Zhenhua Luo [Wed, 12 Sep 2012 08:20:46 +0000 (16:20 +0800)]
valgrind: fix debug info reading error when do memcheck on ppc targets
following is the error message:
--2263-- WARNING: Serious error when reading debug info
--2263-- When reading debug info from /lib/ld-2.13.so:
--2263-- Can't make sense of .got section mapping
--2263-- WARNING: Serious error when reading debug info
--2263-- When reading debug info from /home/root/lzh:
--2263-- Can't make sense of .data section mapping
--2263-- WARNING: Serious error when reading debug info
--2263-- When reading debug info from /usr/lib/valgrind/vgpreload_core-ppc32-linux.so:
--2263-- Can't make sense of .data section mapping
--2263-- WARNING: Serious error when reading debug info
--2263-- When reading debug info from /usr/lib/valgrind/vgpreload_memcheck-ppc32-linux.so:
--2263-- Can't make sense of .data section mapping
--2263-- WARNING: Serious error when reading debug info
--2263-- When reading debug info from /lib/libc-2.13.so:
--2263-- Can't make sense of .data section mapping
Signed-off-by: Zhenhua Luo <b19537@freescale.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bruce Ashfield [Tue, 11 Sep 2012 15:04:41 +0000 (15:04 +0000)]
linux-yocto/3.4: add x32 configuration fragment
When x32 is the tuning for a x86 MACHINE, the kernel should also have
CONFIG_X86_X32=y. This adds a x32 fragment that can be used to trigger
the right ABI.
The commit also contains a check for mx32 in TUNE_FEATURES, and if
present, the new fragment will be appended to KERNEL_FEATURES and
trigger the support in the kernel.
cc: Saul Wold <sgw@linux.intel.com> Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Phil Blundell [Tue, 11 Sep 2012 18:14:25 +0000 (19:14 +0100)]
eglibc: Restore ${PN} to before ${PN}-dev in PACKAGES
Commit 13544fbc6217fee1731a6da1e2cf94901a500842 changed the ordering
of PACKAGES so that ${PN}-dev came before ${PN}. However, this caused
the FILES matching to go wrong if ${libdir} == ${base_libdir}. Fix this
by moving ${PN} ahead of ${PN}-dev once again.
Signed-off-by: Phil Blundell <pb@pbcl.net> Signed-off-by: Saul Wold <sgw@linux.intel.com>
sysvinit-inittab_2.88dsf.bb: only run serial checks at boot if we have items to check
Right now, we delay running the serial console checks to we boot up. This causes
issues for read only file systems. So, if have not configured any serial ports to
check via SERIAL_CONSOLES_CHECK we can skip the check at boot. This fixes any
issues with read only file systems and ipk packaging.
Signed-off-by: Matthew McClintock <msm@freescale.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Bruce Ashfield [Tue, 11 Sep 2012 13:05:14 +0000 (09:05 -0400)]
linux-yocto*: append to KERNEL_FEATURES instead of assigning
It is sometimes useful for KERNEL_FEATURES to be set in a machine
or other configuration file. The linux-yocto recipes currently
initialize the variable, which clobbers any values set by .conf
files.
Appending to the variables allows these settings to propagate to
the kernel configuration, while maintaining the existing set of
added kernel features.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Richard Purdie [Tue, 11 Sep 2012 14:22:29 +0000 (15:22 +0100)]
autotools.bbclass: Add functionality to force a clean of ${B} when reconfiguring (and ${S} != ${B})
Unfortunately whilst rerunning configure and make against a project will mostly
work there are situations where it does not correctly do the right thing.
In particular, eglibc and gcc will fail out with errors where settings
do not match a previously built configuration. It could be argued they are
broken but the situation is what it is. There is the possibility of more subtle
errors too.
This patch adds removal of the build directory (${B}) when configure is
rerunning, the sstate checksum for do_configure has changed and ${S} != ${B}.
We could simply use a stamp but saving out the previous configuration checksum
adds some data at no real overhead.
If we find there are things where we want to disable this behaviour with
CONFIGURESTAMPFILE = "" in the recipe, or users could disable it globally.
[YOCTO #2774]
[YOCTO #2848]
This is particularly helpful for eglibc and gcc which use split builds by default and
are a particular source of reconfigure type problems.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org> Signed-off-by: Saul Wold <sgw@linux.intel.com>
User mode emulation binaries are linked using a local linker script. The
nativesdk ones were not used and the resulting binaries did not have the
interp section resized. Hence, those binaries could not be relocated.
Martin Ertsaas [Tue, 11 Sep 2012 08:30:18 +0000 (10:30 +0200)]
gettext: Make gettext 0.16.1 extend native and nativesdk.
gettext 0.16.1 is a GPLv2 version of gettext. Making that extend native and
nativesdk makes sure we use the same version of gettext for compiling internally
as well as in our toolchain.
Signed-off-by: Martin Ertsaas <mertsas@cisco.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Bruce Ashfield [Mon, 10 Sep 2012 17:51:00 +0000 (13:51 -0400)]
linux-yocto/3.4: v3.4.10 and uprobes/kprobes configuration updates
Updating to 3.4.10 which has been soaking for a bit now, as well
as picking up the following meta commits from Tom Z:
a82db2f meta: have systemtap use kprobes and uprobes feature d5d5b80 meta: add kprobes support to ktypes/standard b32d373 meta: add kprobes feature d40ed99 meta: have uprobe feature use uprobe.cfg a69d1db meta: add uprobe.cfg
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
gcc-4.7: Backport libgcc fixes to appease the new build sequence
This makes the libgcc builds identical when done with gcc-cross-initial
or final gcc-cross. Since eglibc only sees gcc-cross-initial it is
important that the final libgcc that appears on root file system is same
as the one against which eglibc was built.
Ross Burton [Mon, 10 Sep 2012 16:02:22 +0000 (17:02 +0100)]
webkit-gtk: work around Make bug by re-running make
GNU make 3.82 has a bug where it drops required dependencies.
https://bugs.webkit.org/show_bug.cgi?id=79498 is the WebKitGTK+
bug, and http://savannah.gnu.org/bugs/?30653 is the GNU Make bug.
Work around this by running make again if it fails just in case the failure is
due to the bug.
Based on a patch by Andreas Müller <schnitzeltony@googlemail.com>.
[ YOCTO #2816 ]
Signed-off-by: Ross Burton <ross.burton@intel.com>
Phil Blundell [Sun, 9 Sep 2012 10:48:45 +0000 (11:48 +0100)]
shadow-native: Ensure that ${sbindir} and ${base_sbindir} are respected
These values need to be passed on the command line to "make install" otherwise
shadow will use its own built-in idea of where those directories are located.
Signed-off-by: Phil Blundell <pb@pbcl.net> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Phil Blundell [Sun, 9 Sep 2012 10:22:53 +0000 (11:22 +0100)]
shadow: Fix various invalid assumptions about directory layout
The makefiles in the shadow package have their own hard-coded paths
for ${base_bindir} and ${base_sbindir} (known as "bindir" and "sbindir"
in shadow-speak). Ensure that they install into our paths rather than
their own.
Also check that ${base_bindir} and ${bindir} are different before trying
to move files from one to the other; likewise for ${base_sbindir} and
${sbindir}.
Signed-off-by: Phil Blundell <pb@pbcl.net> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Phil Blundell [Sun, 9 Sep 2012 09:13:08 +0000 (10:13 +0100)]
perl-native: PROVIDE libmodule-build-perl-native for consistency with non-native perl
This module is, apparently, included in the standard perl distribution
since 5.10.1 or so. The regular perl recipe has had this PROVIDES for a
while but it seems to have been overlooked in the native version.
Signed-off-by: Phil Blundell <pb@pbcl.net> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Mark Hatle [Fri, 7 Sep 2012 22:07:11 +0000 (17:07 -0500)]
qt4: Update qt4.inc to remove staticdev deps in -dbg packages
It appears that the qt4.inc had a copy/paste error relating to creating
a list of staticdev packages, that caused them to show up as dependencies
in the -dbg package.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Mark Asselstine [Sat, 8 Sep 2012 13:42:21 +0000 (08:42 -0500)]
base-files: provide a mechanism to skip creation of the hostname file
The existence of a /etc/hostname file causes any hostname provided on
the kernel command line or via dhcp to be overwritten by the
initscripts 'init.d/hostname.sh'. This change allows you to set a
value of "" for 'hostname' which will skip the creation of the
/etc/hostname file by the base-files package.
Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com> Signed-off-by: Jason Wessel <jason.wessel@windriver.com> Signed-off-by: Mark Hatle <mark.hatle@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Mark Hatle [Fri, 7 Sep 2012 22:05:44 +0000 (17:05 -0500)]
package_rpm.bbclass: Avoid unnecessary installs in complementary pass
When called with the complementary install option, the first step is to
backup the install manifest so that we can avoid installing items previously
installed. However, this backup process skipped the initial_install portion
of the manifest, causing early install items like libc6, bash, and base-files
to be installed a second time.
Fix this by cating the files to original_solution. This is done as an append to
allow multiple calls to package_install_internal_rpm to work.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Paul Eggleton [Mon, 10 Sep 2012 12:58:10 +0000 (13:58 +0100)]
classes/packageinfo: use better method to check if package exists
Instead of using a rather error-prone method of looking for output
package files in order to determine if a package got created, use the
.packaged file within pkgdata.
This fixes two separate issues:
* Some packages apparently not being found by this code e.g. all
apm/apmd packages when using ipk packaging.
* Buggy implementation of this checking code which triggered an
exception during the event handler if PKGV was overridden on a
per-package basis (as it is with external-sourcery-toolchain), which
blocked Hob from completing parsing at 99% - fixes [YOCTO #2651].
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Paul Eggleton [Mon, 27 Aug 2012 20:45:14 +0000 (20:45 +0000)]
lib/oe/sstatesig.py: add signature data query function
Add a function that can be used from BitBake code which will find
signature data (sigdata/siginfo) files based on specified criteria, and
hook it into BitBake as bb.siggen.find_siginfo.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
LSB needs wget to download packages but wget provided by busybox doesn't
support some options such as '-N'.
LSB perl test 4.1.6-2 case all/tst_perlModPresent.pl,
../lib/Class/ISA/t/00_about_verbose and ../lib/Class/ISA/t/01_old_junk.t
fail because of lack of these modules, add them to make test pass.
File CORE/config.h which is provided by perl-dev and file
unicore/version which is provided by perl-doc are required by LSB perl
test cases.
Add perl-dev and perl-doc to packagegroups-core-lsb.
[Yocto #3030 #3031 #3052 #3054 #3055]
Signed-off-by: Kang Kai <kai.kang@windriver.com>
Rebased for packagegroup change -sgw
These perl libraries are being added directly to OE-Core for 4.1
LSB Complainace, when 5.0 comes out early next year (2013), we will
remove these changes.
core-image: allow root login when debug-tweaks is enabled
This allows root to login over ssh with an empty password just like
dropbear when the debug-tweaks are enabled, it's important to disable
debug-tweaks for a production system as this will leave open a security
hole!
Thanks to Marc for the settings. Cc: Marc Ferland <marc.ferland@gmail.com>
[Yocto #3078]
Martin Jansa [Wed, 5 Sep 2012 01:05:37 +0000 (03:05 +0200)]
pixman: ignore NEON, IWMMXT, LOONGSON_MMI variables for class-native
* pixman-native can have different do_configure sstate checksums if it's built with armv4t machine and armv7a
OE @ ~ $ bitbake/bin/bitbake-diffsigs
before-mgmt/stamps.1346795706/nokia900/x86_64-linux/pixman-native-*.do_configure.sigdata.*
after-mgmt/stamps.1346801508/om-gta02/x86_64-linux/pixman-native-*.do_configure.sigdata.*
basehash changed from 27e577de60880a788c7aaba797ef83e0 to c6799807eb3e767daf1e75738fc753f7
Variable NEON value changed from to --disable-arm-neon
* so if you start building with different machine then last time (wrt
NEON setting) all recipes which depends on pixman-native will be rebuilt too
* this explains why sstate-cache-management.sh wanted to remove many
native sstate packages when --stamps-dir option was used (see comment
28 in https://bugzilla.yoctoproject.org/show_bug.cgi?id=2897)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
* Check ${target_rootfs}/etc/passwd rather than
${target_rootfs}${rpmlibdir} to make sure that it has been previously
installed.
* Remove the "--nodeps" when incremental image generation, it should
take care of the dependencies. Still use "--replacefiles --replacepkgs" in
case there are conflicts.
[YOCTO #3047]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Kang Kai [Wed, 5 Sep 2012 09:31:04 +0000 (17:31 +0800)]
lsb: update version
Update package lsb version to be same with current lsb test suit
version. Because when install the suit, it warns that need lsb
version >= 3.0 at least.
Drop the duplicated creating files under /etc/lsb-release.d.
Provides directories /etc/opt and /var/opt that they are required by
package lsb-dist-checker in lsb test suit.
Signed-off-by: Kang Kai <kai.kang@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Robert Yang [Sun, 2 Sep 2012 04:15:47 +0000 (12:15 +0800)]
bin_package.bbclass: binary package recipe class
This is used for the binary package recipe, it's been suggested that it
would be a useful feature to be able to easily take an RPM or similar
containing a software binary from a 3rd party software vendor and
integrate it into an image created by the build system.
* Brief introduction
- The binary pkg can be .rpm, .deb, .ipk and other formats which can
be unpacked by bitbake fetcher.
- Let bitbake unpack the bianry package, just like unpack the source
package.
- Skip the do_configure and do_compile.
- Install the files to ${D}
- Other steps are similar to the source package's recipe.
* Note:
- The "subdir" parameter in the SRC_URI is useful for the binary
package recipe, so I added an example in the comment.
- I have sent a patch to bitbake-devel mailing list to support
unpack the .rpm, .ipk, and .deb files.
[YOCTO #1592]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Saul Wold <sgw@linux.intel.com>
Robert Yang [Mon, 10 Sep 2012 07:58:52 +0000 (15:58 +0800)]
package_rpm.bbclass: fix the arch (replace "-" with "_")
rpm can't use the "-" as the arch, which causes problem, e.g., when
MACHINE = "beagleboard":
* The arch should be armv7a-vfp-neon, but rpm only takes the armv7a,
this is incorrect since it is mixed with real arch armv7a.
* The nativesdk's arch should be i686-nativesdk (or x86_64-nativesdk),
but rpm only takes the i686 (or x86_64), this in incorrect since it is
mixed with the arch i686 (or x86_64).
Replace "-" with "_" when rpm package and the rootfs generation would
fix the problem, I think this is fine since it doesn't change the tune's
arch, the package manager doesn't care about the arch's name, but it
needs a unify arch system to avoid confusing. This is similar to what we
have done on the deb which fixed the arch i486, i586 and so on to i386.
[YOCTO #2328]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Paul Eggleton [Fri, 7 Sep 2012 13:23:05 +0000 (14:23 +0100)]
classes/mirrors: remove bogus gnutls mirror
This mirror entry which maps to itself plus a slash, if matched, put the
fetcher into a circular loop until the stack space is exhausted. A patch
has been sent to fix this issue in BitBake, but we should remove the
bogus entry as well.
(Note that this entry does not actually trigger the issue with current
master because the gnutls recipe now uses GNU_MIRROR instead of
ftp.gnutls.org, thus the bogus mirror entry is not matched.)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
