We were passing empty string '' to make as option when
ISA was not thumb this breaks build for machines/distros
which do not use ARM_INSTRUCTION_SET = thumb
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
klibc failed to build with armv7+ when using thumb2
it needed to resepct TUNE_CCARGS
Add patches to fix build with musl, to sort out
the linux header maze.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Cc: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
On modern buidsystems with many cores there are sporadic build failures.
The issue is known and a fix in the sources is awaited (new Kbuild).
http://www.zytor.com/pipermail/klibc/2013-May/003420.html
For the moment, disable PARALLEL_MAKE.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Recipe does build with the pending changes for meta-initramfs.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
As done in libtool-cross from oe-core we prefer to be specific even when
there is just one file.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Instead of deleting the packaging tasks we simply inherit
nopackages.bbclass.
Done that we do set PACKAGES to empty.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The klcc script encodes the paths with custom escaping.
As stated in the patch "Turn the horribly encoded paths into something which
sstate can transform using its ususal magic by removing all the crazy escaping".
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
getVar() now defaults to expanding by default, thus remove the True
option from getVar() calls with a regex search and replace.
Search made with the following regex: getVar ?\(( ?[^,()]*), True\)
Signed-off-by: Joshua Lock <joshua.g.lock@intel.com>
klibc installs everything into /usr/lib by default. But the recipe assumes
it installs to ${libdir} which is /usr/lib64 in 64bit environment. Package
generation fails badly in this case. Set INSTALLDIR to fix this.
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Revert commit dd1ca18 removing ${PR}. Bad idea.
The shared lib could not be found giving this error:
Collected errors:
* satisfy_dependencies_for: Cannot satisfy the following dependencies for
klibc-utils-XXX:
* libklibc (= 2.0.4) *
* opkg_install: Cannot install package klibc-utils-XXX.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Fixes:
WARNING: File 'usr/lib/klibc-HWWIvOaMdZy4N4-6rLLHAB2TjFQ.so'
from klibc was already stripped, this will prevent future debugging!
WARNING: File 'usr/lib/klibc/lib/klibc-HWWIvOaMdZy4N4-6rLLHAB2TjFQ.so'
from klibc was already stripped, this will prevent future debugging!
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Following debian packaging-style we have a copy of klibc-xxx.so in libklibc-dev.
(Comments can be found in klibc.spec.in)
Fixes:
ERROR: QA Issue: -dev package contains non-symlink .so:
ibklibc-dev path 'work/i586-oe-linux/klibc/2.0.4-r0/packages-split/
libklibc-dev/usr/lib/klibc/lib/klibc-HWWIvOaMdZy4N4-6rLLHAB2TjFQ.so'
-dev package contains non-symlink .so:
libklibc-dev path 'work/i586-oe-linux/klibc/2.0.4-r0/packages-split/
libklibc-dev/usr/lib/klibc/lib/libc.so' [dev-elf]
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Minimally rebase staging.patch and follow upstream commits.
Fixes:
| usr/klibc/socketcalls/socket.o: In function `socket':
| .../usr/klibc/socketcalls/socket.S:8: multiple definition of `socket'
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
32bit x86 is covered by the x86 over-ride, so duplicate over-rides for
i486/i586/i686/pentium are not required.
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
For the recipes built with klcc-cross it is necessary to
pass --sysroot otherwise we default to the one encoded in gcc-cross which
actually is the 'first one' built.
The issue was revealed when building for armv4 after having built for armv5te:
the produced binaries did contain Illegal Instruction (bx lr).
Use ${TOOLCHAIN_OPTIONS} variable to specify --sysroot=${STAGING_DIR_TARGET}
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The two recipes for klibc-utils (static and shared) have been split-off
from a single original recipe for more granular packaging and share the
same SRC_URI set in the .inc file so there is an implicit dependency.
Make it explicit for the utils needing the shared klibc lib at runtime.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Klcc-cross is properly a crosscript so avoid to inherit cross.bbclass and stage
it in target sysroot under usr/bin/crossscripts.
Minor adjustment is necessary for INSTALLDIR.
Finally some magic mangling is applied fixing the paths of sstate so that it is
invalidated in case of subsequent builds targeting machines with the same
arch and in case of builds sharing the sstate cache.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
No runtime changes.
Reorder and fix some typos in the comments.
Remove unneeded empty do_install from utils.
Fix bogus PACKAGES_${PN} = "${PN}".
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
In case of subsequent builds for machines belonging to the same arch
we have to rebuild the klcc-cross wrapper because it is harcoding
the path to the headers in machine sysroot.
This hack is necessary to avoid to mark the klibc infrastructure as
machine-specific.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
It's not necessary to specify the protocol parameter when it's the
default protocol for the fetcher, e.g. the default protocol for
git fetcher it git, "protocol=git" isn't needed.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* This change is only aesthetic (unlike indentation in Python
tasks).
* Some recipes were using tabs.
* Some were using 8 spaces.
* Some were using mix or different number of spaces.
* Make them consistently use 4 spaces everywhere.
* Yocto styleguide advises to use tabs (but the only reason to keep
tabs is the need to update a lot of recipes). Lately this advice
was also merged into the styleguide on the OE wiki.
* Using 4 spaces in both types of tasks is better because it's less
error prone when someone is not sure if e.g.
do_generate_toolchain_file() is Python or shell task and also allows
to highlight every tab used in .bb, .inc, .bbappend, .bbclass as
potentially bad (shouldn't be used for indenting of multiline
variable assignments and cannot be used for Python tasks).
* Don't indent closing quote on multiline variables
we're quite inconsistent wheater it's first character on line
under opening quote or under first non-whitespace character in
previous line.
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Acked-by: Koen Kooi <koen@dominion.thruhere.net>
* linker fix is needed to avoid 'bx, lr' instruction on armv4
* patch is applied only for armv4 for eabi builds
* http://wiki.debian.org/ArmEabiPort#Choice_of_minimum_CPU
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Used sed expression given here:
http://lists.linuxtogo.org/pipermail/openembedded-core/2011-November/012373.html
Plus an additional expression for .expand. Full expression is:
sed \
-e 's:bb.data.\(setVar([^,]*,[^,]*\), \([^ )]*\) *):\2.\1):g' \
-e 's:bb.data.\(setVarFlag([^,]*,[^,]*,[^,]*\), \([^) ]*\) *):\2.\1):g' \
-e 's:bb.data.\(getVar([^,]*\), \([^, ]*\) *,\([^)]*\)):\2.\1,\3):g' \
-e 's:bb.data.\(getVarFlag([^,]*,[^,]*\), \([^, ]*\) *,\([^)]*\)):\2.\1,\3):g' \
-e 's:bb.data.\(getVarFlag([^,]*,[^,]*\), \([^) ]*\) *):\2.\1):g' \
-e 's:bb.data.\(getVar([^,]*\), \([^) ]*\) *):\2.\1):g' \
-e 's:bb.data.\(expand([^,]*\), \([^ )]*\) *):\2.\1):g' \
-i `grep -ril bb.data *`
Some minor correction in systemd.bbclass was needed for some expressions
that didn't quite match the regex in the desired way; additionally a few
instances were manually changed.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* 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)
* use += instead of = in most cases to keep ${PN}-locale from
bitbake.conf:PACKAGES_DYNAMIC = "^${PN}-locale-.*"
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* and remove bogus dependencies for the static utils
* introduced by do_split_packages
* bump INC_PR
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
* include last patches of may 2012
* tracking upstream wrt forthcoming 2.0 release
* bump INC_PR
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
