Unlike the big jump in version (4 -> 4-6), blktool is actually quite stable since
version 4 (2005). All the changes in 4-6 are about typo fix and debian package
control enhancement.
Fixed metadata accordingly.
Signed-off-by Kevin Tian <kevin.tian@intel.com>
- update metadata. license checksum is skipped, as pax-utils is depended by do_package task itself.
I.e. the 1st do_package will be run after pax-utils-native is built. So far checksum verification
is included in do_package task
- no need to override {S} since {S} uses {BPN}
- no need to define "_GNU_SOURCE" which has been included by source
tar ball. Also remove do_compile override accordingly
- remove recipe internal md5sum, which is not used by Poky
Signed-off-by Kevin Tian <kevin.tian@intel.com>
Run udhcpc results in
udhcpc (v1.15.3) started
/etc/udhcpc.d/50default: line 37: syntax error: unexpected "done" (expecting "fi")
run-parts: /etc/udhcpc.d/50default exited with code 2
Signed-off-by: Enric Balletbo i Serra <eballetbo@gmail.com>
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
VMware SVGA driver needs to have same depth between the host and the guest. Or put in
other word, the depth read by the guest is the value read from host. The guest is not
allowed to change virtual depth to other value. With DefaultDepth option xorg.conf,
vmware driver rejects to work with suggestion "Please do not specify a depth on the
command line or via the config file".
Signed-off-by Kevin Tian <kevin.tian@intel.com>
VMware SVGA presents to the guest with the depth of the host surface it renders
to, and rejects to work if the two sides are mismatched. One problem is that
current VMware VGA may calculate a wrong host depth, and then memcpy from virtual
framebuffer to host surface may trigger segmentation fault. For example, when
launching Qemu in a VNC connection, VMware SVGA thinks depth as '32', however the
actual depth of VNC is '16'. The fault also happens when the host depth is not
32 bit.
Qemu <4b5db3749c5fdba93e1ac0e8748c9a9a1064319f> tempts to fix a similar issue, by
changing from hard-coded 24bit depth to instead query the surface allocator
(e.g. sdl). However it doesn't really work, because the point where query
is invoked is earlier than the point where sdl is initialized. At query time,
qemu uses a default surface allocator which, again, provides another hard-coded
depth value - 32bit. So it happens to make VMware SVGA working on some hosts,
but still fails in others.
To solve this issue, this commit introduces a postcall interface to display
surface, which is walked after surface allocators are actually initialized.
At that point it's then safe to query host depth and present to the guest.
Signed-off-by Kevin Tian <kevin.tian@intel.com>
On an F13 host with glib-config installed pkgconfig-native can get into a
horrible state with recursive calls between pkg-config and glib-config.
The patch adds a configure time option to disable legacy script support in
pkgconfig and makes use of the option for Poky.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
cross-canadian packages need to look for their SOLIBS in the nativesdk
sysroot so that dependencies are correctly picked up and meta-toolchains are
correctly built.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Due to different stack contents in sdl_display_init on Ubuntu vs other distros,
an uninitialized structure is causing a crash. Zeroing the structure makes the
behavior uniform across distros, avoiding the Ubuntu crash, but doesn't fix the
underlying bugs, notably:
the return value of SDL_GetWMInfo needs to be checked, as it's currently
failing silently
the underlying reason for the failure of SDL_GetWMInfo needs to be found -
there is a GetWMINfo method in the internal SDL structure which is NULL, and
the reason for this needs to be found.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Pull time.h patch from upstream Linux kernel
(commit 38332cb98772f5ea757e6486bed7ed0381cb5f98)
The patch fixes the following build failure:
LD .tmp_vmlinux1
kernel/built-in.o: In function `timespec_add_ns':
undefined reference to `__aeabi_uldivmod'
kernel/built-in.o: In function `do_gettimeofday':
undefined reference to `__aeabi_uldivmod'
undefined reference to `__aeabi_uldivmod'
kernel/built-in.o: In function `timespec_add_ns':
undefined reference to `__aeabi_uldivmod'
undefined reference to `__aeabi_uldivmod'
kernel/built-in.o: more undefined references to `__aeabi_uldivmod'
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
include/scsi/scsi.h is not userland parsable and research indicates this is
because the header should not be exposed to userspace. Therefore remove it
in the install.
Research done by Tom Rini <tom_rini@mentor.com> in OE commit
91d3d92a626da89dfe13d63e68a90dbafdbaef1d
This has been the case since kernel 2.6.31
Bump glibc and uclibc PR's so that users have sane <scsi/scsi.h>
Signed-off-by: Joshua Lock <josh@linux.intel.com>
use NO_TCP_WRAPPER=1 in EXTRA_OEMAKE to disable -lwrap
also remove the redundent oe_compile()
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
update license fields
adjust ordering
change to use new patch parameter
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
update homepage and license
change to use new patch parameter
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
update homepage
change license to reflect GPL|BSD dual licensing of the header
update description to reflect the legacy version
change to use new patch parameter
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
One tricky issue for license part. This package doesn't ship with its own COPYING file and autotools will install a GPLv2 one instead of the actual MIT-style license here. So create a MIT-style license on the fly to reflect the fact.
Use new patch parameters.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Saul Wold <Saul.Wold@intel.com>
2.6.33 removed the Hayes ESP driver. The presence of these ioctls
makes setserial believe that ESP support should be built in, breaking its
build.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Fedora 13 switched the default behaviour of the linker to no longer
indirectly link to required libraries (i.e. dependencies of a library
already linked to). Therefore we need to explicitly pass the depended on
libraries into the linker for building to work on Fedora 13.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Back to commit ea45876d7b, LD_LIBRARY_PATH
is disable for cross-build, however it's required for native version. So
force noldlibpath.patch for non-native case only
Signed-off-by Kevin Tian <kevin.tian@intel.com>
Added a entry for bugtracker.
updated LICENSE
updated LIC_FILES_CHKSUM
The x86 build of binutils is working fine.
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
We need to ensure the data directory (/usr/lib/opkg) is created and shipped in
the package as it's used by opkg to create a lock file.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Phenomena: there is udev warning in boot process
tar: can't open '/etc/dev.tar': Read-only file system
The reason is that the init script /etc/rcS.d/S04udev will try to tar the /dev as cache to speed up udev at next boot time. Unfortunately, S04udev is too early and the filesystem is not writable yet.
To fix it, this patch split the cache action to another init script, and register it as /etc/rcS.d/S36, which is after the S35mountall, and the filesystem is already writable.
Signed-off-by: Yu Ke <ke.yu@intel.com>
* Use the -nostdinc++ to CXX fixing libstdc++
* Generate libgcc in gcc-cross, save the result and use in gcc-runtime
* Fix the layout of the crt*.o files so the SDK compiler can find them
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
Added usb tablet options to poky-qemu-internal script, and adjusted the
xorg.conf script for x86 to use VGA screen and tablet input device
Signed-off-by: Saul Wold <saul.wold@intel.com>
Renamed the recipe to 0.12.4.
Moved the patches directory to qemu-0.12.4 - when it was named
qemu-0.12, bitbake silently ignored the patches, and only failed when
it came time to apply them.
Added a FILESDIR to the git recipe and changed its PV to 0.12.4 so it
can use the same patches as the normal recipe for now.
Removed the series file - it wasn't needed.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
--disable-libuuid is added to the configure command line. This has
the effect of disabling the compilation of uuidgen and libuuid.so.
These two are specified as making up their own packages, which turn
out to be empty, and thus aren't created in the first place. The
error which manifests itself is at the final creation of the rootfs
when the global package depends on the e2fsprogs-uuidgen and libuuid,
which turn out not to exist, because they weren't created.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Move the populate-volatile.sh call to below the offline check.
Otherwise, it will try to run during the build, on a system which
likely doesn't have populate-volatile.sh.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
When building the native version, I get errors from ld complaining
about not being able to find a static libc.
Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Needed to increase version to one which includes libuuid and change do_install
for building a native version.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
e2fsprogs requires libuuid to build and we have disabled its built in version
in favour of the one in util-linux, therefore we need util-linux in the DEPENDS
Drop old configure option --enable-dynamic-e2fsck
Signed-off-by: Joshua Lock <josh@linux.intel.com>
This recipe has different names in the different distros. So defining
these aliases for the distro_check task use.
Signed-Off-By: Nitin A Kamble <nitin.a.kamble@intel.com>
The DATE variable in this bb files was distrubing the DATETIME variable
used by the distro_check task.
Renamed the DATE variable to SRCDATE
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
* The no-hardlinks patch is no longer required
* There was an autoconf macro issue which required a new patch (details in patch)
* libuuid was disabled, this should be provided by util-linux on modern systems
* Not providing libuuid is useful for the -native tools as it stops library symbol
confusion with those from the host system.
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
The patch.bbclass changes default to applying listed patches without an
apply/patch parameter. Set the apply parameter to no.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
The import from OE had included some mdev configuration, whereas Poky switched
to udev and a static (tarred) dev some time ago.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
qemu-native was failing to link on my 64bit Fedora 13 machine with this error:
| /usr/bin/ld: libqemu.a(helper_opengl.o): undefined reference to symbol 'dlsym@@GLIBC_2.2.5'
| /usr/bin/ld: note: 'dlsym@@GLIBC_2.2.5' is defined in DSO /lib64/libdl.so.2 so try adding it to the linker command line
| /lib64/libdl.so.2: could not read symbols: Invalid operation
So I did as the linker told me and added -ldl to the linker flags
Signed-off-by: Joshua Lock <josh@linux.intel.com>
There's no need to carry around two copies of the same patches for qemu-git and
qemu-0.12 so drop the qemu-git directory and update the git recipe to use the
qemu-0.12 directory for patches.
Move common code from the two recipes to an inc file.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
fweh
Add kernel headers to match our preferred/default kernel version and set them
as the preferred version for the Poky and Moblin distributions.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Use the same kernel version as the moblin distro so that we provide a similar
kernel for all QEMU machines and the netbook machine.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
The different kernel recipes encapsulate functionality groups for machines,
therefore it makes sense to have all the QEMU machines using the same kernel
recipe.
Switch the QEMU machines to default to the "linux" recipes for their kernel
and bump the latest recipe from linux-2.6.32 to 2.6.33.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Fix various kernels to build with our toolchain, this includes well known fixes
for:
* sumversion.c: compilation failing with a 'PATH_MAX' undeclared (fixed by
adding limits.h to sumversions includes
* a patch taken from oe.dev to stop GCC >= 4.3 from optimizing a loop which
causes compilation to fail
* Fixing the KERNEL_OUTPUT for mx31 and nokia800 kernel recipes
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Added a new variable in recipe : LIC_FILES_CHKSUM
It is a required field for every recipe.
It describes license text location in the source files. And also stores
md5sum of that license text. Any change in this license text triggers build
error. Which enables developer to review any changes in the license and
update the license fields in the recipe accordingly.
For Example: contents of zlib_1.2.3.bb
LICENSE = "zlib"
LIC_FILES_CHKSUM = "file://README;md5=ae764cfda68da96df20af9fbf9fe49bd \
file://zlib.h;beginline=1;endline=30;md5=6ab03f03a5ee92d06b809797d4d5586d "
Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Apply a patch from upstream that fixes the build, patch should be able to be
dropped when we update openssl version.
Signed-off-by: Joshua Lock <josh@linux.intel.com>