If this isn't set, pkg-config will use its inbuilt paths which mean it can
end up looking in the native sysroot.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
Move the environment-setup script generation into a separate class and adapt
meta-toolchain to use it.
Add a new dummy target, meta-ide-support, to install an appropriately
configured script to TMPDIR and ensure all of the required packages for target
development are available.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
from 3.0.3
changes:
- rebased patch configure.patch and pcimodules-pciutils.diff
- fix a bug in parameter order in config that causes invalid
search path and library version
- remove unnecessary kernel version reference in configure
- use DESTDIR mechanism for build and installation
- pciutils should RDEPENDS on pciutils-ids
- enable shared library build and package a new package series,
libpci
- conceive a new patch to fix shared library build, default build
would embedd invalid library dependencies
- add another patch to fix build on arm
- fix a bug in previous do_install_depends that used invalid
installation mode
- symbolic link /usr/sbin/pciutils to /usr/bin, so non-root user
can use it as well
Signed-off-by: Qing He <qing.he@intel.com>
for ppc glibc build failes with -os (optimize for size) gcc option. so
the fix disables this gcc options on ppc.
MIPS fix is added in the hope of helping webkit-gtk build failure (gcc
seg fault).
And fix the PR bumps for all affected recipes
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
use autoconf to generate jsautocfg.h instead of preparing
different headers for different archs
size_of and align_of tests also work under cross-compilation.
Signed-off-by: Qing He <qing.he@intel.com>
Performance counters for Linux are a new kernel-based subsystem
that provide a framework for all things performance analysis. It
covers hardware level (CPU/PMU, Performance Monitoring Unit) features
and software features (software counters, tracepoints) as well.
To enforce the coupling between userspace and kernel, this
commit introduces perf as a subpackage of the linux-wrs tree.
perf is built directly inside the kernel tree (and hence picks
up all appropriate patches), but is packaged as a separate
utility that can be added to the rootfs and used to analyze
the kernel.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Adding new kern_tools to allow the compilation of a meta series
and the patching of a kernel tree.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
This change adds the ability to dynamicaly modify a git based
linux repo during the do_patch and do_configure changes.
The bulk of the work is done by the kern_tools, and the recipe
simply needs to add createme and modifyme scripts to go along
with the existing configme.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The patch adds a script in /etc/network/if-pre-up.d,
the script returns 1 (effectly skip the ifup on the iface) iff:
1. there is a "* / (nfs|nfs4) *" entry in the /proc/mounts
2. the addr field in that entry is routed using $IFACE
Signed-off-by: Qing He <qing.he@intel.com>
version 2.2.49
- reuse attr build system include file
- the default build will add attr's rpath into libacl.so, a
hack is used here to prevent that hardcode. The hack is ugly
but simple, without the need to do considerable autoconf hacks.
An alternative approach is to use chrpath, but it doesn't have
support for multilib thus not usable in its current state.
- manual fix to .la
Singed-off-by: Qing He <qing.he@intel.com>
copy from qemux86 to make xserver works on qemu x86-64 platform. However
full GUI doesn't pop up yet due to other keycode/theme problems in matchbox.
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
memset is required after malloc in grab_module, or else random segfault
may happen. The fix is from Mark Hatle <mark.hatle@windriver.com>.
Upstream bug is reported as:
https://bugzilla.kernel.org/show_bug.cgi?id=16528
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Move enchant from meta-moblin to meta
Remove "S" since it is same as default path.
Also fix the metadata
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
changes:
- enable extras, this includes pci/usb and acl related utilities
- pci.ids automatic search doesn't work, specifying manually
- package udev glib binding into libgudev{,-dev,-dbg}
- libudev RDEPENDS on udev, this is somehow counter-intuitive:
libudev is the library to access udev information
- should RRECOMMENDS usbutils-ids and pciutils-ids
bug: RDEPENDS (RRECOMMENDS) on a specific version still doesn't work
Signed-off-by: Qing He <qing.he@intel.com>
from 0.72
changes:
- package new usbutils.pc
- move binaries into /usr/{bin,sbin} as default. udev uses only
usb.ids instead of binaries
- remove DEFAULT_PREFERENCE
- symbolic link /usr/sbin/lsusb to /usr/bin, lsusb can be used by
non-root, inspired by Debian
- separate usb.ids to a new package usbutils-ids, for finer grained
control, inspired by pciutils, also usbutils RDEPENDS on
usbutils-ids
- don't rm ${S}/libusb any more, since usbutils now doesn't bundle
an included libusb
Signed-off-by: Qing He <qing.he@intel.com>
changes:
- separate the build system to ea-acl.inc, it is also used by
upcoming acl (ea stands for extended attributes)
- the .so.* is moved to /lib, so it can be used by packages like
udev
- change absolute symbolic links to relative path, code inspired
by udev
- manual fix to .la
Signed-off-by: Qing He <qing.he@intel.com>
from 1.5
use version string 1.7.0+1.8.0rc1
aka. spidermonkey
changes:
- rebase jsautocfg.h to add JS_HAVE_LONG_LONG, fixing the build
failure
- document the limitation of jsautocfg.h
- jskwgen is a host script, and should be built by BUILD_CC
- remove dependency of readline. readline or editline is only
used in standalone js command line, which is not shipped in
package. Removing the dependency to avoid any concern of
GPL3-ization
- set BUILD_OPT=1 for optimized build
Note on version choice:
js is known to have long rc cycles. e.g. the version 1.5:
js-1.5-b1: 12/16/1999, js-1.5-rc1: 03/17/2000,
js-1.5-rc6a: 06/16/2004, js-1.5 release: 10/22/2005
Current release version is 1.7.0: 10/19/2007, latest version
1.8.0-rc1: 03/09/2009, so 1.8.0-rc1 is used.
Note on standalone js vs. xulrunner:
xulrunner also includes a js version, that is more updated than
standalone spidermonkey. It also uses autotools so jsautocfg.h
patch may not be necessary. However the version of xulrunner is
different from js, and it's difficult to identify the mapping
between them. It needs further investigation on xulrunner's
bundled js
Signed-off-by: Qing He <qing.he@intel.com>
Previous update was reverted due to some reason. This commit re-introduce the
update after following efforts:
* rebase owl-window-menu.patch - Translate the menu of gtk_ui_manager to
original gtk menu, to enable owl feature.
* fix cross-compile issue - 0.9.7 use a temp binary, built from its own source
code, to optimize the data file size. This binary for target can't run on
build system. As this size optimization only gain 0.2K improvement for one
data file(used for about dialog), disable it to work around this issue. No
perf drop found after this fix.
Signed-off-by: Zhai Edwin <edwin.zhai@intel.com>
On an x86_64 host, both 32-bit and 64-bit libraries need to be generated.
It is fairly common that a user will have at least a few 32-bit programs
on their x86_64 host system, so a pseudo wrapper for 32-bit is required
to allow those programs to be successfully wrapped.
Signed-off-by: Mark Hatle <Mark.Hatle@windriver.com>
Change the pseudo integration:
* Uprev to latest open source version
* Restructure the patches to allow for many local DBs, as well as
pseudo specific lib dirs.
Signed-off-by: Mark Hatle <mhatle@windriver.com>
Use the virtual fakeroot program when installing a package and also
during packaging. This is important as it allows us to track full
permissions, owners, groups and special files generated by packages.
Signed-off-by: Mark Hatle <mhatle@windriver.com>
gcc hard codes the linker path to lib64 for x86_64 machines, update our
64bithack patch to change this to lib.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Most notable change is the move to creating symlinks to patches in the metadata
tree rather than copying them.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
mesa-dri first generates glsl compiler and then use it to compile GL
shader sources. Target glsl compiler can't run on build system, and
thus port from openembedded by introducing a native recipe dedicated
to creating a native glsl-compiler.
This has to be an explicit -native recipe since its un-native sibling
is already part of mesa-dri
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
opkg_unarchive.patch: This patch was trunkating the filenames silently
for no reason. took it out.
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
A previous commit added AutoAddDevices as a separate ServerFlags
section. This is apparently overridden by the one at the end of the
file, so they need to be merged.
[BUGID #141]
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
[BUGFIX#158] tcf-agent launched with misleading options
In the commit 90e80aafbd, with the image built by
poky-image-sdk(MACHINE=qemux86), tcf-agent is launched automatically with
the option "-s SSL:" in script /etc/init.d/tcf-agent, which is misleading to
users. This option tells the agent to use SSL instead of TCP for communication.
Because tcf-agent is not built with the SSL option enabled (libssl is not
installed in the target), the agent will fall back to TCP.
This could be misleading to users, so we may as well remove it.
Thanks Lianhao Lu for reporting the bug and clarifying it.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Add conf/distro/include/private/distro_sdk_tools.inc;
Move some sdk tools from distro_x11_libs.inc to the new .inc file;
Add tracking info for lttng-ust.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
CLIENT_PATH is the only environment variable of dhclient-script,
change it to include major run path, [BUGFIX#132]
Signed-off-by: Qing He <qing.he@intel.com>
Once ipk meet errors, the log will show "Collected errors:\n xxxx ".
so keyword "Collected errors" is more general to catch errors.
Signed-off-by: Yu Ke <ke.yu@intel.com>
avahi will create an empty /var/run dir, which will conflict with base-files package.
This patch fix this by using populate-volatiles.sh approach recommended by OE handbook.
Signed-off-by: Yu Ke <ke.yu@intel.com>
The X server on some netbooks don't interact well with HAL with their
keyboard or mouse and consequently one or both is unusable. This
patch adds AutoAddDevices = False to a new netbook-specific
xorg.conf. If this breaks any systems which currently work, we'll
have to figure out what's really going on with hotplug and the X
server.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
genext2fs 1.4.1 creates minimal inode number based on specified rootfs
directory. This is desired in some embedded devices as storage/memory
are precious. However it's not suitable in development phase where target
rootfs may be changed heavily on the fly (remote debug, test, ...).
Sometimes this may even cause "No space on device" error due to limited
free inodes exhaulted at the 1st boot.
Here a new option is added to allow falling back to original 1.3 behavior,
i.e caculating inode number based on specified bytes-per-inode parameter.
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Fix parse error with packages such as spectrum-fw whose name has no
version string. Later we may have per-recipe option to disable automatic
check for those known with troubles, to reduce complexity in this part.
Signed-off-by Kevin Tian <kevin.tian@intel.com>
default cmake behavior for FIND_***, is to first search listed directories prefixed
with CMAKE_FIND_ROOT_PATH, and if not found then go to unprefixed directories. This
is not desired as build system may be referenced.
this commit refine the search path only on prefixed directories, and thus fix
libmusicbrains build failure on some systems
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
install a proper script /etc/init.d/tcf-agent
make it autostart in runlevels 3 and 5
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Signed-off-by: Lianhao Lu <lianhao.lu@intel.com>
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Imported OE patches into the poky recipes.
- Use elfutils instead of libelf
Converted OE's svn source into tar ball & a patch.
Added a patch to revert the baseversion to 4.5.0
Merge OE & poky extra_oeconf_options
Fix the zlib (inside gcc) make issue by providing the --with-system-zlib switch in EXTRA_OECONF
Found out that some header file dirs were soft linked to non-existing
locations like c_std. Changed the configure options to point them to
existing locations like c_global.
gcc-cross-canadian_4.5.0: fix configure issue
Thanks to Saul Wold for providing poky gcc patches rebased to the 4.5.0
sources.
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Import the recipe from OE
Upgrade from 0.8.1 to 0.8.2, and it solved the build failure
Cleaned up the libmpc-native recipe for poky tree
Add nativesdk target for gcc-cross-canadian_4.5.0
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Add the new cross prelink support as the prelink-native integration.
Set the prelinker to use the prelink-cross.git repository as the upstream.
Note: libiberty.a is required on the host at this time
Signed-off-by: Mark Hatle <mhatle@windriver.com>
Add elfutils native, also change the binaries being installed to be
prefixed with 'eu-' to avoid conflict with binutils
Signed-off-by: Mark Hatle <mhatle@windriver.com>
Enable use of BBCLASSEXTEND to generate a native verison of binutils for the
libiberty, libbfd and libopcodes libs.
Also stop installing useless libs in the cross-canadian recipe.
Partially based on a patch by Mark Hatle <mhatle@windriver.com>.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
When using the fake chroot ability of pseudo, it will expand absolute
symlinks to their full non-fake path by default. The simple change disables
that behavior, as it is undesired when generating a rootfs.
Signed-off-by: Mark Hatle <mhatle@windriver.com>
Re implement the rootfs generation using rpm5. This also gets rid of the
need for yum, and handles all dep resolving internal to the script itself.
The new file scripts/rootfs_rpm-extract-postinst.awk comes from the original
yum integration work. It has been unchanged, but since yum is no longer used
we needed to move the script somewhere else.
Signed-off-by: Mark Hatle <mhatle@windriver.com>
Switch to RPM5 as our rpm provider of choice and update the recipe to the latest
stable release.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Modify the package_rpm.bbclass to understand the macro and command line changes
present in rpm5.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Add new patches to fix visiblity of C++ methods and to disable running a
program to detect the ICU version.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Enable changing the data directory on the fly from the environment and then use
this feature within poky to confine pseudo usage to each WORKDIR.
This fixes issues that could be seen under heavy inode reusage e.g.
with rm_work.
Work based mainly off a patch from Joshua Lock but finished by Richard
Purdie.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
* tar 1.23 is GPLv3
* tar 1.17 is latest GPLv2
Recipes and patches dervied from OpenEmbedded and Debian
Signed-off-by: Scott Garman <scott.a.garman@intel.com>
* at 3.1.12 is GPLv3
* at 3.1.10.2 is latest GPLv2
Recipes and patches derived from OpenEmbedded
Signed-off-by: Scott Garman <scott.a.garman@intel.com>
New dependency on libfm & menu-cache, which are added in previous commits
following patches are removed as they either don't apply or this new version
has no those problems:
desktop.patch
no-fam.patch
no-warnings.patch
pcmanfm-mips-fix.patch
the pending one is aowl-window-menu.patch, which need more work as the sources
have changed a lot. So far w/o this patch it still works. but we need more
verification later
Signed-off-by: Zhai Edwin <edwin.zhai@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
older versions are kept untouched for now
changes:
- udev has diverged a lot, the udev.inc was based on a quite dated
version, thus doesn't reflect current status of udev, create a
new include file udev-new.inc to adapt
- separate libudev{,-dev,-dbg} apart from udev
- --with-udev-prefix is removed. change ${libexecdir} to
${base_libdir}/udev, which is described in INSTALL file.
- udev.pc is provided in ${datadir}/pkgconfig, since it's used
to indicate the existence of udev itself instead of development
headers and libs, pack it in udev instead of udev-dev
- udev-extras is disabled since we are missing libacl, this also
means gudev is missing as for now
- other cleanups, e.g. do_stage, do_install, EXTRA_OEMAKE
todo:
- udev-extras
- remove old udev versions
- rebase udev-145 on udev-new.inc as well
Signed-off-by: Qing He <qing.he@intel.com>
* Update the following files with 1.3.2 version in upstream:
GNUmakefile.am
Makefile
Makefile.shared
autogen.sh
configure.ac
* Update the .bb with the one in openembeded(91eefa1e)
* inherit pkgconfig and remove dolt stuff according to RP's comments
Signed-off-by: Zhai Edwin <edwin.zhai@intel.com>
New file for tracking the following new packages:
* at
* cpio
* cronie
* crontabs
* grep
* mailx
* openssh
* tar
Signed-off-by: Scott Garman <scott.a.garman@intel.com>
expat 2.0.1 archive is generated with DOS format, which causes CRC
error for some version of gzip (<1.4), e.g. on Unbuntu 10.04. This
commit adds do_unpack dependency to gzip-native, and thus to use
local compiled binary to address this issue. Since 1.4 is a known
version solving this issue, also specify a default version for
gzip-native
also move most lines into expat.inc as cleanup
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
borrow from OE gzip 1.3.13 after some cleanups:
- no ${S} redefinition
- configure.patch is not required
also add native support
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
libusb, libusb1, libusb-compat
remove libusb, upgrade libusb1 and libusb-compat. libusb-compat
now provides libusb
util-linux
update the latest version to 2.17.2
hal
hal-info
udev
dhcp
Signed-off-by: Qing He <qing.he@intel.com>
from version 4.1.0
changes:
original .bb file is based on dhcp3, whose build system differs from
dhcp 4.x, add a new dhcp4.inc and preserve dhcp3.inc for reference,
this should fix the conf file location bug of dhclient/dhcpd 4.1.0
- defining _PATH_DHC{LIENT,PD}_CONF in make parameter no longer works,
put these definition to includes/site.h
- delete upstream version of conf files, which are not used, and
with an installation path hardcoded to ${sysconfdir}
- similar thing happen for leases file, use new configure option to
specify
- RANLIB, LIBDIR, etc. as make parameter are no longer necessary
Signed-off-by: Qing He <qing.he@intel.com>
from 2.17
changes:
- clarify license. the package does contains a standalone utility
called lscpu that is under GPLv3+, fortunately it's note packaged
as for now
- the installation path of getopt has changed
todo:
- move libblkid and libuuid into separate packages
Signed-off-by: Qing He <qing.he@intel.com>
from 20090414
changes:
- consolidate hal-info.inc and hal-info_git.bb, git version is used
as a reference and is not preferred
- preferred version updated to 20091130
Signed-off-by: Qing He <qing.he@intel.com>
from 0.5.13
changes:
- udev rules dir has been changing to /lib/udev/rules.d in udev
upstream, add an override to stay in /etc/rules.d for compatibility
- consolidate hal.inc and hal_git.bb, hal_git.bb is used as a
reference thus should have a negative preference
- change preferred version of hal to 0.5.14
- remove hal_0.5.11+0.5.12rc1
- remove RRECOMMENDS on udev-utils, the package is to be removed
Signed-off-by: Qing He <qing.he@intel.com>
from 0.0.9, also remove libusb recipe (see below)
changes:
- remove some unnecessary logics derived from libusb-0.1
- now PROVIDES libusb
- bluez4 depends on libusb instead of libusb-compat
libusb-0.1 vs. libusb-compat
libusb-compat is intended to be a drop-in replacement of libusb-0.1,
building on top of libusb-1.0 (which has different API from -0.1).
Few known packages don't work with libusb-compat, notable example
is libmtp. Since most packages work very well with libusb-compat,
and libusb-compat doesn't suffer from power consumption issue of
libusb-0.1 (which is often very important for embedded systems),
drop libusb-0.1 and move to libusb-compat completely. If we see
any problems in the future, either the package should be fixed
or we can add back libusb-0.1 as needed.
Signed-off-by: Qing He <qing.he@intel.com>
from 0.9.2
changes:
- the previous recipe was derived from libusb-0.1 and thus had many
inconsistencies, e.g. binconfig and lib_packages are not necessary,
libusb1 doesn't provide libusb++, etc. Remove these legacies
- remove the patch 0.9.0-gcc3.4-compat-fix
Signed-off-by: Qing He <qing.he@intel.com>
Required when updating pcmanfm to 0.9.7
[with cleanup from Richard Purdie]
Signed-off-by: Zhai Edwin <edwin.zhai@intel.com>
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
required when adding libfm
[with cleanup from Richard Purdie]
Signed-off-by: Zhai Edwin <edwin.zhai@intel.com>
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
startup-notification_0.10 depends on xcb-util, so I add this package.
xcb-util.inc and xcb-util_0.3.6.bb are based on recipes from OpenEmbedded;
The changes I made are: updating HOMEPAGE,BUGTRACKER,LICENSE and
LIC_FILES_CHKSUM, and changing "include xcb-util.inc" to
"require xcb-util.inc", and using the latest version 0.3.6.
add gperf dependency and remove do_staging
Signed-off-by: Saul Wold <Saul.Wold@intel.com>
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>