussion]

task-core-sdk is too big - for example, I simply don't want to ship tcl, ever =)
Historically distcc caused a large dependency graph explosion because it has
a utility which uses gtk+, although that was fixed.

packagegroup-core-sdk also includes coreutils, which is a bit
confusing; conceptually things dependent on coreutils should pull it
in explicitly, or possibly we just declare coreutils to be in the
minimal build set.

So packagegroup-core-buildessential is intended to be similar to
Debian's "build-essential" package.  It's the stuff needed by say 80+%
of components, not worth repeating over and over.

(From OE-Core rev: 7d6cc169c95fecf6388a275281eb8b8f5d8eb4a2)

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