classes/buildhistory: do git garbage collection after committing

We don't normally perform any operations (such as "git pull") that
trigger "git gc --auto", thus garbage collection never happens which
means performance of accessing the repository degrades noticeably over
time. Add an explicit "git gc --auto" to clean things up when needed.

Thanks to Elijah Newren and Ross Burton for suggesting this.

(From OE-Core rev: 4a45a999e0ad2e99581428a5a6d34f483c00544f)

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

classes/buildhistory: reduce parsing time

Disable several python functions if not parsing within the worker
context. This avoids executing expensive operations while parsing
recipes (which is unnecessary).

(Thanks to Richard Purdie for pointing out the issue and suggesting the
workaround.)

(From OE-Core rev: 540a2a30be21c3eca4323efbe91e7dcfc31a4c97)

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

classes/buildhistory: improve collection of package info

Use a function added to SSTATEPOSTINSTFUNCS and read the necessary
information out of pkgdata, instead of using a function executed
during do_package that reads the data directly. This has two benefits:

* The package info collection will now work when the package content is
  restored from shared state

* Adding/removing the inherit of buildhistory will no longer change the
  do_package signatures and force re-execution of that function for
  every recipe.

Fixes [YOCTO #5358]

(From OE-Core rev: cd7f7efcd5f297d876823b8f579ecefb9542b089)

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

classes/buildhistory: add additional variables to image information

Add PACKAGE_EXCLUDE and NO_RECOMMENDATIONS to the info we track for
images, since these can change what ends up in the image.

(From OE-Core rev: a10189366f180b87f5be20b66834b7e7a9bb8c12)

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

classes/buildhistory: drop cruft from old SRCREV tracking implementation

This should have been removed when the implementation was rewritten in
OE-Core commit 2179db89436d719635f858c87d1e098696bead2a. The collected
values weren't being used anywhere since then.

(From OE-Core rev: cbc23a87c1897b7fda40f452dd36acb0bca3d197)

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

classes/package: write PE and PKGE out to pkgdata

These are important parts of the version for every package, so we should
include them in PKGDATA just as we include PV/PR/PKGV/PKGR.

(From OE-Core rev: 5ceed97ba02a698f1c260c3f56cdf2cc156e6d8b)

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

classes/package: record PKGSIZE as total file size in pkgdata

We were using "du -sk" to collect the total size of all files in each
package for writing out to PKGSIZE in each pkgdata file; however this
reports the total space used on disk not the total size of all files,
which means it is dependent on the block size and filesystem being used
for TMPDIR on the build host. Instead, take the total of the size
reported by lstat() for each packaged file, which we are already
collecting for FILES_INFO in any case.

Note: this changes PKGSIZE to be reported in bytes instead of kilobytes
since this is what lstat reports, but this is really what we should be
storing anyway so that we have the precision if we need it.

Fixes [YOCTO #5334]

(From OE-Core rev: 29615b36fca696822a715ece2afbe0bf9a43ed61)

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

classes/package: fix FILES_INFO serialisation in pkgdata

The FILES_INFO entry in each pkgdata file stores the list of files for
each package. Make the following improvements to how this is stored:
* Store paths as they would be seen on the target rather than
  erroneously including the full path to PKGDEST (which is specific to
  the build host the package was built on)
* For simplicity when loading the data, store complete paths for each
  entry instead of trying to break off the first part and use it as the
  dict key
* Record sizes for each file (as needed by Toaster)
* Serialise the value explicitly using json rather than just passing it
  through str().

Fixes [YOCTO #5443].

(From OE-Core rev: ca86603607a69a17cc5540d69de0e242b33382d3)

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

babeltrace: correct PV variable

(From OE-Core rev: ceba66859f87904066e67504a53fc8b07f4b1111)

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

classes/icecc: fix typos in header comments

Looking at the code, it should be ICECC_USER_PACKAGE_BL and not
ICECC_PACKAGE_BL here, and also fix "localy" -> "locally".

Thanks to Scott Rifenbark for pointing this out.

(From OE-Core rev: b325e46059efe6974abf01131dca3f2094a6bf90)

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

linux-dummy: catch up with image.bbclass changes

image.bbclass now depends on virtual/kernel:do_deploy, so add a task for that.

This fixes errors like this:

ERROR: Task do_build in /build/linaro/build/meta-linaro/meta-linaro/recipes-linaro/images/linaro-image-lng.bb depends upon non-existent task do_deploy in /build/linaro/build/openembedded-core/meta/recipes-kernel/linux/linux-dummy.bb

(From OE-Core rev: 26d07f2a5bef42a113c9c81f2b5701b4f3d10d47)

Signed-off-by: Koen Kooi <koen.kooi@linaro.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: hob: fix rotating progress icon animation

I've always found it somewhat annoying that this icon's rotation
animation wasn't quite correct; this was because it was looping around
to the second position instead of the first, which made it appear to
stutter.

(Bitbake rev: 78ad15b669b9c7cde41f7bd1ab884c1d2e0db91b)

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

bitbake: imagedetailspage: Fix crash with more than 15 layers

If you had more than 15 layers the system would crash since one more
value is added to one array than the other. This fixes the code
so equal numbers of values are added to the arrays and hence
doesn't crash when many layers are enabled.

(Bitbake rev: 4e65463886a2ef245b2f8974e82e9cb942af224b)

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

bitbake: bitbake.lock: Add host:port to bitbake.lock for memres server

The idea is to build on the --status-only option for bitbake and
expose a mechanism where the oe init scripts can easily switch between
memres server and the non-memres server.

In the case of the standard oe init script the following
can shut down the server:

if [ -z "$BBSERVER" ] && [ -f bitbake.lock ] ; then
    grep ":" bitbake.lock > /dev/null && BBSERVER=`cat bitbake.lock` bitbake --status-only
    if [ $? = 0 ] ; then
       echo "Shutting down bitbake memory resident server with bitbake -m"
       BBSERVER=`cat bitbake.lock` bitbake -m
    fi
fi

A similar function can be used to automatically detect if the server
is already running for the oe memres init script.  This new
functionality allows for the memres init script to be started in a new
shell and connect up to an alaready running server without seeing the
error of trying to start the server multiple times.

(Bitbake rev: b1803958de8d7c3c3279841e38604a08dc2316cc)

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

bitbake: bitbake, xmlrpc.py: Implement memory resident auto port configuration/restart

This patch adds the ability to dynamically select a port for the
bitbake memory resident server when the BBSERVER port is set to -1.
This allows for running multiple instances of the bitbake memory
resident server on the same system in different build directories.

The client portion of the bitbake instance can also request that the
server automatically start when using the auto port feature.  This is
to deal with a bitbake instance that eventually times out and exits or
that has died for some unknown reason.

The new functionality allows for lazy startup of the server after
sourcing the init script for the memory resident functionality.

(Bitbake rev: d6abc07ff385357d312d8435b89e0a9c1f965433)

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

bitbake: bitbake: Add --status-only option

The --status-only option is purely to check the health of the server.
The idea is to use it from the oe environment script so as to know if
the server needs to be started or not.

(Bitbake rev: 46b26ef5c27bfa9c91680f494f02750305b8098c)

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

oe-init-build-env-memres: Add auto port functionality

In order to run multiple bitbake memory resident servers on the same
machine they must each use different ports.

This patch works in conjuction with bitbake to make the auto port
selection and lazy server startup the default.

(From OE-Core rev: 9cf1ac73c4e35101a4f5c01a5e1c53f9d567bc58)

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

oe-init-build-env: Allow startup with and without memres

Use the bitbake --status-only and the fact that bitbake.lock will
contain the host name and port to determine when to activate or
shutdown the stay resident bitbake server.

This allows a end developer to cleanly switch between the two ways to
use bitbake as well as enter the memres bitbake server from multiple
shells without starting the server if it is already running.

(From OE-Core rev: d71059c86a8160f39af6ddfdd30c86835f4eb959)

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

gcc-cross-canadian: Fix fortran build

When fortran was enabled, builds were failing due to a extra files.
For now we can remove these and avoid the build failure.

(From OE-Core rev: 2e60ef7fe63974e443a9ddc25c5eb4249ec37963)

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

eglibc-initial.inc: Drop duplicate include

There is little point in including the file twice so lets not. The
main recipe already included it.

(From OE-Core rev: b3cccee0c66ce744a79843a5dd9798475c84e23c)

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

bitbake: hob: disable layer drag and drop outside the containing widget

[YOCTO #5577]
(Bitbake rev: d39d957c77b9804670cad3bc9f7742084ead27e8)

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

bitbake: parse/ConfHander/BBHandler/utils: Fix cache dependency bugs

Currently bitbake only adds files to its dependency list if they exist.
If you add 'include foo.inc' to your recipe and the file doesn't exist,
then later you add the file, the cache will not be invalidated.

This leads to another bug which is that if files don't exist and then
you add them and they should be found first due to BBPATH, again the
cache won't invalidate.

This patch adds in tracking of files we check for the existence of so
that if they are added later, the cache correctly invalidates. This
necessitated a new version of bb.utils.which which returns a list of
files tested for.

The patch also adds in checks for duplicate file includes and for now
prints a warning about this. That will likely become a fatal error at
some point since its never usually desired to include a file twice.

The same issue is also fixed for class inheritance. Now when a class
is added which would be found in the usual search path, it will cause
the cache to be invalidated.

Unfortunately this is old code in bitbake and the patch isn't the
neatest since we have to work within that framework.

[YOCTO #5611]
[YOCTO #4425]

(Bitbake rev: 78d285871e4b8c54ccc4602d571e85f922e37ccd)

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

qt4-x11-free: depend on ICU

ICU presence is auto-detected at configure time and until recently (e68850 and
d61230) was pulled into most builds through harfbuzz and beecrypt.  Now it's
floating and this leads to build failures.

As in all likelihood the majority of people were building this with ICU enabled,
add an explicit dependency.

(From OE-Core rev: 46dcec6fd455584d9b5d0d7ff1e5b36fbe5a2d62)

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

sqlite3: Fix configure option warning

WARNING: QA Issue: sqlite3: configure was passed unrecognised options: --disable-tcl

There is no tcl option or support now so remove the option.

(From OE-Core rev: 7d1fd11e50a280bcc3e6c2160871cdef14864cc2)

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

util-linux: Remove obsolete configure options

WARNING: QA Issue: util-linux-native: configure was passed unrecognised options: --enable-rdev --enable-init --enable-arch --with-fsprobe

Remove obsolete configure options to silence warning.

(From OE-Core rev: ec2b080e0698ffc7e12faf69a60bd81a1d94bea0)

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

linux-yocto/3.10: -rt, ebtables and e1000 fixes

Updating the 3.10 SRCREVs for the following fixes:

  f47ea28 bridge: enable EBTABLES
  a9ec82e e1000: prevent oops when adapter is being closed and reset simultaneously
  a4e1bd7 Revert "arm: add dummy swizzle for versatile with qemu"

(From OE-Core rev: 20139a693677c23324c89c963735f3b8c18eeb84)

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

linux-yocto/3.4: update to v3.4.69

Bumping the 3.4 kernel to the latest korg -stable release.

(From OE-Core rev: b9049f6394d58610ecd3b0c77ccee5ef95ebdb73)

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

linux/yocto-3.10: merge v3.10.19

Updating the linux-yocto-3.10 SRCREVs to the latest korg -stable
release.

(From OE-Core rev: 1da738722c9f81260dfbada6a97a21eb812d20a1)

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

webkit/midori: block to build webkit on mips64 with libn32 ABI

To mips64 libn32, The HOST_SYS is mips64-*-linux-gnun32, the COMPATIBLE_HOST
is mips64.*-linux, the regular expression matching both will return true. so
append the ending position word into COMPATIBLE_HOST to make match failure

(From OE-Core rev: f5c4d18c1b0d8a252abb3b1f67f05a1689b3ea67)

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

alsa-state: add status command

(From OE-Core rev: c8d687f409a19312b34e215e7caaa39199598ed0)

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

apmd: add status command

(From OE-Core rev: a3f3f44f9a064a0158ce9ad12af7a7cb26042c97)

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

ppp: add status command

(From OE-Core rev: 12c4094ce118569f9518e1b0625d110251595ef0)

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

busybox: add busybox_git.bb recipe

Add busybox_git.bb recipe so that it would be easier to hack with
busybox. Set DEFAULT_PREFERENCE to "-1" in the git recipe so that we
still use the busybox_1.21.1.bb recipe by default.

(From OE-Core rev: d290bbe6d9826fbcfa2e0a7624886284697ed7eb)

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

linux-firmware: Remove bash scripts from target package

The carl9170fw is unbuilt, needs specialise toolchains, cmake and so on
so we might as well delete it (and lose the bash dependnecy).

Equally, the top level bash dependency from the empty configure script is
pointless.

[YOCTO #5555]

(From OE-Core rev: e2b62393497a0aaffb21161336a42249db61a10d)

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

hdparm: make wiper appear before hdparm in PACKAGES

Make wiper appear before hdparm in PACKAGES so that wiper is packaged
correctly.

[YOCTO #5555]

(From OE-Core rev: d294f1de082a5aadefa8efb3485ca3f091ef83bf)

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

libsdl: Clean up configure option QA warnings

WARNING: QA Issue: nativesdk-libsdl: configure was passed unrecognised options: --disable-video-gem --disable-video-xbios --enable-dlopen --disable-debug --enable-endian

The gem and xbios options exist but their helptext is wrong and there
is no "video" in the name. This patch removes the now obsolete options
and corrects the names of the others to match reality.

(From OE-Core rev: 8cea5dd7d6804a8d739f5db6a5814ab6c802d538)

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

adt_installer: Fix missing autogen.sh error

With the transition from svn -> tarball release, this issue was
missed. Instead of using autogen.sh, we can call configure
directly.

[YOCTO #5606]

(From OE-Core rev: 776e3b19160e5e1cceec1a8941c831be4d1f82b0)

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

lsb: de-bash scripts to remove bash dependency

De-bash some script in lsb to make it not dependent on bash.

[YOCTO #5555]

(From OE-Core rev: 5ba55f9936d1f036518a1722f9c1e551477e167c)

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

image.bbclass: Depend on virtual/kernel:do_deploy

Now that none of the packagegroups depend on virtual/kernel, we have the problem
that MACHINE=qemumips bitbake core-image-minimal doesn't put a kernel
into the deploy directory. This breaks many common usecases and
user expectations.

To avoid this, add a dependency on the kernel deploy to image do_build tasks.
This should avoid any circular dependency issues but equally ensure users
have their expectations met.

[YOCTO #5581]

(From OE-Core rev: fe26b2379ecdbdb56acde8592bc0c2d95092a207)

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

base/gcc-common: Ensure umask setting is consistent for shared workdir

gcc has cross and target components with a shared workdir. The unpack umask
settings need to match for all of these. We need to use strings in each
case to ensure the sstate code matches them correctly.

This patch tweaks various things to ensure the change adding the unpack umask
change doesn't break the compiler builds.

(From OE-Core rev: 67162438ee9c402b23c32853af9d313949eb6e4a)

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

sstatesig.py: Fix image regeneration issue

With the "ABI safe" recipes, we've been excluding those from signatures. This
is fine in the general case but in the specific case of image recipes it breaks.

A good test case is the interfaces file. Editting this causes init-ifupdown
to rebuild but not an image containing it (e.g. core-image-minimal).

We need to ensure the checksums are added to the image recipes and this change
does that.

(From OE-Core rev: fd085f15e7cd093953f974f69277e130174d551d)

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

bitbake: cooker: Fix support for wildcards in bbappend filenames

The 'bbappend in f' incorrectly compares the current recipe with the avaliable
bbappends recipes. This comparsion causes unrequested bbappend files to be
appended, e.g. in the case of 'libgcc_4.8.bb', the bbappends for 'libgcc_4.8.bb'
and 'gcc_4.8.bb' are added to the filelist (because 'gcc_4.8.bb' is contained in
the 'libgcc_4.8.bb' string) which in turn causes the gcc_4.8.bbappend files to
be appended to the libgcc_4.8 recipe.

This should be a 'bbappend == f' to match the previous implementation of this
function, such that if no wildcard is present the recipe names must match
exactly.

This issue was introduced by commit 31bc9af9cd56e7b318924869970e850993fafc5f,
which it related to [YOCTO #5411].

(Bitbake rev: 991cbeedbde8bd25ce08c669b1bfac8b99e33149)

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

base.bbclass: Set umask 022 also for do_unpack task

* when git checkouts files from fetched clone it respects system umask
  and creates files with different permissions, if such files are copied
  to packages, resulting target images have also different permissions
  on them.
* we need reproducible builds across different builders with different
  system umask, so set 022 umask

[YOCTO #5590]

(From OE-Core rev: c9289c506633ffe5c482000d8d225e45454c064d)

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

xinput-calibrator: add formfactor to RDEPENDS

* 30xinput_calibrate.sh is calling ". /etc/formfactor/config"
  breaking Xsession for images without formfactor

(From OE-Core rev: 181a46da02d6ae74a8d1b5d06c547e0d213767ea)

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

module-base: Fix misleading comment

The comment was originally written for module.bbclass and is now
slightly misleading. This updates it to match the current code.

(From OE-Core rev: 434277ed156c1685283f6fd681062d265f4fa6d0)

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

Revert "kernel: restore scripts in the sysroot"

This reverts commit 6a6735cb98d529c8c5376aedb5c951339f54fe96.

The module class already ensures the scripts are rebuilt correctly. Running
this at sstate installation time is problematic since it can require the
cross compiler. Adding such a dependency would cause issues of its own.

(From OE-Core rev: b2c948d56241ff7cdea2e9e68b740f305c72f5ca)

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

rpcbind: Fix rpcbind restart failure when rpcbind is not running

"rpcbind restart" executes stop and then start function. However, if
rpcbind is not started, "exit 0" shall be run in stop function, so start
function will not be run at all. This patch changes "exit 0" to "return 0".

(From OE-Core rev: ce17144b736b364175ab76e19e720292edfd2d81)

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

elfutils: fix the condition check before remove eu-objdump

"grep -q" does not write anything to standard output, the result is same as
a empty string, The second test becomes [!""] which is a fixed true value

(From OE-Core rev: 79b0979ecd33ce15563354c90d8bcc857ddf95ad)

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

libav: install libraries to right directory when multilib is enabled

Explicitly set libdir and shlibdir to ${libdir} in EXTRA_OECONF. Otherwise, default library path of ${prefix}/lib is used which is incorrect in a multilib build.

(From OE-Core rev: e16b6bab8d5286cdf58d808ef4c195127d69a8c8)

Signed-off-by: Nick D'Ademo <nickdademo@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

weston: upgrade to 1.3.1

(From OE-Core rev: 4a9f855ecb209aebe37f9a26f3f4129a570a3af4)

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

chrpath: Improve crazy code

The current code is a little bit overcomplicated, deficient and also
possibly broken.

Issues include:

a) Not maximally optisming rpaths (e.g. a lib in usr/lib might get an
   rpath of $ORIGIN/../../usr/lib)
b) The return in the middle of the for loop look suspiciously like
   it might break on some binaries
c) The depth function, loops of "../" prepending and so on can
   be replaced with a call to os.path.relpath

This patch cleans up the above issues.

Running binaries should result in less "../" resolutions which can't
hurt performance either.

[YOCTO #3989]

(From OE-Core rev: feea54df6768036649ca6c57524e2a1f480ad249)

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

icecc: Disable icecc for allarch recipes

* Fix icecc.bbclass failing with:
  ERROR: recipe-name NULL prefix
  when it's used with empty TARGET_PREFIX.
* Allarch recipes cannot use compiler at all (even the local one)
  so there is no point of using icecc for them.

(From OE-Core rev: a956f9d91c8128e43b55c6bc01337472e47fe43a)

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

cronie: fix init.d/crond status return value

(From OE-Core rev: da9c74917804a823bcf122b778aef273c3b64ede)

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

tcf-agent: add init.d/tcf-agent status command

(From OE-Core rev: 5b3c4863665eabf7750ff1bdd975ea1f4772006b)

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

zlib: Add ptest

Install zlib tests and run them as ptest

(From OE-Core rev: 2988cef2f0ad857b5bbf6a0189ffb0fb88795f8c)

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

tcl: fix version string and make recipe multilib build compatible

Explicitly set libdir in EXTRA_OECONF so that the correct library folder is generated in a multilib build.

The version string (VER) has been changed to 8.6.1 and the library paths have been updated accordingly so that the related tk recipe can correctly detect tcl (this search is done using the tclConfig.sh script which contains the tcl version number).

(From OE-Core rev: e840f526e7223c9d393aab818c7a5a446b89c503)

Signed-off-by: Nick D'Ademo <nickdademo@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

file: remove useless patches

no user uses these two patches

(From OE-Core rev: 47e7b47def5b06a1d825bf0883409510cb4da36f)

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

clutter-1.0: upgrade to 1.16.2

(From OE-Core rev: eba3c24036f69a84ec3f0bf4ab23b8e7db7b24f8)

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

librsvg: upgrade to 2.40.1

(From OE-Core rev: 8f5f99a9c403d10e480d21f65147d9f67dfc60a0)

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

fontconfig: upgrade to 2.11.0

Drop sysroot patch as a rewritten form has been accepted upstream.

Update license checksums - Google added to COPYING, and the fccache license has
moved.

(From OE-Core rev: e24659fc039c9f54b841ed01c3d5ff407921398b)

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

pango: upgrade to 1.32.6

Drop automake macro patch, merged upstream.

(From OE-Core rev: 22f1eb5165592b47baa35e28c5cb3d4faff36bae)

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

libdrm: upgrade to 2.4.49

(From OE-Core rev: 03f6f0ca01d83f633dcdcd692cc2df7bb0613d72)

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

bitbake: crumbs/builder.py: Fix typo for indefinite article

(Bitbake rev: bc3c69e50d36ea5e43f55f7b3b1ff3b0ab3696ee)

Signed-off-by: Yi Zhao <yi.zhao@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

libpng: set reasonable SUMMARY

Also don't set DESCRIPTION to the same value, it's superfluous.

(From OE-Core rev: f991d2d60b74f5ebd990f77aecd3324b1a4533e9)

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

xproto: upgrade to 7.0.25

(From OE-Core rev: 2509a9b8f1486fc3b7a16672d0c8618a168f7c6d)

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

git: upgrade to 1.8.4.4

(From OE-Core rev: 0565b1906d59a05b43e8658d9776e7ffbd8fe4ae)

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

libxcb: upgrade to 1.9.3

Added 2 new extensions: dri3 and present. Moved libxcb and xcb-util
recipes to xorg-lib directory.

Removed the following patch(es):
 * automake_1.14_fix.patch (backport)

(From OE-Core rev: 657ffd8bd6115fa2d61a918a67d6b8f162d50c5a)

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

xcb-proto: upgrade to 1.9

Also move xcb-proto recipes from xcb directory to xorg-proto directory.

(From OE-Core rev: 2c17403eb0cd3a966249a70327c9879375c6a7ee)

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

powertop: upgrade to 2.5

(From OE-Core rev: 879cb228e4ad4a8901caf8eda630bc9355fd8dfa)

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

tar: upgrade to 1.27.1

(From OE-Core rev: 58151b7e7ce651619e09fd8129f252332a342521)

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

python-setuptools: upgrade to 1.4

(From OE-Core rev: 0d47d440a3674b3eafb605f738e98f82f5698f05)

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

xtrans: upgrade to 1.3.2

(From OE-Core rev: f279a1fc1b6b7e72d8eec782af81cb198ec07e76)

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

systemtap: upgrade to 2.4

(From OE-Core rev: fb8ccbd06929aa539c39e65fb8926a945db922ea)

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

libav: use CCLD as ld

The libav uses gcc as the ld, but it doesn't use CCLD, it may have
problems when target arch is 64 bit since it doesn't use the "-m64", the
poky's toolchain is fine since use "x86_64-poky-linux-gcc" without
"-m64" is default to 64 bit, but external toolchain *may* default to 32
bit (for example, when multilib is enabled and both 64 and 32 bit use
the same gcc, then the default arch can be either of them), then there
would be errors, the error is just like we run this in poky:

$ x86_64-poky-linux-gcc -m32 <file.c>

ld: skipping incompatible /path/to/sysroot/usr/lib64/x86_64-poky-linux/4.8.2/libgcc.a
when searching for -lgcc

Use CCLD as the ld will fix the problem since CCLD has been set
correctly.

(From OE-Core rev: 7afbc62be0e4720fb7cd2e44ec9e438a7e4ff78f)

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

libpng: upgrade to 1.6.7

License is the same.

(From OE-Core rev: a1e8def7b335a69fb1ca412ad82cafa15350422e)

Signed-off-by: Valentin Popa <valentin.popa@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

coreutils 6.9: fix coreutils.texi

This is used for fixing coreutils 6.9 (GPLv2+) do_installed failed:

[snip]
| coreutils.texi:2499: @itemx must follow @item
| coreutils.texi:2636: @itemx must follow @item
| coreutils.texi:2644: @itemx must follow @item
| coreutils.texi:2654: @itemx must follow @item
| coreutils.texi:2677: @itemx must follow @item
| coreutils.texi:2689: @itemx must follow @item
| coreutils.texi:2820: @itemx must follow @item
| coreutils.texi:3058: @itemx must follow @item
| coreutils.texi:3253: @itemx must follow @item
[snip]

Use '@item' instead of '@itemx' in several places, as Texinfo 5 refuses
to process an '@itemx' that is not preceded by an '@item'.  Ensure that
node extended names in menus and sectioning are consistent, and that
ordering and presence of nodes in menus and in the actual text are
consistent as well.

[YOCTO #5593]

(From OE-Core rev: 04fab782f42b8f5047390042618f9c841b8c3a96)

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

bitbake: Hob: force notebook refresh

Redraw the entire notebook widget after
number-of-issues drawable is incremented
(because for this case, the notebook widget
doesn't refresh it's children automatically).

[YOCTO #5596]

(Bitbake rev: 347b2ead091f00ee60703f6f3d17cfdd9075ac07)

Signed-off-by: Valentin Popa <valentin.popa@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: data/codeparser: Improve handling of contains functions

One of the current frustrations with the sstate checksums is that
code like base_contains('X', 'y',...) adds a full dependency on X
and varies depend even on whitespace changes in X.

This patch adds special handling of the contains functions to expand
the first parameter and check for the flag specified by the second
parameter (assuming its a string).

The result is then appended to the value of the variable with a "Set"
or "Unset" status. If the flag is added/removed, the stored variable
value changes and hence the checksum changes. No dependency on X
is added so it is free to change with regard to other flags or
whitespace.

This code is far from ideal, ideally we'd have properly typed variables
however it fixes a major annoyance of the current checksums and
is of enough value its worth adding in a stopgap solution. It shouldn't
significantly restrict any propely typed variable implementation in
future.

(Bitbake rev: ed2d0a22a80299de0cfd377999950cf4b26c512e)

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

bitbake: bitbake: Share BB_TASKDEPDATA with tasks

Currently tasks have no knowledge of which other tasks they depend
upon. This makes it impossible to do at least two things which would be
desirable/interesting:

a) Have the ability to create per recipe sysroots
b) Allow the aclocal files to be present only for the entries in
   DEPENDS (directly and indirectly)

By exporting task data through this new variable, tasks can inspect
their dependencies and then take actions based upon this.

(Bitbake rev: 84f1dde717dac22435005b79d03ee0b80a3e8e62)

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

bitbake: runqueue: Optimise next_buildable_task()

This unlikely looking function was found to be eating a lot of CPU time
since it gets called once per trip through the idle loop if we're not
running a maximum number of processes. This was particularly true in
world builds of 13,000 tasks.

Calling the computation code is pretty pointless because until some
other task finishes nothing is going to become available to build.
We can know when things become available so this patch teaches the
scheduler this knowledge.

It also:

* skips any coputation when nothing can be built
* if there is only one available item to build, ignore the priority map
* precomputes the stamp filenames, rather than doing it every time
* saves the length of the array rather than calculating it each time
  (the extra function overhead is significant)

Timing wise, initially, 5000 iterations through here was 20s, with
the patch 200000 calls takes the same time. The end result is that
builds get up and running faster.

(Bitbake rev: 4841c1d37c503a366f99e3a134dca7440e3a08ea)

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

bitbake: codeparser: Drop unneeded variable separation

There is no good reason to separately track var_references and
references so merge them and remove the unneeded variable.

(Bitbake rev: 64d4cbd6360c96574cece70205ea3aecc3f8bae6)

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

classes/image: write image manifest

Write a list of installed packages to a .manifest file next to the
image, so we can find out what went into the image after it has been
constructed without necessarily having to have buildhistory enabled
(although that will provide more detail.) We can make use of this for
example in the testimage class associated code that checks for installed
packages for determining whether or not to run specific tests.

Note: this replaces the previous ipk-specific manifest code with
something that works for ipk, rpm and deb, and instead of a pruned
status file, packages are listed one per line, in the following format:

<packagename> <packagearch> <version>

Tests for all three backends have shown that the performance impact of
this change is negligible (about 1.5s max).

Implements [YOCTO #5410]

(From OE-Core rev: 2978d1f2617a33e2e3a77e249d73e998d79b4ec9)

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

harfbuzz: Add PACKAGECONFIG for icu

harfbuzz can be built without icu. We don't need harfbuzz-icu for any
default OE-Core configuration so default to icu being disabled for
performance improvements.

(From OE-Core rev: d61230ac70158dd9a33fcfac4eea768d21ccc61d)

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

beecrypt: Add PACKAGECONFIG for cplusplus

Add a PACKGECONFIG to control the building of the beecrypt C++
bindings. The only user of beecrypt in OE-Core is rpm and this doesn't
need the C++ bindings so default the option to be off. This means
we can lose the icu dependency by default which is a significant
performance win.

(From OE-Core rev: e6885069e2af833ebacfd33a04147b095af92d20)

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

qemu: add PACKAGECONFIG for vnc, libcurl, nss, uuid, curses, gtk+, libcap-ng

Use PACKAGECONFIG to explicitly address vnc, libcurl, nss, uuid, curses, gtk+,
libcap-ng dependencies rather than tested by configure.

It avoided potential errors while multiple builds shared a common state_cache.

(From OE-Core rev: 4482af07df26644885bae49b98f5d765a5caa68c)

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

bzip2: Remove stange copy line, automake does this for us anyway

(From OE-Core rev: 2ba32f7bd9deec0d977d7d2ff30275af54a41892)

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

linux-firmware: add missing linux-firmware-iwlwifi-7260-7 package

The FILES / RDEPENDS lines were added for this package, but not the
entry in PACKAGES, so it was never being created.

(From OE-Core rev: 25a75e83550fab0f9d2486b13ec9ab6339b6a8b0)

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

sanity: Use random filename for maximum path length test

check_create_long_filename used a fixed filename for its test files. This
meant that os.remove(testfile) could fail with ENOENT if two instances were
running at the same time against the same sstate directory. Using a
randomly generated filename stops this from happening.

(Although it might seem unlikely, this race did appear to occur multiple
times with Jenkins - presumably because the matrix jobs were all kicked off
at the same time.)

(From OE-Core rev: bc28e3f26e7f85af82f403924c0ae29e1ad34a87)

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

sanity: Don't hard code value of ENAMETOOLONG

Although ENAMETOOLONG is 36 on Linux x86 and x86_64 it does isn't on other
architectures so the value shouldn't be hard coded.

(From OE-Core rev: 11a9cf5ee0daf82097fb2f36b58016f20a5968f3)

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

bluez4: added dependency on 'libsndfile1'

bluez4 detects and uses libsndfile1 and the compilation can fail with

| sbc/sbctester.c:32:21: fatal error: sndfile.h: No such file or directory
| ...
| compilation terminated.
| make[1]: *** [sbc/sbctester.o] Error 1

in rebuilds (image with libsndfile1 was built, then some change ->
bluez4 do_configure runs with libsndfile1 -> libsndfile1 gets removed
-> bluez4 do_compile fails).

As there is no trivial way to disable its detection and to make it a
PACKAGECONFIG option, 'libsndfile1' was put into static DEPENDS.

(From OE-Core rev: b9571256f8996d1eb4b9a09b3b5b862a13f1b414)

Signed-off-by: Enrico Scholz <enrico.scholz@sigma-chemnitz.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

avahi: add leading space to RRECOMMENDS append

* in case update-rc.d is already in RRECOMMENDS it fails with
  ERROR: Nothing RPROVIDES 'update-rc.dlibnss-mdns' (but
  meta/recipes-connectivity/avahi/avahi_0.6.31.bb
  RDEPENDS on or otherwise requires it)

(From OE-Core rev: 70dedb67c2b8b7302dc4c51e8c607e57f61f530a)

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

webkit/midori: fix COMPATIBLE_HOST

webkit can not be built on mips64 with n32 ABI, but can be built on mips64
n64 and o32 ABI whose TARGET_SYS's name is mips-*-linux

(From OE-Core rev: 66cf1cc01b8e4f6284e13d57d9fdcb9f228a6846)

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

x264: install libraries to right directory when enable multilib

x264 use [EPREFIX/lib] as default libdir. When multlib is enabled that
is not right. Packages depends on x264 such as libav configure fails
that can't find library x264.

Pass the right libdir to configure script to fix it.

(From OE-Core rev: d1deb07d158cf27bce2ee95e2f02b4fd1d00fe21)

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

attr: attr_2.4.47 release tarball missing configure.ac

The release tarball of attr 2.4.47 is missing the configure.ac file.
This prevents the autotools bbclass from being able to regenerate the
configure script which leads to other side affects. (e.g. using
outdated config.sub)

This patch adds the configure.ac file via a patch.
(obtained via the source repository at the v2.4.47 tag)

See the mailing list thread for additional information:
http://lists.nongnu.org/archive/html/acl-devel/2013-05/msg00024.html

(From OE-Core rev: b28f12a272a9e2f0c3084a58b91605acb05f58f8)

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

bitbake: fetch2: Correct a few calls to latest_revision()

In 6a48474de9505a3700863f31839a7c53c5e18a8d the url parameter to a
number of functions was removed. However, not all calls to
latest_revision() were fixed...

(Bitbake rev: 7c94ca56b2fd85a989089f58b3dcce3172a778f2)

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

bitbake: bitbake: fetch2: Remove the unused urldata from the git fetcher.

(Bitbake rev: e0a2e9699e9f338dd7ade8c9eef0a12c7639ec05)

Signed-off-by: Florin Sarbu <florin.sarbu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>

bitbake: data: Fix output inconsistencies for emit_var

VAL = ""     (not shown)
VAL = " "    (shown as "")
VAL = " x"   (shown as "x")

would all show up rather differently to what would be expected in the
bitbake -e output. This fixes things so they appear consistently.

The output for running some shell functions may also change slightly
but shouldn't change in a way that is likely to cause problems.

[YOCTO #5507]

(Bitbake rev: fcba5ef0053dc0ef5360e4912609e5d52f5046b0)

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

bitbake: perforce: Fix path subdirectory issues

With a SRC_URI = " \
p4://depot/folder/...;module=localfolder/localsubfolder;changeslist=${P4CHANGELIST} \
"

the subfolders of //depot/folder/... get renamed when mapped to the
local folder structure. They lose the first 3 letters. This
patch fixes that.

Issue reported by and patch sent from katutxakurra@gmail.com

[YOCTO #5380]

(Bitbake rev: 40e06dc459d9c0b5d42d65b2d2c846196fd36b1f)

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

bitbake: bitbake-layers: avoid loading configuration when not needed

In recent versions of bitbake, it is not possible to initialise a
BBCooker object without having it load the configuration first. Thus we
should avoid creating the Tinfoil object here in bitbake-layers which
does that internally until we actually need to, so you can run
"bitbake-layers help" and not have to wait several seconds for the
output.

(Bitbake rev: 8f1e280fbbb6432d7bcc1fb4241f402668c6c5ea)

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

bitbake: HOB: Show warnings in Issues tab

Proper update of warnings view during build.

[YOCTO #3496]

(Bitbake rev: 124428ba8a0b1cc85d4b96053bf71bce14a5774c)

Signed-off-by: Valentin Popa <valentin.popa@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>