Keep the la files in source directories unchanged,
and do relocation in destination directories (sysroot-destdir).
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
{e}glibc should install the scsi/*.h files, which are needed by hal.
Currently hal can build because eglibc-initial (which has its own do_install)
installs the files. In future eglibc will have its own sysroot, so hal
won't build.
BTW: in OE side, eglibc's do_install also doesn't remove the files.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Fixes [BUGID #595]
Because of the QEMU booting slowness issue(see bug #646 and #618), autobuilder
may suffer a timeout issue when running sanity test. We introduce variable
SHARE_IMAGE here to fix the issue. It is by default set to 1. Poky will copy
latest built-out image and keep using it in sanity testing. If it is set to 0,
latest built-out image will be copied and tested for each case, which will take
much time.
Signed-off-by Jiajun Xu <jiajun.xu@intel.com>
This makes 'kernel-modules' installable again when perf isn't enabled in the kernel config
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We use host perl while building target perl modules. When some of the
prebuilt perl modules such as scalar::util are used in the build process
directly or indirectly, perl needs to load the arch specific .so library
file. If perl can not find the .so library files, then perl
tries to use the target pm files which ends up in circular perl module
depedancy causing build time perl invocation to fail.
Adding this new parameter viz EXTRA_PERLFLAGS to cpan.bbclass allows perl
module recipes to specify the location of such host .so files for hostperl,
so that build time invocation of perl does not fail.
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
perl 5.12.2 does not use /usr/share/perl path, and all that stuff goes
in /usr/lib/perl. This commit fixes cpan class which depends on
/usr/share/perl.
cpan.base/build.bbclass: change /usr/lib/perl5 to /usr/lib/perl
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
configme used to be able to calculate the output/build directory
when branches were always <machine>-<kernel type>. Branch names
can now be widely different and to avoid embedding complexity
in the scripts it is easier to just pass ${B} from the build system
down to the scripts.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
During a poky-image-minimal build I ran into the following error:
rm: cannot remove `/vol/1/dvhart/poky.git/build/tmp/work/atom-pc-poky-linux/poky-image-minimal-initramfs-1.0-r0/rootfs/install': Is a directory
This is caused by a missing -r (recursive) argument to rm. This patch
adds that to the rm command of rootfs_rpm.bbclass.
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
CC: Bruce Ashfield <bruce.ashfield@windriver.com>
This patch intend to fix the random unpack failure of linux-libc-headers-yocto
and linux-yocto.
The root cause of the unpack failure is that: these two recpies has the same URL, thus
has the same dest file during the fetch and unpack phase:
do_fetch : create tar ball ${DL_DIR}/git_git.pokylinux.org.linux-yocto-2.6.37.tar.gz
do_unpack : extract tar ball ${DL_DIR}/git_git.pokylinux.org.linux-yocto-2.6.37.tar.gz
fetch phase is protected by lockfile, so it works fine. but unpack phase is not lock protected,
thus there is race condition like: when linux-yocto do_unpack is extracting the tar ball,
linux-libc-headers-yocto do_fetch starts to create tar ball thus overwrite linux-yocto's
tar ball and cause linux-yocto do_unpack failure
To fix this issue, do_unpack also need to be protected by lock
Signed-off-by: Yu Ke <ke.yu@intel.com>
previously you could only build one set of packages for multiple machines:
MACHINE=foo bitbake task-base
MACHINE=bar bitbake task-base
would only create task-base packages for foo, but not for both foo and bar.
Doing MACHINE=bar bitbake task-base -c cleanall would remove the packages for foo.
The solution is to use MULTIMACH_ARCH as suggested by Richard Purdie.
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
This fixes various incorrect sysroot usages observed by people with sstate.
Thanks to Kevin Tian for figuring out where the problem was and Gary Thomas
for confirming the fix.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
BSPs are built from a particular branch of the kernel repository
which is specfied via the mapping of MACHINE to KMACHINE. Unless
a global branch is being forced (like libc headers), KMACHINE
is an override on a per machine basis.
Because KMACHINE is typically override we must first try the
most specific variant KMACHINE_<machine> and if that is undefined
look for a fallack default. This allows any combination of
variables to work (and at the time the anonymous python
executes) safely and get us a properly defined branch for the
fetcher and build.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The recent environment changes mean TOPDIR/BUILDDIR need to be exported
specifcially to the enviromnent so the qemu scripts can find the correct
build directory.
Without this, qemu can fail to run.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
There needs to be a space at the end of DEPENDS_prepend or otherwise it can
collide with the contents of DEPENDS. The OE version of this file has this.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The existing 2.6.34 kernel tree uses _ where poky typically
uses -. This is a historical artifact, since working with
gnu Make and shells means avoiding - is wise. The opposite
is true in Yocto.
To avoid using the _ reserved character wherever possible
we can simply remove it from the branch names in the
new 2.6.37 kernel, but to keep the content stable in the
0.9 2.6.34 kernel, we map _ to - for the purposes of
packaging.
To further faciliate this switch, the branch names no
longer need to be shortened in the KMACHINE mappings, but
can be fully specified and the tools/processing adapt as
required. This gives us the flexibility to map multiple
boards to a single branch for building.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
environment files.
[BUGID #565] Fixing bug #565, added package of
meta-environment-${TARGET_ARCH} for environment files used by
cross-canadian toolchain.
Also corrected the situation of empty config site file for target.
Signed-off-by: Lianhao Lu <lianhao.lu@intel.com>
This fixes the cleanall task to actually delete the sstage cache files. This
was not working correctly before due to attempting to use the current task's
hash in the package file name to be deleted, which of course is the hash of
the cleanall task when doing cleanall. The siginfo files are now deleted as
well.
Signed-off-by: Paul Eggleton <paul.eggleton@intel.com>
Detect and fail if filesystem in use for TMPDIR or SSTATE_DIR has an
unreasonably short file name length limit (eg. eCryptFS). This can cause
"file name too long" errors during poky builds (e.g. when writing sstate
files for packages with a git revision as the version).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Add a mechanism to avoid running more invasive checks every time bitbake is
invoked. Adds a file in conf/ (i.e. under user's build directory) that
tracks a new variable SANITY_VERSION (set in sanity.conf) as well as TMPDIR
and SSTATE_DIR, allowing checks to be run when these are changed.
Fixes [BUGID #562]
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Exclude the DATETIME variable from the KERNEL_IMAGE_BASE_NAME variable.
This avoids the last road block linux-yocto for sstate.
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Fixes [BUGID: 594]
In imagetest-qemu.bbclass, we depend on DISPLAY to start qemu in
a X desktop. When user wants to run qemu sanity test, we need to
check if the variable is set.
Signed-off-by Jiajun Xu <jiajun.xu@intel.com>
An investigation into undocumented variables uncovered a mispelled
instance of PARALLEL_MAKE in image-swab.bbclass.
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
In order to extend and create more kernel recipes based on the
supported yocto kernel common routines need to be placed in
re-usable blocks.
To accomplish this meta/recipes-kernel/linux/linux-yocto_git.bb
is broken into three parts:
- meta/classes/kernel-yocto.bbclass: contains common routines
for checking out and configuring a yocto kernel git repository.
This should be inherited by recipes that need this functionality.
- meta/recipes-kernel/linux/linux-yocto.inc: Contains the machine
mappings, compatibility, build directives and common task
definitions for a yocto kernel based recipe. This inherits
kernel-yocto, and is the typical point of entry for other recipes.
- meta/recipes-kernel/linux/linuux-tools.inc: tasks and function definitions
for kernel recipes that want to build/export perf
It also updates the linux-yocto recipe to default to 2.6.37.
As part of the update to 2.6.37 the branch naming and conventions
have been modified to show inheritance, and be more generic.
For example:
master
meta
yocto/base
yocto/standard/arm_versatile_926ejs
yocto/standard/base
yocto/standard/beagleboard
yocto/standard/common_pc/atom-pc
yocto/standard/common_pc/base
yocto/standard/common_pc_64
yocto/standard/fsl-mpc8315e-rdb
yocto/standard/intel_atom_z530
yocto/standard/intel_core_qm57_pch
yocto/standard/mti_malta32_be
yocto/standard/preempt_rt/base
yocto/standard/preempt_rt/common_pc
yocto/standard/preempt_rt/common_pc_64
yocto/standard/preempt_rt/intel_atom_z530
yocto/standard/preempt_rt/intel_core_qm57_pch
yocto/standard/qemu_ppc32
yocto/standard/routerstationpro
In this structure:
master: tracks the mainline kernel
meta: meta information for the BSPs and kernel features
yocto/base: baseline kernel branch
yocto/standard/base: 'standard' kernel, contains features
and configs for all BSPs
yocto/standard/<machine>: represents a BSP with specific
features or configurations
The tools, tree and libc-headers have all been updated to
deal with this new structure. Also in addition to dealing with
the new structure, they continue to work with the existing
tree and will adapt at runtime to the differences.
The linux-yocto-stable_git.bb recipe continues to build the
2.6.34 based tree,and linux-yocto_git.bb builds 2.6.37. As
boards are enabled for the new kernel they will move from
-stable to the development kernel. As of now, only the
emulated targets have moved to 2.6.37-rcX
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Becaues of the error library path definition in cpan-base.class, libxml-parser-perl package is empty.
Signed-off-by: Jingdong Lu <jingdong.lu@windriver.com>
This patch massively simplifies the canadian packaging and allows
multiple toolchain targets to be parallel installed into the same
nativesdk sysroot without package name conflits. Since we now do
this, we can simplify cross-canadian to become more like nativesdk.
This is a first pass over this task, similar changes would be
desireable to cross and the whole MULTIMACH_ARCH mess can then
probably be similified much further.
Signed-off-by: Richgard Purdie <rpurdie@linux.intel.com>
Using TRANSLATED_TARGET_ARCH instead of TARGET_ARCH for
cross-canadian packages. This is due to the TARGET_ARCH of x86_64
would results incorrect packaging in cross-canadian packages.
The pacakge name appendix of x86_64 target in cross-canadian
packages is x86-64.
Signed-off-by: Lianhao Lu <lianhao.lu@intel.com>
One bitbake invocation can cause multiple images to be generated. We should test
each one and we cna do this by running after the rootfs task. Running the tests
standlone is also still possible with a new separate standlone task.
Acked-by: Scott Garman <scott.a.garman@intel.com>
Acked-by: Jiajun Xu <jiajun.xu@intel.com>
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
We default to gnome-terminal for TERMCMD (and TERMRUNCMD) so should check that
this program actually exists during sanity checking.
As a corollary document how to change these variables in the local.conf
Signed-off-by: Joshua Lock <josh@linux.intel.com>
OEROOT isn't used outside the init script so stop exporting it into the environment
where it can "corrupt" sstate packages.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
Most of the d.keys() used in file parsing are variables in
distro_tracking_fields.inc, which are not used in normal build.
Therefore remove the inclusion of distro_tracking_fields.inc from
poky.conf. Besides, move distro related tasks to distrodata.bbclass,
which includes that tracking field file.
By this change, the file parsing time could save about 25%.
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Or else absolute paths may be expanded which then breaks sstate signature
comparison.
I'm not sure whether pre/post funcs can be expanded or not. They are
invoked at image build or on the target, and thus need be expanded some
place. But if one recipe does use directories under TMP in its postinst
helpers, this would be an issue again.
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Remove the temp sstate-build-* directoies.
Take poky-image-minimal as an example, this saves about 17% disk space.
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
scp test is to check if file copying via network work or not in target.
shutdown test is to check if target can be poweroff with qemu process off.
Signed-off-by Jiajun Xu <jiajun.xu@intel.com>
Timestamp is kept in UTC
Remove superfluous 'create_etc_timestamp()' function
- seems to be a duplicate of 'rootfs_update_timestamp()'
Remove External function reference
[sgw@linux.intel.com: merged 2 patches and cleanup commit message]
Signed-off-by: Gary Thomas <gary@mlbassoc.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
The custom uImage produced by the kernel classes by default
does not universally boot and working with existing uboot
configurations and setups.
This changes the preference to the uImage constructed by
the kbuild processes and falls back to the other techniques
if uImage is requested by the kernel does not produce one.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Add a test case for error log check with command dmesg in target. The
case introduces a new folder in target, "/opt/test", which holds test
scripts running in target.
Signed-off-by Jiajun Xu <jiajun.xu@intel.com>
package sstate requires a lock under sysroots/, which however may not be
created when sstate_setscene functions are executed and then causes failures.
here we make sure 'sysroots' created before do_package_setscene is executed.
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
The fetcher has special handling for file:// mirror urls, being efficient and
just providing an updated path. Unfortunately the sstate fetching code wasn't
able to handle this. This patch detects this and injects a symlink to ensure
everything works. It also fixes some datastore references to be correct and
ensures the sstate download directory exists if it doesn't already.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
If sstate was used to accelerate a build, the pseudo directory might not have
been created leading to subsequent task failures.
Also, sstate packages were not being installed under pseudo context meaning
file permissions could have been lost.
Fix these problems by creating a FAKEROOTDIRS variable which bitbake ensures
exists before running tasks and running the appropriate setscene tasks under
fakeroot context.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
If a sstate package exists for the package task but not for the rpm packaging
task, the output from the package task will be used. The directory pointed
to by dvar will not exist under this scenario.
Since the directory is never used by the packaging process remove the
check, substituting the pkgd variable which is always present and used.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
The message "WARNING: Function do_build doesn't exist" doesn't look professional,
so fix the underlying problem even if this warning is harmless.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
When installing a sstate package, directories tracked by plaindirs were being installed
to the incorrect location. With the current implementation this was limited to
the do_package task.
This patch ensures plaindirs tracked files are created in the correct location, fixing
the bug where these files would go missing.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
This change didn't work as expected as ld.so.cache doesn't preseve any search
path ordering, therefore revert the change.
This reverts commit b2c3ce6d17.
Fix the library search path so that libs in the toolchain in /opt/
have priority over those in base system but ones from the base system
can still be found.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
This class needs to wipe out the contents of the pseudo database but also
ensure the pseudo directory exists for any subsequent tasks and also ensure
any pseudo server has shut down before removing the database. This patch
does all of these things.
[BUGID #222]
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
[BUGID #281]
Add the ability for the deb, ipk and rpm classes to use the new summary
and description fields. The Description is wrapped around 75 characters
to ensure a reasonably nice, presentable description.
(Summary defaults to the description if Summary is not defined.)
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
This addresses [BUGID #410], require implict action by someone to enable
functionality that may have license implications.
By default this diables mp3 and mpeg decoding
Signed-off-by: Saul Wold <Saul.Wold@intel.com>
Along with qemu this provides another option to generate locales for
[e]glibc. The new method is to use cross-localedef with appropriate arch
specific parameters.
The cross-localedef method is found to be 15 times faster than qemu's
emnualted method.
LOCALE_GENERATION_WITH_CROSS-LOCALEDEF : This is new config variable
introduced to selet qemu or cross-localedef method for locale generation.
Thanks to Mark Hatle from Windriver for providing the rich
information for cross locale generation.
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
[BUGID #376]
Update to the insane class to detect the new gcc and binutils poison
messages located within the config.log. (Continue to scan for the old
style message.)
Add a new compile and install log check to the package_qa that scans the
logs for the same types of messages "unsafe for cross-compilation".
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
[BUGID #379]
In a strange configuration, siteconfig could have already generated it's
directory, but decided it needs to regenerate the site config. The lack
of -p in the mkdir causes a failure.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
[BUGID #327]
RPM attempts to validate all of the directories on the system are
owned by a package. While the root "/" directory was not owned, so
in some cases an error could be thrown. Resolve this by informing
RPM that yes, "/" is in fact a directory on the system that was
properly constructed and is "owned".
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
This reverts commit 138df217ef.
We still need the vm_mmap_min_addr set to 0 to run locale generation for
qemu-arm
Signed-off-by: Joshua Lock <josh@linux.intel.com>
The directory list in the manifest file needs to be sorted such that child
directories are deleted first. Fortunately as the list is generated by walking
the directory tree achieving this is as simple as reversing the directory list.
Fixes [BUGID #269]
Signed-off-by: Joshua Lock <josh@linux.intel.com>
With new batch ability, we can use below simple 2 lines for multiple alternatives
updates:
ALTERNATIVE_LINKS = "${bindir}/cmd1 {sbindir}/cmd2 ..."
ALTERNATIVE_PRIORITY = "100"
Then for each command "/path/cmd" listed in ALTERNATIVE_LINKS, below is done
automatically:
${D}/path/cmd is renamed to ${D}/path/cmd.{PN}
a new alternative named 'cmd' is created which:
links /path/cmd to /path/cmd.{PN} with priority specified in ALTERNATIVE_PRIORITY
This way the recipe with multiple alternatives could be simplified a lot. There
are still some cases where above assumptions may break, but I expect more recipes
should benefit from this simple enhancement
Fix [BUGID #257]
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Fix PKG_CONFIG_SYSROOT_DIR and PKG_CONFIG_PATH to make them point
to the correct directory path.
This fixes [BUGID #340].
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
siteconfig: use SITECONFIG_SYSROOTCACHE and use sstate-interceptfuncs
Use the new sstate-interceptfuncs functionality to interpose do_siteconfig
between the sstate_install and package.
Signed-off-by: Jeff Polk <jeff.polk@windriver.com>
Adds the ability to impose intercept function call(s) after
sstate_install and before sstate_package in sstate_task_postfunc.
Signed-off-by: Jeff Polk <jeff.polk@windriver.com>
This solution is what Mark Hatle recommended.
To disable pseudo, while already running, you need to set:
PSEUDO_RELOADED=YES, and then exec something...
This causes pseudo to disable itself from LD_PRELOAD, and thus fall
out of memory on the exec.
This Fixes [BUGID #226]
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
The exit 0 in the workaround seems to have causes the filesystem creation
process to stop. Instead change it to an if that can never succeed.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
This change allows the IMAGE_OVERHEAD_FACTOR (default to 1.2) to
adjust dynamicly how much space to add for each image, thus there
is 20% overhead space by default.
Signed-off-by: Saul Wold <Saul.Wold@intel.com>
By defining the perf package in the shared kernel class bitbake will be able to
find the package provider even when the preferred kernel doesn't build perf,
preventing dependency resolution failures.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Export the location of the native sysroot for use in the scripts to
determine the location of native binaries.
Also add a POKY_DISTRO_VERSION to the external toolchain environment as a
method of determining if we're running under a build directory or not
i.e. not if the variable is empty
Signed-off-by: Joshua Lock <josh@linux.intel.com>
In order to resolve a host-contamination problem, we re-work the way that
autoconf and friends are invoked during the compilation of RPM.
This has a side effect of fixing another bug where RPM was being renamed
HOST_ARCH-HOST_OS-rpm. So we remove the "fixes" for that behavior as well.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
With the new checksum changes, the system is no longer able to determine
if the function resolve_package is used. So it doesn't get emitted. The
workaround resolves this by triggering an exit 0, followed by a bogus usage
of the function.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Richard Purdie
Dongxiao Xu
Dexuan Cui
Jiajun Xu
Koen Kooi
Beth Flanagan
Lianhao Lu
Nitin A Kamble
Bruce Ashfield
Saul Wold
Darren Hart
Yu Ke
Richard Purdie
Paul Eggleton
Paul Eggleton
Alex deVries
Kevin Tian
Jingdong Lu
Joshua Lock
Gary Thomas
Qing He
Saul Wold
Koen Kooi
Mark Hatle
Jeff Polk