Richard Purdie [Wed, 8 Jul 2015 11:56:44 +0000 (12:56 +0100)]
pseudo: Exclude SITEINFO_BITS from checksums
We really want the same sstate checksums for pseudo-native on 32 and 64 bit platforms
but the use of SITEINFO_BITS prevents this. Since other things would change if
the bit size changes, we can safely exclude this variable and rely on others
(e.g. BUILD_ARCH included in WORKDIR) to handle this.
[YOCTO #5970]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Wed, 8 Jul 2015 11:56:13 +0000 (12:56 +0100)]
oeqa/sstatetests: Add test for comparing 32 and 64 bit sstate checksums
The sstate checksums for both native and target should not vary whether
they're built on a 32 or 64 bit system. Rather than requiring two different
build machines and running a builds, override the variables calling uname()
manually and check using bitbake -S.
[YOCTO #5970]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Robert Yang [Mon, 6 Jul 2015 10:00:35 +0000 (03:00 -0700)]
file: fix long-options
Backport two patches to fix bug with long options:
* 0001-Fix-bug-with-long-options-and-explicitly-number-them.patch
* 0002-fix-bug-with-5.23-long-options.patch
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ross Burton [Mon, 6 Jul 2015 14:19:41 +0000 (15:19 +0100)]
ca-certificates: remove Debianism in run-parts invocation
ca-certificates comes from Debian but not all distros (i.e. Fedora) have a
leaner run-parts that doesn't support the -- separator between options and
paths, which causes this error:
| Running hooks in [...]/rootfs/etc/ca-certificates/update.d...
| [...]/usr/sbin/update-ca-certificates: line 194: Not: command not found
Signed-off-by: Ross Burton <ross.burton@intel.com>
image_types.bbclass: add support for multiple ubi sizes
Add new FSTYPE called 'multiubi'. This new type has the ability to
create multple ubi's, something that's extremely useful on boards that
can include varying flash sizes (e.g. 256MB vs 2GB).
Because we don't want to reuse code, I extracted the mkfs and ubinize
commands out to a new function that will handle the creation of both
multiple ubi's, as well as the 'ubi' FSTYPE (so as not to break current
boards).
Due to the single processes nature of the OE when doing it's
'create_image', this new multiubi_mkfs function also creates it's own
symlinks as well as cleans up it's own cfg file.
Signed-off-by: Pushpal Sidhu <psidhu@gateworks.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
1. Show GNU unique symbols as provided symbols
2. Remove dependency on dpkg
Both have been submitted to mklibs maillist:
https://lists.debian.org/debian-boot/2015/07/msg00018.html
https://lists.debian.org/debian-boot/2015/07/msg00004.html
Signed-off-by: Yuanjie Huang <yuanjie.huang@windriver.com> Signed-off-by: Roy Li <rongqing.li@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Joe Slater [Thu, 2 Jul 2015 09:10:51 +0000 (17:10 +0800)]
security_flags: eliminate FORTIFY_SOURCE for debug builds
If -D_FORTIFY_SOURCE=2 is included in CFLAGS for debug builds,
many warnings will be generated and some packages will fail to
build. So, only conditionally include it.
Signed-off-by: Joe Slater <jslater@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Yue Tao [Fri, 3 Jul 2015 08:53:56 +0000 (16:53 +0800)]
man: replace ',' with '#'
Sometimes, the parameters of CC/BUILD_CC contains the ',', which
cause the sed command failed, so replace the ',' with '#'
Signed-off-by: Yue Tao <Yue.Tao@windriver.com> Signed-off-by: Wenzong Fan <wenzong.fan@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
ldconfig-native: Add 64-bit flag for ELF64 entries
ldconfig-native was grepped from an old version of glibc, and its output
lacks neccessary 64bit flag in entries.
Due to this defect, ctypes.util.find_library() python function fails to
detect any library due to the old file format that ldconfig-native
creates. This fix sets architecture-dependent 64bit flags for 64-bit ELF.
Since the host's elf.h may not have definition for new AArch64 machine
type, a work-around is added to correctly flag 64-bit ARM libraries.
Signed-off-by: Yuanjie Huang <yuanjie.huang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Puustinen, Ismo [Fri, 3 Jul 2015 11:36:15 +0000 (14:36 +0300)]
combo-layer: modified to generate better commit messages.
This patch includes support for a global section in combo-layer.conf
called [combo-layer-settings]. Supported in this section is key
"commit_msg"; its value is the template for the git commit message
that updates the last_revision. The template can include substitution
for the updated component list: ${components}. The substituted value
will either be a comma-separated list of components or "all components",
if combo-layer was invoked without component list argument.
If the key is not present, the old default value is used for the commit
message.
Configuration file example:
[combo-layer-settings]
commit_msg = pulled in the latest changes for ${components}.
Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ross Burton [Fri, 3 Jul 2015 10:18:00 +0000 (11:18 +0100)]
libdrm: remove git recipe, merge inc
The git recipe hasn't been touched since 2012 and is out of date, so simplify
the recipe by removing the git form and merging the remaining bb and inc file.
Signed-off-by: Ross Burton <ross.burton@intel.com>
If 'rc.debug' is not in kernel parameters, the functions script would
exit 1 which causes other init scripts that source it exit 1. This is
not what we want.
[YOCTO #7948]
Signed-off-by: Chen Qi <Qi.Chen@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Zhixiong Chi [Wed, 1 Jul 2015 07:54:21 +0000 (15:54 +0800)]
rpm: check if the argument(rootpath) exists or be writable
When user execute the command "rpm -qai --root=$dir",if $dir doesn't
exist or is unwritable as result of making a typo in rootpath,then
it will create dirent $dir and subdirectory.
So we should add the check function to fix it before creating relational
subdirectory,and warn the incorrect rootpath to user. It just checks the
rootpath reasonableness when the user input the argument(--root=/-r=).
Signed-off-by: Zhixiong Chi <zchi@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Yue Tao [Thu, 2 Jul 2015 07:18:59 +0000 (15:18 +0800)]
image_types_uboot: remove the middle file
The .gz, .bz2 or .lzma are middle files for making gz.u-boot,
bz2.u-boot, or lzma.u-boot. It should be removed once the final
image is generated. Otherwise, even RM_OLD_IMAGE = 1, it can't
be removed, moreover more and more middle files are genrated via
'bitbake *-image'
Signed-off-by: Yue Tao <Yue.Tao@windriver.com> Signed-off-by: Roy Li <rongqing.li@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
This ensures that in cases where the preference value changes when the
multilib override is applied, we correctly expand it in that context.
For example, for `PREFERRED_PROVIDER_${TARGET_PREFIX}gcc
= "gcc-external-cross-${TARGET_ARCH}"`, when it sets the prefixed version of
this, we want TARGET_ARCH expanded with the multilib applied, otherwise the
arch suffix will be incorrect for that context.
We ran into this trying to use preferences in meta-sourcery along with
multilibs. We worked around it there via PNBLACKLIST, but this fix should
still go into the core.
Signed-off-by: Christopher Larson <kergoth@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Saul Wold [Tue, 30 Jun 2015 22:48:02 +0000 (15:48 -0700)]
core-image-kernel-dev: Define a correct subset of TEST_SUITES to run for this image
Since this image can be built with both poky and poky-lsb, ensure the correct
tests are run in both cases, the auto detection does not quite work right in the
poky-lsb case since not all the Posix cmdline utilites are added. Add connman and
connman test for network sanity
Signed-off-by: Saul Wold <sgw@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Martin Jansa [Thu, 4 Jun 2015 12:44:37 +0000 (14:44 +0200)]
squashfs-tools: upgrade to newer revision from git
* 4.3 release has nasty hard to reproduce bug in LZO compression
which in few cases results in one of these 2 errors:
1) Lseek failure when creating the FS:
| writer: Lseek on destination failed because Bad file descriptor, offset=0x1f72306
| FATAL ERROR:Probably out of space on output filesystem
| Parallel mksquashfs: Using 32 processors
| Creating 4.0 filesystem on rootfs.squashfs, block size 262144.
..
2) failing to read the filesystem in runtime, kernel 3.16.0 showing
errors like this:
[ 46.720568] SQUASHFS error: lzo decompression failed, data probably corrupt
[ 46.730003] SQUASHFS error: squashfs_read_data failed to read block 0x1f72305
[ 46.740076] SQUASHFS error: Unable to read fragment cache entry [1f72305]
it's fixed in upstream git repo:
https://github.com/plougher/squashfs-tools/commit/de03266983ceb62e5365aac84fcd3b2fd4d16e6f
https://github.com/plougher/squashfs-tools/commit/9c1db6d13a51a2e009f0027ef336ce03624eac0d
the official repo is:
https://git.kernel.org/cgit/fs/squashfs/squashfs-tools.git
but author's github is kept in sync so we can use just that
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Nicolas Guyomard [Thu, 18 Jun 2015 07:43:04 +0000 (09:43 +0200)]
grub-efi: Allow to use file name different from grub.cfg in GRUBCFG
To generate a same image both in live image type and boot-directdisk image type
and make boot-direct image boot directly on the rootfs partition without using
an initramfs we need to have the ability to generate several grub.cfg files.
Signed-off-by: Nicolas Guyomard <nicolas.guyomard@open.eurogiciel.org> Signed-off-by: Ross Burton <ross.burton@intel.com>
Daniel Istrate [Mon, 29 Jun 2015 09:04:59 +0000 (12:04 +0300)]
oeqa/selftest: Added new testsuite for image features.
Automated 5 oe-selftest testcase:
- 1107: Check if non root user can connect via ssh without password
- 1115: Check if all users can connect via ssh without password
- 1114: Check rpm version 4 support on image
- 1101: Check if efi/gummiboot images can be buit
- 1103: Check that wic command can create efi/gummiboot installation images
Signed-off-by: Daniel Istrate <daniel.alexandrux.istrate@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Daniel Istrate [Mon, 29 Jun 2015 15:31:55 +0000 (18:31 +0300)]
scripts/oe-selftest: Added mechanism for including/removing bblayers.inc
When oe-selftest starts it includes bblayers.inc into bblayers.conf
When oe-selftest ends it deletes bblayers.inc and the included line
from bblayers.conf
Signed-off-by: Daniel Istrate <daniel.alexandrux.istrate@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Daniel Istrate [Mon, 29 Jun 2015 08:59:41 +0000 (11:59 +0300)]
oeqa/selftest: Add methods to manipulate bblayers.conf in base.py
Added methods for manipulating bblayers.conf file in the same manner as local.conf file:
- write_bblayers_config
- append_bblayers_config
- remove_bblayers_config
Signed-off-by: Daniel Istrate <daniel.alexandrux.istrate@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Mariano Lopez [Fri, 26 Jun 2015 05:56:00 +0000 (05:56 +0000)]
report-error.bbclass: Added file syncronization.
errorreport_handler would fail if several errors are
triggered at the same time because of two proccess
writting to the same file. This patch add the required
syncronization to handle concurrent process.
[YP #7899]
Signed-off-by: Mariano Lopez <mariano.lopez@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Richard Purdie [Tue, 7 Jul 2015 12:22:15 +0000 (13:22 +0100)]
oeqa/bbtests: Fix race over DL_DIR and SSTATE_DIR
Running "-c cleanall" on shared DL_DIR and SSTATE_DIR is antisocial.
It leads to hard to debug races where we wonder why files disappear
and reappear from those directories.
Fix this by using a specific set of directories for these tests. This
avoids a long standing bug on the autobuilder where aspell and man
sources would disappear.
[YOCTO #6276]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Andre McCurdy [Tue, 23 Jun 2015 23:27:57 +0000 (16:27 -0700)]
tune-core2.inc: set X86ARCH32 to i686 (instead of i586)
Use i686 as TARGET_ARCH for 32bit core2 (and corei7 and atom) builds.
In most cases, i586 and i686 are equivalent values for TARGET_ARCH, however
one important exception is glibc. When configured for i686, glibc enables
optimised string functions (SSE, SSE2, etc), which are not used when
building for i586.
The benefits of i686 optimised string functions vary depending on the
application and the CPU, however in some cases the improvements are
significant. In one test, a 50% increase in FPS was seen when running the
'smashcat' benchmark [1] in a qtwebkit browser on an Intel Atom based SoC.
The gain seems to comes from a 3x improvement in memcpy performance when
copying graphics buffer lines (5120 bytes, or 1280 x 4 bytes/pixel), from
the CPU to GPU. Note that very large memcpy's (e.g. 32MB) on the same
machine show no particular performance increase between i586 and i686.
[1] http://www.smashcat.org/av/canvas_test/
Warning: The change in TARGET_ARCH means that _i586 architecture specific
over-rides will no longer take effect. Both oe-core and meta-oe have been
updated to replace _i586 over-rides with _x86, however other layers may
still need review and updating.
Signed-off-by: Andre McCurdy <armccurdy@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ross Burton [Wed, 1 Jul 2015 22:04:45 +0000 (23:04 +0100)]
wayland: always build wayland-scanner
By passing --disable-scanner to use a native wayland-scanner binary a target
wayland-scanner binary isn't built, which is a problem if you want to use it on
the target or in a SDK.
Instead, always build a target wayland-scanner binary, and have an option to
control whether that binary or a host-provided binary is used at build time.
[ YOCTO #7931 ]
Signed-off-by: Ross Burton <ross.burton@intel.com>
Khem Raj [Wed, 3 Jun 2015 19:56:23 +0000 (12:56 -0700)]
libinput: Upgrade 0.7.0 -> 0.18.0
Upgrade is also required for wayland/weston 1.8.x
libinput license changes are
- COPYING: note that having linux/input.h in the tree does not make libinput GPL
- Updated to 2015 where appropriate, added where missing.
- filter: add Simon's copyright
This code was largely lifted from the X server in bb25b2ad297891430606c367bfabc but didn't take the copyright messages
that applied to that code.
Martin Jansa [Tue, 30 Jun 2015 14:15:00 +0000 (16:15 +0200)]
feature-arm-thumb.inc: Fix ARMPKGSFX_THUMB value
* my previous thumb related commit:
commit 3e760031f91fb87c3e2f62b77a117eb41164f259
Author: Martin Jansa <martin.jansa@gmail.com>
Date: Wed Feb 18 15:40:35 2015 +0100
feature-arm-thumb.inc: respect ARM_INSTRUCTION_SET when adding thumb
suffix
unfortunately removed conditional on "thumb" in TUNE_FEATURES, when
setting ARMPKGSFX_THUMB
* in case we have MACHINE without "thumb" in TUNE_FEATURES and distro
setting ARM_INSTRUCTION_SET to "thumb" we end with:
ARM_INSTRUCTION_SET="thumb"
ARM_THUMB_OPT="thumb"
ARM_M_OPT="thumb"
Saul Wold [Fri, 26 Jun 2015 18:23:06 +0000 (11:23 -0700)]
rootfs.py: Add check for kernel modules before running depmod
Add a check for kernel modules so we don't un-necessarily run the depmods, this
will also handle the case with linux-dummy does not place the kernel-abiversion
since it also does not have kernel modules.
[YOCTO #7884]
Signed-off-by: Saul Wold <sgw@linux.intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Cristian Iorga [Mon, 29 Jun 2015 15:03:11 +0000 (18:03 +0300)]
runqemu-internal: correctly set format for root FS as raw
qemu guesses via probing the format of root FS, but
gives a warning and restricts write operations on block 0.
Fix it by setting correctly the format as raw for more
machines and non-KVM machines.
In some cases, replaced the way machine disk is set for qemu.
Fix for [YOCTO #7918]
Signed-off-by: Cristian Iorga <cristian.iorga@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Koen Kooi [Wed, 3 Jun 2015 06:11:54 +0000 (08:11 +0200)]
bootimg bbclass: don't hardcode 'bzImage' name
In the EFI case there are 2 more architectures besides x86(_64) that may use this class: aarch32, which tends to use 'zImage' and aarch64, which uses 'Image'. Instead of hardcoding those values like the luvOS fork does, use the proper variable.
Signed-off-by: Koen Kooi <koen.kooi@linaro.org> Signed-off-by: Ross Burton <ross.burton@intel.com>
Chuang Dong [Tue, 30 Jun 2015 09:02:45 +0000 (17:02 +0800)]
ltp: vma03 fix the alginment of page size
the offset the param of mmap2() doesn't align the page size, but,
this param allow must be a multiple of the page size as returned
by sysconf(_SC_PAGE_SIZE).meanwhile offset * 4096 must be a
multiple of the system page size, so modify the input param of offset
pgoff = (ULONG_MAX - 1)&(~((pgsz-1)>>12));
Signed-off-by: Chuang Dong <Chuang.Dong@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ed Bartosh [Mon, 29 Jun 2015 10:30:36 +0000 (13:30 +0300)]
wic: Fix confusing error message
Wic throws this message when any of the build artifacts are
not provided:
Build artifacts not completely specified, exiting.
(Use 'wic -e' or 'wic -r -b -k -n' to specify artifacts)
It was not clear which artifact was not specified.
Reworked the code to specify list of missed artifacts.
Now the message looks like this:
The following build artifacts are not specified:
bootimg-dir, kernel-dir, native-sysroot
[YOCTO #7912]
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
Ed Bartosh [Sat, 27 Jun 2015 14:18:10 +0000 (17:18 +0300)]
wic: Add --uuid partition option
Added --uuid option to the configuration of wks parser.
This option specifies partition UUID. The code to process
it is already in place. It was implemented for --use-uuid
option.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
Ed Bartosh [Sat, 27 Jun 2015 12:08:15 +0000 (15:08 +0300)]
wic: Fix naming conflict
Image file name is not unique for the partitions without label.
This causes image being rewritten and used as a source for all
partitions without label. Wic produces broken or incorrect result
images because of that.
Added wks line number to the image name to make it unique.
Ed Bartosh [Sat, 27 Jun 2015 08:12:50 +0000 (11:12 +0300)]
wic: Turn off debug output for 'bitbake -e'
Switched debug level to 'normal' to prevent huge 'bitbake -e'
output to go into wic debug output. This should help to make
wic debug info much more clean and easier to read.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
Ed Bartosh [Sat, 27 Jun 2015 08:05:19 +0000 (11:05 +0300)]
wic: Remove annoing debug message
Wic tries to find plugins in every layer and prints a message
'Plugin dir is not a directory or does not exist' if layer
doesn't have plugin dir. It causes a lot of duplicated
messages in the debug output, which makes it hard to find
useful info there.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
Ed Bartosh [Fri, 26 Jun 2015 18:27:31 +0000 (21:27 +0300)]
wic: Refactor getting bitbake variables
Wic gets bitbake variables by parsing output of 'bitbake -e' command.
This implementation improves this procedure as it runs 'bitbake -e' only
when API is called and does it only once, i.e. in a "lazy" way. As parsing
results are cached 'bitbake -e' is run only once and results are parsed
only once per requested set of variables.
get_bitbake_var became the only API call. It replaces find_artifacts,
find_artifact, find_bitbake_env_lines, get_bitbake_env_lines,
set_bitbake_env_lines and get_line_val calls making API much more clear.
Signed-off-by: Ed Bartosh <ed.bartosh@linux.intel.com>
Robert Yang [Tue, 30 Jun 2015 03:06:38 +0000 (20:06 -0700)]
bootimg.bbclass:hddimg: check the size of rootfs.img for hddimg
The hddimg uses FAT, so the single file size should be less than 4GB,
otherwise errors, check that and error out.
Another way might be use ext2/3/4 rather than FAT, but EFI only supports
FAT, if we make EFI use FAT, and non-EFI use extX, that would the code
very different, which seems not worth.
[YOCTO #6449]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Robert Yang [Tue, 30 Jun 2015 03:06:37 +0000 (20:06 -0700)]
bootimg.bbclass:iso: use mkisofs -iso-level 3 for large iso
There will be problems when rootfs.img is larger than 4GB:
mkisofs: Value too large for defined data type. File /path/to/iso/rootfs.img is too large for current mkisofs settings - ignoring
And will get a wrong iso.
Check the size of ${ISODIR}/rootfs.img, use mkisofs -iso-level 3
when it exceeds 3.8GB, the specification is 4G - 1 bytes, we need
leave a few space for other files.
[YOCTO #6449]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Li Wang [Mon, 29 Jun 2015 06:22:35 +0000 (14:22 +0800)]
nfs-utils: debianize start-statd
make start-statd command to use nfscommon configure, too.
Signed-off-by: Henrik Riomar <henrik.riomar@ericsson.com> Signed-off-by: Li Wang <li.wang@windriver.com> Signed-off-by: Roy Li <rongqing.li@windriver.com> Signed-off-by: Wenzong Fan <wenzong.fan@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Yuanjie Huang [Mon, 29 Jun 2015 08:07:48 +0000 (16:07 +0800)]
mklibs-native: avoid failure on symbol provided by application
Undefined symbols in a library can be provided by the application
that links to the library, such as `logsink' in libmultipath.so.0.
This fix checks the type of object in which the symbol is needed
and the existence of the symbol in application, when a symbol
cannot be provided by libraries. It prevents false alarm on absence
of symbols.
Signed-off-by: Yuanjie Huang <yuanjie.huang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Robert Yang [Mon, 29 Jun 2015 04:42:29 +0000 (21:42 -0700)]
gdbm 1.8.3: install ndbm.h and dbm.h
There are gdbm 1.8.3 and gdbm 1.11, while 1.8.3 is GPLv2, it should
install ndbm.h and dbm.h as what 1.11 does, to make perl build OK when
non-gplv3 build.
Fixed when perl build: (non-gplv3 buld)
| NDBM_File.xs:18:60: fatal error: gdbm/ndbm.h: No such file or directory
| # include <gdbm/ndbm.h> /* RedHat compatibility version */
| ^
| compilation terminated.
And:
| ODBM_File.xs:8:19: fatal error: dbm.h: No such file or directory
| # include <dbm.h>
| ^
| compilation terminated.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
