Richard Purdie [Tue, 17 Apr 2018 14:42:31 +0000 (15:42 +0100)]
uninative: Add allow-shlib-undefined to BUILD_LDFLAGS and drop other workarounds
We have a problem when for example, a glibc 2.27 based system builds some
library like libpopt-native and puts it into sstate then it is reused
on a pre glibc-2.27 system to build something which depends on popt like
rpm-native. This results in an error like:
recipe-sysroot-native/usr/lib/libpopt.so: undefined reference to `glob@GLIBC_2.27'
In the past we've had this problem with new symbols like getrandom and
getentropy, here its with a more complex symbol where there is an old
version and a newer version.
We've looked into various options, basically we cannot link against our
uninative libc/ld.so since we don't have the right headers or compiler
link libraries. The compiler doesn't allow you to switch in a new set
either, even if we did want to ship them. Shipping a complete compiler,
dev headers and libs also isn't an option.
On the other hand if we follow the ld man page, it does say:
"""
The reasons for allowing undefined symbol references in shared libraries
specified at link time are that:
- A shared library specified at link time may not be the same as the one
that is available at load time, so the symbol might actually be
resolvable at load time.
"""
which is exactly this case. By the time the binary runs, it will use
our uninative loader and libc and the symbol will be available.
Therefore we basically have a choice, we get weird intermittent bugs,
we drop uninative entirely, or we pass this option.
If we pass the option, we can drop the other workarounds too.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Fri, 13 Apr 2018 16:08:10 +0000 (17:08 +0100)]
bitbake.conf: Set and export TZ envvar to UTC
We just ran into an issue where tar failed to build on one server setup
but built everywhere else just fine.
It was running makeinfo to regenerate some docs files and makeinfo was too
old for the host it was running on. There was no dependency on makeinfo-native
as it was not meant to be regenerating the docs.
It was being regenerated as a date from a timestamp used in the docs
was different in Asian timezones than in the other timezones our builds
were being tested in.
I added an entry to https://wiki.yoctoproject.org/wiki/TipsAndTricks/
about how this was debugged.
As such, lets default to setting and exporting TZ to 'UTC' as was already
pioneered by the reproducibile builds work. This makes the builds
deterministic.
[YOCTO #12665]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Martin Jansa [Wed, 11 Apr 2018 17:55:32 +0000 (17:55 +0000)]
scripts/test-dependencies.sh: remove
* with RSS used in pyro this script isn't very useful anymore
* RSS makes sure that the dependencies are almost always deterministic
the only case known to me where dependencies are different based on
what was already built in TMPDIR are runtime dependencies resolved
by shlibs code in package.bbclass (which is using global pkgdata, not
specific to given recipe and its RSS) as described here:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=9217#c4
but for this case it's not worth running complete test-dependencies.sh
runs
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ramon Fried [Tue, 10 Apr 2018 11:20:40 +0000 (14:20 +0300)]
xserver-nodm-init: Respawn service in case of failure
It appears that sometimes xserver-nodm.service is starting before
display driver finished loading causing the following failure in
Xorg log:
(EE) open /dev/dri/card0: No such file or directory
The introduced by this patch is to restart the service,
hopefully the display driver will finish loading.
Signed-off-by: Ramon Fried <rfried@codeaurora.org> Signed-off-by: Ross Burton <ross.burton@intel.com>
Zhixiong Chi [Wed, 11 Apr 2018 08:26:18 +0000 (16:26 +0800)]
valgrind: fix the shared object issue while prelink ptest
If valgrind-ptest is installed, we will get the some prelink error
like below at do_image:
.../usr/sbin/prelink: /usr/lib64/valgrind/ptest/memcheck/tests/wrap7:\
Could not find one of the dependencies: \
.../usr/sbin//prelink-rtld: error \
while loading shared libraries: wrap7so.so: cannot open shared \
object file: No such file or directory
The wrap7 needs to link the shared object in the path
/usr/lib64/valgrind/ptest/memcheck/tests, but it fails.
So we correct the path for ptest.
Signed-off-by: Zhixiong Chi <zhixiong.chi@windriver.com> Signed-off-by: Chen Qi <Qi.Chen@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Otavio Salvador [Wed, 11 Apr 2018 14:32:52 +0000 (11:32 -0300)]
distro: default: Change Go default version to 1.9
For the time being, there is a serious bug[1] in Go 1.10 when it comes to
use the shared runtime support which cases problems in multiple projects.
1. https://github.com/golang/go/issues/24640
It is still unclear if the problem arises from a bug inside the
compiler itself or it makes a real problem more visible. Either way,
using 1.10 as default seems to be a risk so we are changing back to
1.9 for now.
Refs: [YOCTO: #12631]
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br> Signed-off-by: Ross Burton <ross.burton@intel.com>
* Fix arbitrary command execution in ed-style patches:
- src/pch.c (do_ed_script): Write ed script to a temporary file instead
of piping it to ed: this will cause ed to abort on invalid commands
instead of rejecting them and carrying on.
- tests/ed-style: New test case.
- tests/Makefile.am (TESTS): Add test case.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
* Fix segfault with mangled rename patch
- src/pch.c (intuit_diff_type): Ensure that two filenames are specified
for renames and copies (fix the existing check).
Signed-off-by: Jackie Huang <jackie.huang@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Improve reporting when the icecream environment cannot be created by
assigning the flock call a specific error number when the lock fails so
it can be distinguished from environment creation errors.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Shared libraries sometimes (frequently?) don't have a program
interpreter specified. The previous code would fail to find the library
dependencies in these cases because no interpreter could be found.
Commonly, this meant that if a library depends on another library, it
might not be included toolchain because dependency scanning stops with
the first one.
Instead, capture the program interpreter from the program or library
that starts the dependency chain and use that interpreter to get all of
the dependencies in the chain, recursively.
Additionally, if no interpreter can be found, fallback to using ldd
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Some newer libraries and programs use RUNPATH to specify the library
search path. These executables were being skipped by the rpath fixup
code because it was grepping the ELF header for RPATH only. A more
correct solution is to ask patchelf to report the rpath, as that tool
will properly report either RPATH or RUNPATH as appropriate.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Modifies the icecc-create-env script so that it can log output to a log
file. In addition, a --debug flag is added that allows verbose logging.
Finally, the silent flag was removed since it was never used in
icecc.bbclass
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Jaewon Lee [Tue, 10 Apr 2018 21:26:47 +0000 (14:26 -0700)]
gconf: fix saving of settings when config folder doesnt exist
In some circumstances, gconf isn't able to save configurations
because ~/.config folder aka root_dir doesn't exist.
For example when saving settings using matchbox-appearance,
the following error is shown:
GConf Error: Configuration server couldn't be contacted: D-BUS error:
Can't overwrite existing read-only value: Value for
`/desktop/poky/interface/font_name' set in a read-only source at the
front of your configuration path
This issue was not seen before because ~/.config directory is shared
between several packages and one of those packages usually creates it
by the time gconf wants to use it.
This patch makes sure that gconf creates the .config directory if it
doesn't exist, along with the gconf directory inside it.
[YOCTO #12632]
Signed-off-by: Alejandro Hernandez <alejandr@xilinx.com> Signed-off-by: Jaewon Lee <jaewon.lee@xilinx.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Tom Hochstein [Mon, 9 Apr 2018 16:20:32 +0000 (11:20 -0500)]
python3: Add recommended modules to nativesdk install
The python3 installation in the SDK did not include the minimum set
of modules to be functional, particularly in the case where Python
is brought in through dependencies. Rather than requiring the user
to explicitly add the modules, it's better to pull in the modules
through RRECOMMENDS. Note that the Python 2 recipe already does
this.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
"srcversion" field inserted into module modinfo section contains a
sum of the source files which made it. However, this field can
be incorrect. Building the same module can end up having inconsistent
srcversion field eventhough the sources remain the same.
This can be reproduced by building modules in a deeply nested directory,
but other factors contribute as well.
The reason for incorrect srcversion is that some source files can be
simply silently skipped from the checksum calculation due to limited
buffer space for line parsing.
This patch addresses two issues:
1. Allocates a larger line buffer (32k vs 4k).
2. Issues a warning if a line length exceeds the line buffer.
Signed-off-by: Juro Bystricky <juro.bystricky@intel.com> Signed-off-by: Juro Bystricky <juro.bystricky@intel.com> Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Updates to 4.14 via the -stable releases have also introduced the same
dependencies to 4.14's "make scripts". As such, we bring the same lines
into 4.14 to restore the ability to build scripts.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
Ross Burton [Mon, 9 Apr 2018 15:17:05 +0000 (16:17 +0100)]
populate_sdk_base: use xz -T instead of pixz
xz has native support for threaded compression now and SDK creation was the only
part of oe-core which is using pixz instead of xz.
Not only does this remove pixz-native from the SDK dependencies, but in my
limited testing xz -T0 is slightly faster and produces smaller archives than
pixz for the same input.
Signed-off-by: Ross Burton <ross.burton@intel.com>
Ross Burton [Tue, 10 Apr 2018 14:32:19 +0000 (15:32 +0100)]
image_types: use pigz to create .gz files
Since pigz is no longer a drop-in replacement for gzip (oe-core 1624b7b) the
image creation has been using gzip instead of pigz, despite still depending on
pigz-native. Fix this by invoking pigz explicitly.
Signed-off-by: Ross Burton <ross.burton@intel.com>
linux-firmware: fix the mess of licenses
* LICENSE_CREATE_PACKAGE functionality in license.bbclass when enabled
adds new package with suffix:
LICENSE_PACKAGE_SUFFIX ??= "-lic"
but then it checks if ${PN}-${LICENSE_PACKAGE_SUFFIX} is included
in PACKAGES before adding it and when found it shows:
WARNING: linux-firmware-1_0.0+gitAUTOINC+4c0bf113a5-r0 do_package: linux-firmware-lic package already existed in linux-firmware.
and doesn't add the ${PN}-lic to PACKAGES and causes another warning:
WARNING: linux-firmware-1_0.0+gitAUTOINC+4c0bf113a5-r0 do_package: QA Issue: linux-firmware: Files/directories were installed but not shipped in any package:
/usr
/usr/share
/usr/share/licenses
/usr/share/licenses/linux-firmware
that's because it was searching ${PN}-lic in PACKAGES as a string
so it found ${PN}-lic as a substring of ${PN}-license, add a split
to search in an list
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Martin Jansa [Mon, 9 Apr 2018 07:51:41 +0000 (07:51 +0000)]
libxcrypt: add -std=gnu99 to BUILD_CPPFLAGS
* add it to allow older distributions e.g. Ubuntu 14.04 with gcc 4.8
to build this, otherwise it fails with:
../git/gen-des-tables.c: In function 'write_table_u8':
../git/gen-des-tables.c:307:3: error: 'for' loop initial declarations are only allowed in C99 mode
for (size_t i = 0; i < m; i++)
^
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add option for dbus in configure.ac, and explicitly build without dbus
for rpm-native.
Previously, the rpm recipe tries to prevent rpm-native from attempting
to inhibit shutdown via session dbus by appending '--disable-plugins'
to EXTRA_OECONF in case of native.
However, some layer may need some functionality via plugin support. And
when it enables it, we would meet the following warning at rootfs time.
Unable to get systemd shutdown inhibition lock: Socket name too long
As plugins/systemd_inhibit.c is the only place that's related to this
dependency, we can be sure that dbus is really not needed for rpm-native.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
license.bbclass: Minor simplification of get_deployed_dependencies()
Since ${SSTATE_ARCHS} now contains ${PACKAGE_EXTRA_ARCHS} there is no
longer any need to add those extra architectures to the list of
architectures handled in get_deployed_dependencies().
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
sstate.bbclass: Add ${PACKAGE_EXTRA_ARCHS} to SSTATE_ARCHS
This makes sure files provided by packages that use any of the extra
architectures defined using ${PACKAGE_EXTRA_ARCHS} are cleaned from
tmp/sysroot-components when sstate_eventhandler2() executes.
Without this, changing a package from using one of the extra
architectures to another architecture would lead to files being
leftover in tmp/sysroot-components, which could later be picked up
when creating the RSS for another package rather than the files that
belonged to the updated package.
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Fri, 6 Apr 2018 12:53:53 +0000 (14:53 +0200)]
nativesdk-glibc: Split glibc and libcrypt to use libxcrypt instead
Fedora28[1] has decided to go ahead and use libxcrypt to replace libcrypt from glibc
despite the change not having merged into glibc upstream yet. This breaks the use of
uninative in OE on fedora28 since binaries there are now using new symbols only found
in libxcrypt. libxcrypt is meant to be backwards compatible with libcrypt but not the
reverse.
Since this will impact OE in the next release cycle, this changes nativesdk only
to use this new model and adds libxcrypt to work in that case. This allows us to
build a uninative which is compatible with fedora28 and previous other OSes.
In order to work, recipes will now need to depend on virtual/crypt where they use
libcrypt since its now a separate library and we can't depend on it from glibc to
preseve backwards compatibility since glibc needs to build first. For now, only the
problematic nativesdk recipes have been fixed up. For target use, the default
provider remains glibc for now. Assuming this change is merged into upstream glibc,
we will need to roll this change out for the target but we will do this in the next
release cycle when we can better deal with the resulting bugs.
Original patch from Charles-Antoine Couret <charles-antoine.couret@essensium.com>,
tweaked by RP to add virtual provides, SkipRecipe for libxcrypt and other minor
tweaks.
Signed-off-by: Charles-Antoine Couret <charles-antoine.couret@essensium.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Now that we see warnings if LAYERSERIES_COMPAT is unset, the auto generated
code from devtool needs to set this to avoid warnings which break
various tests.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Sat, 7 Apr 2018 10:32:03 +0000 (11:32 +0100)]
devtool/oeqa: Ensure added layers set LAYERSERIES_COMPAT
Now that we see warnings if LAYERSERIES_COMPAT is unset, the auto generated
code from devtool/oeqa needs to set this to avoid warnings which break
various tests.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Although the package will get an automatic prefix "kernel-module", so
the package kernel-module-hello does exist, populating rootfs can
generate an error:
- nothing provides kernel-module-hello ...
This is quite unfortunate, as this recipe is used as a sample.
Adding RPROVIDES_${PN} += "kernel-module-hello" to the recipe fixes
the problem.
[YOCTO #12641]
Signed-off-by: Juro Bystricky <juro.bystricky@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Victor Kamensky [Thu, 5 Apr 2018 18:25:30 +0000 (11:25 -0700)]
crosstap: replace script with new python based implementation
New crosstap python implementation is total replacement for
crosstap shell script, that has superseding capabilities.
New script support cross compiling of SystemTap scripts
for user-land, by using supplied image rootfs. Whereas old
script could only deal with scripts against kernel. New script
has more complex logic and additional capabilities.
As invocation interface new script support old "legacy"
mode and provides alternative new regular options interface
to access additional functionality.
Signed-off-by: Victor Kamensky <kamensky@cisco.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Victor Kamensky [Thu, 5 Apr 2018 18:25:29 +0000 (11:25 -0700)]
systemtap: create translator packageconfig
For cases when systemap module compilation happens on host in
cross-compilation mode, and it is desirable to minimize systemtap
presense on target we need to have just smallest possible set of
utilties that are required to run compiled modules.
Introduce new "translator" PACKAGECONFIG, if it is not set
it would mean that just minimal set of run-time utilities will
be included in the package.
For run-time only systemtap build variant use
PACKAGECONFIG_pn-systemtap = "" or
PACKAGECONFIG_pn-systemtap = "monitor"
Suggested-by: Taras Kondratiuk <takondra@cisco.com> Signed-off-by: Victor Kamensky <kamensky@cisco.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Victor Kamensky [Thu, 5 Apr 2018 18:25:26 +0000 (11:25 -0700)]
image: add image-combined-dbg helper class
There is IMAGE_GEN_DEBUGFS="1" variable that enables build of
additional rootfs-dbg and additional archive that contains
complimentary symbols files for a given image. But the issue
with this resulting directory and tarball that before use it
has to be combined with original image content. It is required
since all cross debugging tools like gdb, perf, and systemtap
need file system that contains both target executables/libraries
and their symbols. Those tools need to find executable/library
first and through it debuglink note find corresponding symbols
file.
image-combined-dbg when added to USER_CLASSES just copies
final resulting rootfs back into rootfs-dbg creating combined
target and debug symbols rootfs that can be used for debugging
directly.
Signed-off-by: Victor Kamensky <kamensky@cisco.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
André Draszik [Thu, 5 Apr 2018 11:08:58 +0000 (12:08 +0100)]
curl: DEPENDS on libidn2 (not libidn)
Since v7.51.0, libidn2 is the only available option, libidn
support was dropped.
The configure option was renamed as of v7.53.0
Therefore, curl unconditionally tries to build against libidn2,
which in particular is a problem for curl-native, as that might
or might not build against the build-machine's libidn2 now,
which furthermore causes problems when trying to share sstate
between multiple build machines.
We therefore see the following in the config log:
...
checking whether to build with libidn2... (assumed) yes
...
checking for libidn2 options with pkg-config... no
configure: IDN_LIBS: "-lidn2"
configure: IDN_LDFLAGS: ""
configure: IDN_CPPFLAGS: ""
configure: IDN_DIR: ""
checking if idn2_lookup_ul can be linked... yes
checking idn2.h usability... yes
checking idn2.h presence... yes
checking for idn2.h... yes
...
IDN support: enabled (libidn2)
...
even though this recipe tries to disable that.
While libidn2 isn't available in OE, this change at least:
* prevents curl-native to silently build against libidn2 if
that is installed on build machine, even if not requested
* alerts people who use the PACKAGECONFIG option that it's
not actually doing what they intend to do
Signed-off-by: André Draszik <andre.draszik@jci.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Tom Hochstein [Tue, 3 Apr 2018 13:51:15 +0000 (08:51 -0500)]
archiver: Fix archive for linked kernel source
When archiving a kernel, if S is ${WORKDIR}/git, then
${WORKDIR}/git is in fact a soft link into work-shared,
and the archive contains just the soft link. Fix by
archiving the real directory.
Signed-off-by: Tom Hochstein <tom.hochstein@nxp.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
glib-2.0/glib.inc: apply MLPREFIX renaming to all package classes
Now that we have postinst intercept execution working for SDK generation,
adding MLPREFIX again makes sense in all cases, as the intercepts do require
that it is there.
Previously this wasn't done, and so any packages installed from
populate_sdk would not have the postinsts fully executed
(particularly generation of various caches via running nativesdk or target
binaries with qemu wasn't working).
[YOCTO #12630]
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
package_manager.py: move intercept running logic from rootfs class to PackageManager class
This allows running the intercepts when creating SDKs, which previously
wasn't possible, as SDK code does not use the rootfs class, and calls
into PackageManager methods directly.
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Its use required a script from an external repo which hasn't been updated
in 4 years, the recipe itself is out of date (doesn't install all
intercepts), and there is no oe-selftest or documentation for this.
If anyone still wants this, please do it in a separate layer.
Signed-off-by: Alexander Kanavin <alexander.kanavin@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The list of patches in SRC_URI was restructured in 49aae1d7, and left
a comment about patches from Debian hanging without context. Move and
reword it to make it remain useful.
There was also a leftover comment that referred to two .h files and
do_configure_prepend() that were removed in f960c026.
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
package_manager.py: Skip gpgcheck while using dnf on target
By default, RPM_SIGN_PACKAGES is not defined. Add gpgcheck=0 to
oe-remote-repo.repo file, otherwise dnf will complain during
install operation on target
Note, RPM_SIGN_PACKAGES is set only when you inherit sign_rpm explicitly
Signed-off-by: Manjukumar Matha <manjukumar.harthikote-matha@xilinx.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
As introduced by a56fb90dc3805494eeaf04c60538425e8d52efc5 ('base.bbclass
wipe ${S} before unpacking source') the base.bbclass uses a python
anonymous function to set the 'do_unpack' varflag 'cleandirs' to either
'${S}' or '${S}/patches' depending on equality of '${S}' and '${WORKDIR}'.
Not that this only differs from the way almost all other recipes set or
modify a tasks 'cleandirs' flag, it also has a significant impact on the
kernel.bbclass (and possibly further ones) and causes incorrect
behavior for rebuilds triggered by source modification, e.g. by a change
of the defconfig file for a kernel build.
The kernel.bbclass tries to extend do_unpack[cleandirs]:
As python anonymous functions are evaluated at the very end of recipe
parsing, the d.setVarFlag('do_unpack', 'cleandirs', '${S}') statement in
base.bbclass will overwrite every modification to cleandirs that is done
as shown for the kernel class above.
As a result of this, a change to a kernels 'defconfig' will lead to an
updated defconfig file in ${WORKDIR}, but as ${B} never gets cleaned and
${B}/.config still exists, it will not be copied to ${B}/.config and
thus not find its way in the build kernel.
This is a severe issue for the kernel development and build process!
This patch changes setting of the cleandirs varflag in base.bbclass to
a simple variable assignment as almost all other recipes do it. This now
again allows overwriting or appending the varflag with common methods
such as done in kernel.bbclass.
This issue affects morty, pyro, rocko and master.
Signed-off-by: Enrico Jorns <ejo@pengutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It's possible that a trailing or extra slash somewhere in the external
source path could result in the directory not being removed from
cleandirs; it's also possible that a cleandirs entry is somewhere
underneath the source tree and that tree should never have parts of it
deleted by the build system. Use oe.path.is_path_parent() (which makes
paths absolute before checking them) to find out if any path in
cleandirs is anywhere underneath the external source path, and drop it
if it is.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Paul Eggleton [Wed, 4 Apr 2018 11:02:35 +0000 (23:02 +1200)]
classes/externalsrc: handle if cleandirs contains python expressions
Use the existing oe.recipeutils.split_var_value() function to split the
unexpanded value of the cleandirs varflag, in case it contains python
expressions - we don't want to split the expression itself as the chunks
will not expand properly individually and we can miss something that
expands to the source tree (and thus it can get deleted, the avoidance
of which is the whole point of this code).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cuero Bugot [Fri, 16 Mar 2018 17:31:30 +0000 (17:31 +0000)]
uninative: add variables to the whitelist so that it does not re-triger recipe parsing
When uninative is activated (poky's default) internal datastore variables are modified (NATIVELSBSTRING and SSTATEPOSTUNPACKFUNCS) to enable uninative
support. This is happening after parsing is done at the beginning of the build. On the next bitbake call the recipe would be parsed if the two
variables above were not added to the parsing whitelist BB_HASHCONFIG_WHITELIST.
The fix is to add these two variables to the recipe parsing whitelist BB_HASHCONFIG_WHITELIST, this is done at recipe parsing time, only when
uninative.bbclass is used.
Signed-off-by: Cuero Bugot <cbugot@sierrawireless.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
python3: Improve logging capabilities for do_create_manifest
Adds a couple of prints to get a nicer log, and creates a
small summary or report after checking every module, it
makes it more feasible for adoption, easier to debug why
a module ended at a certain package and see how the
manifest was created.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandr@xilinx.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It is not using autoconf completely, therefore there
is no autoreconf happening, so when we depend on latest
gnu-config changes e.g. new architectures like riscv
the build does not see them and fails.
Installing these files from native sysroot helps
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Define cpu type for Qemu in QEMU_EXTRAOPTIONS.
Otherways Qemu will emulate some virtual qemu32/64 processor,
which has very basic set of features, and programs built
for later CPU may crash, due to lack of new features (e.g. SSSE3).
Signed-off-by: Ruslan Ruslichenko <rruslich@cisco.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It can be installed to some non standard path in which
case build will be broken.
As python2.7 is specified in HOSTTOOLS we can rely
that it is present in the PATH, so no need to hardcode
it to /usr/bin.
Signed-off-by: Ruslan Ruslichenko <rruslich@cisco.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ming Liu [Sun, 1 Apr 2018 21:15:27 +0000 (23:15 +0200)]
recipetool: create: fix port number parsing issue
A flaw was found when I run:
$ recipetool create "ssh://git@xxx.xxx:7999/xxx.git"
the url turned out to be: "git://git@xxx.xxx/7999/xxx.git;protocol=ssh"
after parsing, the port number was parsed as part of the path, this is
definitely wrong and lead to fetching failures.
This issue could be fixed in reformat_git_uri, by filtering out port
numbers when formatting ":".
Signed-off-by: Ming Liu <liu.ming50@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Otavio Salvador [Sun, 1 Apr 2018 19:50:37 +0000 (16:50 -0300)]
cmake: upgrade 3.10.2 -> 3.10.3
,----[ Changes in 3.10.3 ]
| Brad King (1):
| CMake 3.10.3
|
| Craig Scott (1):
| GoogleTest: Rename TIMEOUT parameter to avoid clash
|
| Sebastian Holtermann (1):
| Autogen: Fix for the empty source file crash in 3.10.2
|
| Tianhao Chai (1):
| ccmake: fix status line buffer overflow on very wide terminals
`----
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Jonathan Liu [Tue, 3 Apr 2018 04:23:31 +0000 (14:23 +1000)]
kernel.bbclass: improve reproducibility
Set KBUILD_BUILD_VERSION=1 to avoid build version being incremented on
every build. It is visible in the output of "cat /proc/version" after
the hash character.
Signed-off-by: Jonathan Liu <net147@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With multi kernel support in the installer we can exceed this limit.
Calculate a sane size by checking the size of the original boot
partition minus some objects we know won't be installed, plus some extra
space for users.
In addition, in the common case where only one small kernel is present
to be installed, we actually get a smaller boot partition with less
wasted space.
Also add VIRTUAL-RUNTIME_base-utils to RDEPENDS where these scripts are
used, as they're needed for the du command.
[YOCTO #12583].
Signed-off-by: California Sullivan <california.l.sullivan@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Amanda Brindle [Thu, 15 Mar 2018 21:06:38 +0000 (14:06 -0700)]
buildhistory_analysis.py: Check if RPROVIDES changed order
Instead of assuming order has changed if no package has been added or
removed, loop through packages to check if order has changed. This will
prevent the script from falsely reporting "changed order" if a version
has increased.
Fixes [YOCTO #12334]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Fri, 30 Mar 2018 13:18:57 +0000 (14:18 +0100)]
pseudo: Upgrade to latest master
This change includes several bug fixes and improvements, including better path
handling (the existance of . and .. for files), handling of the sticky bit, and
syscall renameat2 handling and interception through syscall() which was breaking
coreutils mv operations on fedora27.
[YOCTO #12594]
[YOCTO #12379]
[YOCTO #11643]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Juro Bystricky [Thu, 29 Mar 2018 20:27:02 +0000 (13:27 -0700)]
libpcre-ptest: skip locale test
If a fr_FR locale is found, it is automatically tested. The test
will fail if the locale is UTF-8, as the test blindly assumes
(and expects) a non-UTF fr_FR locale.
The remedy is to skip the test.
[YOCTO #12215]
Signed-off-by: Juro Bystricky <juro.bystricky@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
python3: Fix do_create_manifest for python3-sqlite3
Some of the sqlite3 files ended up in python3-misc incorrectly,
this is caused becuse we couldnt add the whole ${libdir}/python3/sqlite3
folder on the package because we also have another sqlite3-tests
package that needs to include another folder from that directory.
This patch not only fixes the do_create_manifest script to handle this
situation, but also patches the manifest (created using the script)
which also fixes a hiddn runtime dependency that we wouldn't have seen.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandr@xilinx.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
python3: fix create_manifest to handle pycache folders
We have a couple of python modules that contain folders themselves,
for that reason they also contain a __pycache__ folder inside those
directories, since we include the whole folder in the manifest, the
pycache directories end up with the files and not the cache files.
This patch catches that and adds the directories to the correct
structure.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandr@xilinx.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
default-distrovars: set default KERNEL_IMAGETYPE(S)
With the change from assuming kernels will be named "vmlinuz"
everywhere, to instead using KERNEL_IMAGETYPE, we require that
KERNEL_IMAGETYPE is set to something. Instead of setting the default in
multiple individual files, set it in default-distrovars.inc.
x86(-64) arches get bzImage as the default. Others get zImage as per
meta/conf/documentation.conf.
Also set KERNEL_IMAGETYPES, as we will eventually be switching away from
KERNEL_IMAGETYPE.
Thanks to Andre McCurdy <armccurdy@gmail.com> for the arch defaults
part.
Signed-off-by: California Sullivan <california.l.sullivan@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
André Draszik [Thu, 29 Mar 2018 15:43:19 +0000 (16:43 +0100)]
ca-certificates: use relative symlinks from $ETCCERTSDIR
update-ca-certificates symlinks (trusted) certificates
from $CERTSDIR or $LOCALCERTSDIR into $ETCCERTSDIR.
update-ca-certificates can call hook scripts installed
into /etc/ca-certificates/update.d. Those scripts are
passed the pem file in /etc/ssl/certs/ that was added or
removed in this run and those pem files are absolute
symlinks into $CERTSDIR or $LOCALCERTSDIR at the moment.
When running update-ca-certificates during image build
time, they thusly all point into the host's file system,
not into the $SYSROOT. This means:
* the host's file system layout must match the one
produced by OE, and
* it also means that the host must have installed the same
(or more) certificates as the target in $CERTSDIR and
$LOCALCERTSDIR
This is a problem when wanting to execute hook scripts,
because they all need to be taught about $SYSROOT, and
behave differently depending on whether they're called
at image build time, or on the target, as otherwise they
will be trying to actually read the host's certificates
from $CERTSDIR or $LOCALCERTSDIR.
This also is a problem when running anything else during
image build time that depends on the trusted CA
certificates.
Changing the symlink to be relative solves all of these
problems. At the same time, we have to make sure to add
$CERTSDIR to SYSROOT_DIRS, so that the symlinks are still
valid when somebody DEPENDS on ca-certificates-native. As
a side-effect, this also fixes a problem in meta-java,
where some recipes (e.g. openjdk-8-native) try to access
certificates from $CERTSDIR to generate the java trustStore
at build time.
Do so.
Upstream-Status: Inappropriate [OE-specific] Signed-off-by: André Draszik <andre.draszik@jci.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Kernels which use tools/objtool can now fail when building external modules
due to objtool being missing, the generated files can also cause problems
for kernel-devsrc.
Ensure objtool is generated in make-mod-scripts by also calling
"make prepare".
For devsrc, delete the generated binaries since they'd be native
binaries and unsuitable for the target.
The oeqa kernel module tests also need to have the additional "make prepare"
step added.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Khem Raj [Fri, 23 Mar 2018 13:34:45 +0000 (06:34 -0700)]
nss: Use snprintf in sign.c
Fies security warnings
| sign.c:86:31: error: 'sprintf' may write a terminating nul past the end of the destination [-Werror=format-overflow=]
| sprintf(fullfn, "%s/%s", tree, tempfn);
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Ross Burton <ross.burton@intel.com>
