| Age | Commit message (Collapse) | Author |
|---|
|
Binary locale generation fails on openSUSE 11.2 and probably would fail
on some other hosts with 2.6.31+ kernels since mmap-ing page zero may be
forbidden for security reasons even with /proc/sys/vm/mmap_min_addr set
to 0 (this also affects Ubuntu 9.10, although it's been fixed there:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/423513
).
This is not a problem for recently introduced qemu 0.12.3 since it can
relocate binaries now, which is tested on openSUSE 2.6.31 kernel with
mmap_min_addr=65536 and Debian lenny 2.6.26 with mmap_min_addr set to
0, 4096, 65536.
Given that, this check is just useless. Unfortunately, it also means
that for reliable builds we have to remove qemu 0.10 which can't work
with some kernels. Patch will follow to do that.
Signed-off-by: Roman I Khimov <khimov@altell.ru>
Acked-by: Tom Rini <tom_rini@mentor.com>
Acked-by: Koen Kooi <koen@openembedded.org>
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
Signed-off-by: Chris Larson <chris_larson@mentor.com>
|
|
|
|
The postinsts for kernel-image and modules run "depmod -a" on the
target, but this only updates the old kernel's modules.dep.
"depmod -a ${KERNEL_VERSION}" updates the files in
/lib/modules/${KERNEL_VERSION}.
Signed-off-by: Michael Smith <msmith@cbnco.com>
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
I ran into a problem yesterday where the ROOTFS_POSTPROCESS_COMMAND
started failing after I turned off ONLINE_PACKAGE_MANAGEMENT. It seems
the problem is that if package management is turned off, then the opkg
directory gets deleted. Subsequent opkg commands in the
ROOTFS_POSTPROCESS_COMMAND fail to open the lock file because the
directory is gone.
This patch simply moves the postprocess command above the destruction of
the opkg directory to allow any such commands to complete successfully.
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
Move the logic to determine what qemu-TARGET to run into qemu.bbclass so
we can check for the right binary in sanity.bbclass. This code was
duplicated by glibc-package and eglibc-package anyhow and with the new fn
we can clean up the usage in these classes a bit. Now that we have a class
for qemu stuff, and the gcc check is just for qemu, move it there.
|
|
- parallel builds need to happen via MOZ_MAKE_FLAGS and it gripes if still
passed -jN, so keep the old value before we clear it.
- Move the HOST_LIBIDL stuff into configure, otherwise bad things happen when
you don't have pkg-config on the build host.
- Prior to 3.6, wireless-tools can be, or not be built already and the Necko
wifi options deals. With how 3.6 is configured, it must be disabled or
DEPENDED on. For consistency, turn it off.
- In 3.5 and newer, libnotify can be used, add it to DEPENDS (could be
disabled).
- Because of both of the above, bump PR.
Signed-off-by: Tom Rini <tom_rini@mentor.com>
Acked-by: Khem Raj <raj.khem@gmail.com>
|
|
glibc-package.bbclass: Add glibc-pic package
By default, in our glibc builds we don't have these libraries. They may
however come from various binary toolchains, or custom source builds.
Signed-off-by: Tom Rini <tom_rini@mentor.com>
Acked-by: Khem Raj <raj.khem@gmail.com>
|
|
Perform qemu-related checks not based on if we're ARM but based on if
we'll be using qemu for binary locale generation. Clarify what the
first of these sanity checks does. Next, change the check for a
provided qemu binary to be generic enough to work on all arches (and catch
distribution or user built versions of qemu). While we're in here, correct
the gcc version check code in base.bbclass.
Signed-off-by: Tom Rini <tom_rini@mentor.com>
Acked-by: Denys Dmytriyenko <denis@denix.org>
Acked-by: Chris Larson <chris_larson@mentor.com>
|
|
need work
Signed-off-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
the magic"
Drop this for now, as apparently the magic file location varies with 'file'
version. I'll just patch file to find it relative to the binary location
instead.
This reverts commit 1b5e7041ae3b26b7e59c76bd2f2fd72e35492940.
|
|
This works around one relocation issue.
Signed-off-by: Tom Rini <tom_rini@mentor.com>
Signed-off-by: Chris Larson <chris_larson@mentor.com>
|
|
(from poky)
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
|
|
commands"
Back this out for the time being, things are exploding now.
This reverts commit df32920678d15c86897b50b752b937210a01edea.
|
|
Signed-off-by: Chris Larson <clarson@mvista.com>
Acked-by: Michael Smith <msmith@cbnco.com>
Acked-by: Tom Rini <tom_rini@mentor.com>
|
|
Patch courtesy the Poky project.
Signed-off-by: Chris Larson <clarson@mvista.com>
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
Acked-by: Tom Rini <tom_rini@mentor.com>
|
|
opkg-cl call with -t parameter
* Use opkg.conf from staging in the same way as do_rootfs does
* When option tmp_dir is used in opkg.conf installed on rootfs then it's
used also in do_rootfs call and points to probably non-existent directory
on buildhost like /var/lib/opkg/tmp.
* The value of tmp_dir from rootfs is used even with another config file
specified with -c parameter
* Before this, it was using default value (/tmp) on buildhost, now it will use
own "${IMAGE_ROOTFS}-tmp" and remove it after do_rootfs finish
(usually already empty inside, cleaned by opkg itself)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Acked-by: Denys Dmytriyenko <denis@denix.org>
Acked-by: Koen Kooi <koen@openembedded.org>
|
|
* When option tmp_dir is used in opkg.conf installed on rootfs then it's
used also in do_rootfs call and points to probably non-existent directory
on buildhost like /var/lib/opkg/tmp.
* The value of tmp_dir from rootfs is used even with another config file
specified with -c parameter
* Before this, it was using default value (/tmp) on buildhost, now it will use
own "${IMAGE_ROOTFS}-tmp" and remove it after do_rootfs finish
(usually already empty inside, cleaned by opkg itself)
* Similar patch for testlab.bbclass will follow
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
|
Signed-off-by: Chris Larson <chris_larson@mentor.com>
|
|
the magic"
Forgot to check a git status / log before pushing :)
This reverts commit bd711cfab09394b6f3064eaed24b8761edc19f19.
|
|
This works around one relocation issue.
Signed-off-by: Tom Rini <tom_rini@mentor.com>
Signed-off-by: Chris Larson <chris_larson@mentor.com>
|
|
|
|
* remove config.rpath stuff from recipes
|
|
* no need for DONT_CHECK_KERNELSIZE
* clean up the affected files
* check happens only if KERNEL_IMAGE_MAXSIZE is set
|
|
STAGING_DIR_HOST is "" when using native bbclass:
native.bbclass:STAGING_DIR_HOST = ""
But in autotools.bbclass there is:
sed -i -e s:${STAGING_DIR_HOST}::g $i
which result in the following code in run.autotools_prepackage_lamangler
sed -i -e s:::g $i
which makes libxml2-native fail like this:
| sed: -e expression #1, char 0: no previous regular expression
I discussed it on IRC:
Feb 11 17:41:29 <RP> GNUtoo: We should only be doing that is STAGING_DIR_HOST isn't empty I guess
So I made a temporary fix which only workarround,because some issues persist:
Feb 11 17:41:02 <pb_> RP: heh, I wonder whether paths with colons in would be considered valid :-}
Signed-off-by: Denis 'Gnutoo' Carikli <GNUtoo@no-log.org>
|
|
* config.vapi is per-app and can be considered as part of the autotools stuff
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
|
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
|
|
There is no point in telling users to move TMPDIR "back" to its current location :-}
|
|
We need to add ${S}/${KERNEL_OUTPUT} with package_stagefile_shell to
avoid playing more complex games with the install line. We should
also use package_stagefile_shell on the modules tarball as that will
not otherwise be reproduced.
|
|
${SYSROOT_DESTDIR} ends on a / but there is also a / after the variable
and a slash as the first character of the next variable.
This patch removes the / that is explicitly added so in the generated
code we get // instead of the even sillier looking (and probably more
confusing) ///
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
Accidently tagged another file to this commit, sorry
This reverts commit b977bbd442f3d00c884b68c1307bd9b14fb96446.
|
|
${SYSROOT_DESTDIR} ends on a / but there is also a / after the variable
and a slash as the first character of the next variable.
This patch removes the / that is explicitly added so in the generated
code we get // instead of the even sillier looking (and probably more confusing) ///
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
This file was not updated when DISTRO_PR was introduced.
Result of it was that packages were not properly added to the packaged staging package
(due to faulty filenames being generated).
This patch adds ${DISTRO_PR} to the filenames involved
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
* use += instead of =
this appends new file paths to include instead of replacing the
current definition
Signed-off-by: Rolf Leggewie <oe-devel@rolf.leggewie.biz>
Acked-by: Graeme Gregory <dp@xora.org.uk>
Acked-by: Martin Jansa <Martin.Jansa@gmail.com>
Acked-by: Michael 'Mickey' Lauer <mickey@vanille-media.de>
|
|
When using base_contains() to check for a string in a variable for a,
if the variable is not defined an exception occurs. By checking the
existance of the variable and returning false if it isn't there,
a value can be checked for a variable regardless of whether or not it
is defined.
Signed-off-by: Brian Pomerantz <bapper@mvista.com>
Signed-off-by: Chris Larson <clarson@mvista.com>
Acked-by: Khem Raj <raj.khem@gmail.com>
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
Signed-off-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
from packaged staging"
This reverts commit 61acb940a7436a43d0894284d25e8d106f461c23.
This patch is not as robust as I thought, will reassess the issue at hand later
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
packaged staging
kernel.bbclass do_deploy has:
install -m 0644 ${KERNEL_OUTPUT} ${DEPLOY_DIR_IMAGE}/${KERNEL_IMAGE_BASE_NAME}.bin
which expands to
install -m 0644 arch/arm/boot/uImage .../deploy/glibc/images/beagleboard/uImage-2.6.29-r51-beagleboard.bin
but in packaged staging arch/arm/boot/uImage does not exist so we get the error:
install: cannot stat `arch/arm/boot/uImage': No such file or directory
This patch only installs uImage if not using packaged staging
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
Signed-off-by: Denys Dmytriyenko <denis@denix.org>
Acked-by: Khem Raj <raj.khem@gmail.com>
Acked-by: Frans meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
-native and -sdk cases as well
|
|
|
|
New checksums system needs a name= in the SRC_URI name, so add a default
one for gnome.
|
|
to package we don't read potentially stale pkgdata from staging.
This is basically the commit 27b9bc946eeb3faa7b10d92689ab527af42727bd that got
lost when merging poky stuff into the repo. The behavior is much like
what we already have with package_ipk.
Signed-off-by: Ricardo Salveti de Araujo <ricardo.salveti@openbossa.org>
|
|
icecc-blacklisted packages currently built not only without icecc environment
set but also without any parallel options enabled. The original intention was
not to kill the build machine, I guess, since having a good icecc cluster it's
easy to set '-j16' or '-j32' or whatever else which could do so. However as
the toolchain packages are blacklisted for icecc they're becoming a
bottleneck in the build process, while they can be built with local-machine
parallelism.
So, in order to solve it, introducing ICECC_FALLBACK_PARALLEL option. It's
simply used as PARALLEL_MAKE for icecc-blacklisted packages.
|
|
Building things with icecc and BB_NUMBER_THREADS more then 1 you could see
a lot of messages:
ERROR: no cross compiler built yet?
before gcc-cross would be properly built. There is technically nothing wrong,
cross compiler is not there yet and it's not needed when bitbake does
fetch-unpack-patch and other things to prepare package build. But it makes
people uncomfortable.
We can safely kill this messages.
|
|
comprehensible
Signed-off-by: Chris Larson <clarson@mvista.com>
|
|
Signed-off-by: Chris Larson <clarson@mvista.com>
|
|
Signed-off-by: Chris Larson <clarson@mvista.com>
|
|
* Apply the same change as c0a3e91845fad1cdda723bdcf962de8bc01f8a2d for
x86 path
* A check was missed in the x86 path of the stage method.
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
|
|
* With 2.6.33-rc2-00252-ge9e5521 on my host I noticed that
"cat /proc/sys/vm/mmap_min_addr" returns now
"cat: /proc/sys/vm/mmap_min_addr: Operation not permitted"
Its probably becuse of
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=0e1a6ef2dea88101b056b6d9984f3325c5efced3
But I'm not sure if checking CAP_SYS_RAWIO even for reading this value
is intentional or just bug which should be fixed in kernel.
This patch makes bitbake ignore reading error. Possible QEmu failure is not
handled yet.
Acked-by: Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>
|
|
Without that patch uicmoc3-native fails to build.
|
Roman I Khimov
Chris Larson
Bernhard Reutner-Fischer
Michael Smith
Chris Conroy
Tom Rini
Marcin Juszkiewicz
Chris Larson
Richard Purdie
Christopher Larson
Ross Burton
Martin Jansa
Koen Kooi
Koen Kooi
Andrea Adami
Denis 'Gnutoo' Carikli
Phil Blundell
Frans Meulenbroeks
Rolf Leggewie
Brian Pomerantz
Denys Dmytriyenko
Michael 'Mickey' Lauer
Graeme Gregory
Ricardo Salveti de Araujo
Aleksey Makarov
Marcin Juszkiewicz