Move logic checking that all packages licenses are only a subset of
recipe licenses from base.bbclass to the insane.bbclass so that it's
evaluated only once, during do_package_qa.
As explained in the linked bugzilla entry, if a package license is not
part of the recipe license, the warning message gets shown an
unreasonable amount of time because it's evaluated every time a recipe
is parsed.
[YOCTO #10130]
This also makes it possible to silence this error with INSANE_SKIP.
(From OE-Core rev: ae404ef230882e442e9390b314e1ce023fdbbd1b)
Signed-off-by: Quentin Schulz <quentin.schulz@streamunlimited.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 852408ed4be1f64c57e196688728b7ed223d3493)
Signed-off-by: Steve Sakoman <steve@sakoman.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There are mutually exclusive PACKAGECONFIGs in recipes. Though it
declares that package configs are exclusive, it can't prevent users to
set them at same time. Extend PACKAGECONFIG to support specifying
conflicted package configs.
(From OE-Core rev: 734475b3f86d88a548bc9eb91d836bd1b9335e9f)
Signed-off-by: Kai Kang <kai.kang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently the test depends on which DISTRO_FEATURES and targets are selected.
Similar to SOURCE_MIRROR_FETCH, add PARSE_ALL_RECIPES to allow the test to see
a much wider range of them. This avoids issues added by the new ucontext musl
specific recipe and allows the current whitelist to be cleared out a bit (more
cleanup can follow).
(From OE-Core rev: c139e058bb2df48c79784cf3fd701e51bbc68427)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When some hosttool is a dead symlink, bitbake any recipe will
fail with error like below.
FileExistsError: [Errno 17] File exists: '/usr/bin/chrpath' -> '/path/to/builddir/tmp-glibc/hosttools/chrpath'
So we remove dead symlink under hostools/ directory to avoid
such error.
(From OE-Core rev: b03577cf85784fc052b89083054d911816f8c6e1)
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Instead of reporting ${LICENSE} when a package is identified as using
an incompatible license, report the license(s) that are actually
incompatible.
(From OE-Core rev: fb3405fac7b933a3f9d23e5faf6a3cf2d1990982)
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
After a number of rewrites, the code checking if a package has been
whitelisted for an incompatible license was calculating the
whitelisted packages twice (as 'whitelist' and as 'incompatwl').
(From OE-Core rev: 01dcc8c94b882d865fc72a067a90d9721ac5fea7)
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This helps fix an issue where python interpreter in hg script could
overflow the BINPRM_BUF_SIZE which is 128 on most of systems, because
interpreter is hardcoded and build paths can be deep.
This patch helps, because now the absolute python interp path in hg can
be replaced with '/usr/bin/env python3' and it will ensure that python3
from native install is used instead of the one on host.
(From OE-Core rev: 3adaaab787bd105cc0049d36c7b8caeaa9b633e5)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Cc: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
do_prepare_recipe_sysroot may perform groupadd, which requires pseudo.
However, do_prepare_recipe_sysroot does not depend on pseudo explicitly,
which sometimes causes a build error when building a recipe that adds
groups.
This issue only occurs when executing do_prepare_recipe_sysroot for a
recipe that adds groups before finishing a task that depends on pseudo
for a recipe that doesn't add groups.
(From OE-Core rev: 427320d7621846d57649af46bf57c88f86cac612)
Signed-off-by: Mattias Hansson <mattihn@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Turn:
ERROR: Unable to parse Var <OE_IMPORTED[:=]>
Traceback (most recent call last):
File "Var <OE_IMPORTED[:=]>", line 1, in <module>
File "/media/build1/poky/meta/classes/base.bbclass", line 35, in oe_import(d=<bb.data_smart.DataSmart object at 0x7f1d941ad208>):
for toimport in oe.data.typed_value("OE_IMPORTS", d):
> imported = __import__(toimport)
inject(toimport.split(".", 1)[0], imported)
File "/media/build1/poky/meta/lib/oe/sstatesig.py", line 267, in <module>:
>class SignatureGeneratorOEEquivHash(SignatureGeneratorOEBasicHashMixIn, bb.siggen.SignatureGeneratorUniHashMixIn, bb.siggen.SignatureGeneratorBasicHash):
name = "OEEquivHash"
bb.data_smart.ExpansionError: Failure expanding variable OE_IMPORTED[:=], expression was ${@oe_import(d)} which triggered exception AttributeError: module 'bb.siggen' has no attribute 'SignatureGeneratorUniHashMixIn'
into:
ERROR: Error importing OE modules: module 'bb.siggen' has no attribute 'SignatureGeneratorUniHashMixIn'
which can then trigger a version mismatch error message.
(From OE-Core rev: 5b01726d1cf478aba8d1acc73e29f6cab0c9371e)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* just to make sure it's expaned by bitbake before it gets
executed in shell
* e.g. with cmake.bbclass and cmake recipe (any recipe without
PACKAGECONFIG options have this issue) it looks like this:
bitbake -e cmake | grep EXTRA_OECMAKE=
EXTRA_OECMAKE=" -DCMAKE_DOC_DIR=share/doc/cmake-3.14
-DCMAKE_USE_SYSTEM_LIBRARIES=1 -DCMAKE_USE_SYSTEM_LIBRARY_JSONCPP=0
-DCMAKE_USE_SYSTEM_LIBRARY_LIBUV=0
-DCMAKE_USE_SYSTEM_LIBRARY_LIBRHASH=0 -DKWSYS_CHAR_IS_SIGNED=1
-DBUILD_CursesDialog=0 -DKWSYS_LFS_WORKS=1
\${PACKAGECONFIG_CONFARGS}"
(From OE-Core rev: 745b63f4e11a6536cabd97013973562631a0e080)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
OE_IMPORTS is not intended to be touched by users, but there are cases
in which layers might want to make additional Python modules available
to Python functions. For example, Python modules defined in the layer
themselves (under meta-layer/lib).
(From OE-Core rev: 00fa8391365863fa7805ad61b2d1a8425b9ea040)
Signed-off-by: Chris Laplante <chris.laplante@agilent.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Several fetchers support named sources that require setting a SRCREV with
the source name as a suffix. These named SRCREV variables are not captured
in the sstate hash calculation because they're only referenced within the
bitbake fetcher function. Several recipes avoid this bug by adding the
different SRCREVs to their packaging versioning but this is not enforced so
it is very trivial to trip this bug case.
Add a snippet to the base.bbclass anonymous python to add all named SRCREV
variables to the vardeps of do_fetch to capture them in the sstate hash
calculation.
Testing of the bug can be shown by running the following bitbake commands
with this patch set not applied:
bitbake vulkan-demos | tee
sed -i 's/SRCREV_gli = ".*"/SRCREV_gli = "xxx"/' \
../meta/recipes-graphics/vulkan/vulkan-demos_git.bb
bitbake vulkan-demos | tee;
The above results in no errors despite a broken SRCREV because the recipe is
considered unchanged by the bitbake sstate hash.
After applying this patch the above commands instead result in a fetcher
error which is correct.
(From OE-Core rev: fdc000b8ae1577f1b29efb2cc4927664ad41e21c)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This has been unused in OE-Core since the introduction of recipe specific
sysroots. Its not so useful since it only runs once upon sstate installation,
not per installation per sysroot.
Remove the weird looking comment left behind in pixbufcache too.
(From OE-Core rev: 2af49716504f65be0cb01f609ea9bfa334926589)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Move prepare_recipe_sysroot's task dependency on populate_sysroot from
base.bbclass (where it was specified in the middle of do_configure's
definition) to staging.bbclass (where the rest of
do_prepare_recipe_sysroot is defined). This was a left-over from when
recipe specific sysroots were introduced in commit 809746f5 and the
task dependency on populate_sysroot was moved from do_configure to
do_prepare_recipe_sysroot.
(From OE-Core rev: fe6d020dd609fbde7e4413cef9bbdd4e3962c82c)
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There might be no bb.event.ConfigParsed event if bitbake server is running, so
check bb.event.BuildStarted too to make sure HOSTTOOLS_DIR exists.
Fixed:
$ export BB_SERVER_TIMEOUT=-1
$ bitbake quilt-native
$ rm -fr tmp
$ bitbake quilt-native
ERROR: Error running gcc --version: /bin/sh: gcc: command not found
This error is caused by enable_uninative(), it runs twice (ConfigParsed and
BuildStarted), the error would happen when there is no ConfigParsed event
(no hosttools is created), but BuildStarted. This patch can fix the problem.
[YOCTO #13022]
(From OE-Core rev: da798db0a48282e3d4f58890a7aec42c3deff0b8)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Display the name of the restricted licenses which caused the recipe to
be skipped. This makes it easy to determine which license or licenses
are missing and need to be checked and whitelisted.
(From OE-Core rev: b71cd1ec45e247db688b784697829c1b485ca9ca)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Due to a bug in find [1], -ignore_readdir_race does not work correctly with
-delete. This can lead to spurious build failures when files disappear
while such a command is running; specifically this was seen in the case of
do_configure and do_populate_lic running concurrently for packages
with ${B} == ${WORKDIR}:
find: '.../sstate-build-populate_lic': No such file or directory
While the issue is fixed in the findutils git master, the find command of
the host system is called here, so we can't ensure that the used version
contains the fix. Many common distros have not updated to a recent enough
findutils version yet (Ubuntu 18.10 contains the fix, while 18.04 is still
affected).
Work around the issue by passing the output of find to 'rm -f' instead of
using -delete.
[1] https://savannah.gnu.org/bugs/?52981
(From OE-Core rev: 8079e2d62e23f7c274f46185e6dad64fa95394c1)
Signed-off-by: Matthias Schiffer <matthias.schiffer@ew.tq-group.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Several distros are now shipping "python" as python v3 contra to the original
python guidelines. This causes users confusion/pain in trying to use our tools.
We can just force "python" to "python2" within HOSTTOOLS to avoid this issue
and hide the complexity from the user.
(From OE-Core rev: b06a6cde5c5503f456f260c773cf126085e18c8d)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We assume that LIC_FILES_CHKSUM is a file: URI but don't actually verify this,
which can lead to problems if you have a URI that resolves to a path of / as
Bitbake will then dutifully checksum / recursively.
[ YOCTO #12883 ]
(From OE-Core rev: e2b8a3d5a10868f9c0dec8d7b9f5f89fdd100fc8)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There doesn't seem to be a clear reason to have two separate
variables to hold whitelisted GPLv3 recipes. Both variables are
treated the same, so adding a recipe to LGPLv2_WHITELIST_GPL-3.0 is
already equivalent to adding it to WHITELIST_GPL-3.0.
Anyone needing to whitelist a GPLv3 recipe should now just use
WHITELIST_GPL-3.0.
(From OE-Core rev: d4dea76fbe9765d489e3e522a9d2c22049610c7b)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If a recipe sets S to ${WORKDIR}/ then the S != WORKDIR test doesn't work as
expected. Use os.path.normpath() to normalise the paths so string comparison
works.
(From OE-Core rev: 06aaafd14f3c8e27faeea0a514f80e1ff5eb4deb)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
As introduced by a56fb90dc3 ('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]:
| do_unpack[cleandirs] += " ${S} ${STAGING_KERNEL_DIR} ${B} ${STAGING_KERNEL_BUILDDIR}"
As python anonymous functions are evaluated at the very end of recipe
parsing, the d.setVarFlag('do_unpack', 'cleandirs', '${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.
(From OE-Core rev: 20901b9783220aa6e7adae4951c531919c20859b)
Signed-off-by: Enrico Jorns <ejo@pengutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Handling of backfilling is trickier than you'd think. We need this to execute
early enough that the user will see the changes in bitbake -e and other output
yet late enough that the virtclass extensions have changed the tunes before
it executes.
It makes more sense to execute this at anonymous python time now bitbake -e
correctly handles this and that unbreaks multilib corruption of these variables.
[YOCTO #12373]
(From OE-Core rev: cefd312541019c1bb0132f41334df58971fd81ef)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Now bitbake is executing anonymous python fragments in bitbake -e,
ensure we don't show the error in that context (where PN would be
unchanged from default).
(From OE-Core rev: 79240b3af32c7fa44751752e9e8f2ff832439643)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The new name is much more consistent with what this actually means. We put
the pieces in place to rename everything a while back but looks like we
forgot to actually do it! Fix that now.
(From OE-Core rev: af9612f5d6b848fceea22d10ee964437299be776)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
.txz is .tar.xz, so add it, as this can actually be found in the
wild.
(From OE-Core rev: 58af8c2e4bd17692274fc5a6ac8f8af84319fec6)
Signed-off-by: André Draszik <git@andred.net>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Although it may not appeal so much to users to prefer 80x24 consoles,
the general trend is for screens to get bigger and the current output
has started to look a little cramped on a modern HD display.
Increasing from 17 to 20 is obviously arbitrary, but does give enough
space to cleanly display layers such as "meta-nodejs-contrib" and
"meta-virtualization" while still keeping the output fairly compact.
(From OE-Core rev: e556b85a452a4ebf34a0c9581608f135bbad2e2c)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
According to the bitbake documentation, the "d" global datastore should be
used instead of "e.data" because the latter is deprecated.
(From OE-Core rev: 43820fe158cbc34c9fd67e84238931331153f127)
Signed-off-by: Leonardo Sandoval <leonardo.sandoval.gonzalez@linux.intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Unzip is required for jar files. Without this patch:
ERROR: junit4-4.3.1-r0 do_unpack: Unpack failure for URL:
'http://downloads.sourceforge.net/junit/junit-4.3.1-src.jar'. Unpack
command
PATH="/var/lib/jenkins/workspace/qt5122-dyspro/build/tmp/sysroots-uninative/x86_64-linux/usr/bin ......."
unzip -q -o '/var/lib/yocto/downloads/junit-4.3.1-src.jar' failed with
return value 127
Suggested-by: Ross Burton <ross.burton@intel.com>
(From OE-Core rev: 96665f8b204b0f0ba571bb6634203aada5983f8a)
Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If we're unpacking any rpm (not just .src.rpm) then we'll need to call
rpm2cpio.sh. As of OE-Core rev a7da1aade118d1ccf1b286f82556cd9f706bd2a4
that script no longer uses file-native, but there is a chance that the
rpm file will be compressed internally using xz (for example, rpms from
Fedora) and therefore the script will need xzcat provided xz-native.
(From OE-Core rev: 2f31f1795bc0c85b1646bc7d9596bbe778cb84e5)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This uses the newly added MultiConfigParsed event to handle problems where
checksums in multiconfig build were not fuctioning as expected.
The issue arises around SIGGEN_EXCLUDE_SAFE_RECIPE_DEPS which contains
entries like:
* gcc-cross-${TARGET_ARCH}->virtual/${TARGET_PREFIX}libc-for-gcc
* gcc-cross-${TARGET_ARCH}->linux-libc-headers
These need to be expanded in the multiconfig data store but then placed
into the shared main datastore used by the siggen code. The only other
alternative would be a siggen instance for each multiconfig however that
seemed even more complex and invasive.
In real world usage, this issue would mean a qemux86 base config with
other armv5 and armv7 configs (e.g. beaglebone and qemuarm) would
try and build gcc-cross twice since dependencies normaly excluded
(e.g. linux-libc-headers) would now be included. This breaks sstate reuse
as well as breaking builds unless separate tmpdirs are used.
This patch adds all the entries for each multiconfig. Whilst there may
be duplicates, this shouldn't be an issue.
(From OE-Core rev: 7267e7c000c76c44d09835d4cd2bc485b6a39a2a)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It can be useful to add RRECOMMENDS to packages created, based
on certain PACKAGECONFIGs.
In particular where a package depends on certain linux kernel
infrastructure (kernel modules) which might or might not be
built as a module, being able to RRECOMMENDS instead of
RDEPENDS on the relevant packages avoids build failures in
case those modules are built statically into the kernel, i.e.
in case no package is being created for them.
Add another field to the PACKAGECONFIG syntax to achieve just
that.
(From OE-Core rev: ec96c985ce1c888c3ce3c4d964d7a106c3c88a5c)
Signed-off-by: André Draszik <adraszik@tycoint.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This code dates from distant times before we had class overrides.
The comments are also rather stale. Rewrite this code using class
overrides which makes it safer, more modern and more easily
understandable.
(From OE-Core rev: fb3b160a6bf20a601d6cecf3f06a2b71c03fa91f)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The path to where to install and find the tools copied from the host
environment is already used in a couple of places. This warrants it to
get its own variable.
(From OE-Core rev: 8164c466943ffedff399009bf5547dba4f06d6c8)
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Use the new executable argument to bb.utils.which() to ensure that the symlinks
point to executable files and not for example directories with the right name
which happened to be on $PATH.
[ YOCTO #11256 ]
(From OE-Core rev: 7f40c934c3aeb1d34f95f30b281e25a17c428fce)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adjust the message slightly to make it clearer, in particular mentioning
the HOSTTOOLS variable.
(From OE-Core rev: e022b27d189c1dcc7b3baea2b99dba8d724e6e2b)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We currently have a determinism problem in that the host tools present
in PATH can influence the build. In particular, the presence of pkg-config
on the build host can mask missing pkgconfig class dependencies.
This adds in a new HOSTTOOLS variable and then uses it to set up a directory
of symlinks to the whitelisted host tools. This directory is placed as PATH
instead of the usual /usr/bin:/bin and so on.
This should improve determinism of builds and avoid the issues which have
been particularly obvious since the introduction of recipe specific sysroots.
If users find there is a tool missing, they can extend HOSTTOOLS from a global
class or global conf file.
Right now the settings should be enough to build everything in OE-Core.
(From OE-Core rev: fa764a403da34bb0ca9fa3767a9e9dba8d685965)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The previous wording, e.g.
ERROR: wget was skipped: incompatible with license GPLv3
isn't very clear and could be taken to imply that the recipe is
incompatible with its own license.
(From OE-Core rev: c2047067daadb40ff3a944f380c10b2a56b8e080)
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Now that the datastore works dynamically we don't need the update_data calls
so we can just remove them. They're not actually done anything at all for
a while.
(From OE-Core rev: 8de0c5d3bd01919e2bf0394f9c485936d6098cec)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This patch is comparatively large and invasive. It does only do one thing, switching the
system to build using recipe specific sysroots and where changes could be isolated from it,
that has been done.
With the current single sysroot approach, its possible for software to find things which
aren't in their dependencies. This leads to a determinism problem and is a growing issue in
several of the market segments where OE makes sense. The way to solve this problem for OE is
to have seperate sysroots for each recipe and these will only contain the dependencies for
that recipe.
Its worth noting that this is not task specific sysroots and that OE's dependencies do vary
enormously by task. This did result in some implementation challenges. There is nothing stopping
the implementation of task specific sysroots at some later point based on this work but
that as deemed a bridge too far right now.
Implementation details:
* Rather than installing the sysroot artefacts into a combined sysroots, they are now placed in
TMPDIR/sysroot-components/PACKAGE_ARCH/PN.
* WORKDIR/recipe-sysroot and WORKDIR/recipe-sysroot-native are built by hardlinking in files
from the sysroot-component trees. These new directories are known as RECIPE_SYSROOT and
RECIPE_SYSROOT_NATIVE.
* This construction is primarily done by a new do_prepare_recipe_sysroot task which runs
before do_configure and consists of a call to the extend_recipe_sysroot function.
* Other tasks need things in the sysroot before/after this, e.g. do_patch needs quilt-native
and do_package_write_deb needs dpkg-native. The code therefore inspects the dependencies
for each task and adds extend_recipe_sysroot as a prefunc if it has populate_sysroot
dependencies.
* We have to do a search/replace 'fixme' operation on the files installed into the sysroot to
change hardcoded paths into the correct ones. We create a fixmepath file in the component
directory which lists the files which need this operation.
* Some files have "postinstall" commands which need to run against them, e.g. gdk-pixbuf each
time a new loader is added. These are handled by adding files in bindir with the name
prefixed by "postinst-" and are run in each sysroot as its created if they're present.
This did mean most sstate postinstalls have to be rewritten but there shouldn't be many of them.
* Since a recipe can have multiple tasks and these tasks can run against each other at the same
time we have to have a lock when we perform write operations against the sysroot. We also have
to maintain manifests of what we install against a task checksum of the dependency. If the
checksum changes, we remove its files and then add the new ones.
* The autotools logic for filtering the view of m4 files is no longer needed (and was the model
for the way extend_recipe_sysroot works).
* For autotools, we used to build a combined m4 macros directory which had both the native and
target m4 files. We can no longer do this so we use the target sysroot as the default and add
the native sysroot as an extra backup include path. If we don't do this, we'd have to build
target pkg-config before we could built anything using pkg-config for example (ditto gettext).
Such dependencies would be painful so we haven't required that.
* PKDDATA_DIR was moved out the sysroot and works as before using sstate to build a hybrid copy
for each machine. The paths therefore changed, the behaviour did not.
* The ccache class had to be reworked to function with rss.
* The TCBOOTSTRAP sysroot for compiler bootstrap is no longer needed but the -initial data
does have to be filtered out from the main recipe sysroots. Putting "-initial" in a normal
recipe name therefore remains a bad idea.
* The logic in insane needed tweaks to deal with the new path layout, as did the debug source
file extraction code in package.bbclass.
* The logic in sstate.bbclass had to be rewritten since it previously only performed search and
replace on extracted sstate and we now need this to happen even if the compiled path was
"correct". This in theory could cause a mild performance issue but since the sysroot data
was the main data that needed this and we'd have to do it there regardless with rss, I've opted
just to change the way the class for everything. The built output used to build the sstate output
is now retained and installed rather than deleted.
* The search and replace logic used in sstate objects also seemed weak/incorrect and didn't hold
up against testing. This has been rewritten too. There are some assumptions made about paths, we
save the 'proper' search and replace operations to fixmepath.cmd but then ignore this. What is
here works but is a little hardcoded and an area for future improvement.
* In order to work with eSDK we need a way to build something that looks like the old style sysroot.
"bitbake build-sysroots" will construct such a sysroot based on everything in the components
directory that matches the current MACHINE. It will allow transition of external tools and can
built target or native variants or both. It also supports a clean task. I'd suggest not relying on
this for anything other than transitional purposes though. To see XXX in that sysroot, you'd have
to have built that in a previous bitbake invocation.
* pseudo is run out of its components directory. This is fine as its statically linked.
* The hacks for wayland to see allarch dependencies in the multilib case are no longer needed
and can be dropped.
* wic needed more extensive changes to work with rss and the fixes are in a separate commit series
* Various oe-selftest tweaks were needed since tests did assume the location to binaries and the
combined sysroot in several cases.
* Most missing dependencies this work found have been sent out as separate patches as they were found
but a few tweaks are still included here.
* A late addition is that extend_recipe_sysroot became multilib aware and able to populate multilib
sysroots. I had hoped not to have to add that complexity but the meta-environment recipe forced my
hand. That implementation can probably be neater but this is on the list of things to cleanup later
at this point.
In summary, the impact people will likely see after this change:
* Recipes may fail with missing dependencies, particularly native tools like gettext-native,
glib-2.0-native and libxml2.0-native. Some hosts have these installed and will mask these errors
* Any recipe/class using SSTATEPOSTINSTFUNCS will need that code rewriting into a postinst
* There was a separate patch series dealing with roots postinst native dependency issues. Any postinst
which expects native tools at rootfs time will need to mark that dependency with PACKAGE_WRITE_DEPS.
There could well be other issues. This has been tested repeatedly against our autobuilders and oe-selftest
and issues found have been fixed. We believe at least OE-Core is in good shape but that doesn't mean
we've found all the issues.
Also, the logging is a bit chatty at the moment. It does help if something goes wrong and goes to the
task logfiles, not the console so I've intentionally left this like that for now. We can turn it down
easily enough in due course.
(From OE-Core rev: 809746f56df4b91af014bf6a3f28997d6698ac78)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
getVar() now defaults to expanding by default, thus remove the True
option from getVar() calls with a regex search and replace.
Search made with the following regex: getVar ?\(( ?[^,()]*), True\)
(From OE-Core rev: 7c552996597faaee2fbee185b250c0ee30ea3b5f)
Signed-off-by: Joshua Lock <joshua.g.lock@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With the changes to the code for extracting source for a recipe, we are
properly executing the tasks for a recipe, which means their stamps (and
therefore signatures) are important. When running devtool extract on
the lsof recipe I noticed that do_fetch and do_unpack were executing a
second time when we called for do_patch, and this turned out to be
because LIC_FILES_CHKSUM in that recipe contains an entry which
is an absolute path (has ${S} at the start). Normally this wouldn't be
an issue since S is under TMPDIR and thus the existing code would ignore
it, however devtool's extraction code extracts to a temporary directory
which is not under TMPDIR; the result was the path to this file was not
being ignored and the second time around when the license file had been
extracted it was incorporated into the signature. We don't want this, so
explicitly exclude S as well as B and WORKDIR for good measure.
(From OE-Core rev: 1c99d74a862f25e23ea6465fab7ddc9ce74d6974)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This sets a good example and avoids unnecessarily contributing to
perceived complexity and cargo culting.
Motivating quote below:
< kergoth> the *original* intent was for the function/task to error via
whatever appropriate means, bb.fatal, whatever, and
funcfailed was what you'd catch if you were calling
exec_func/exec_task. that is, it's what those functions
raise, not what metadata functions should be raising
< kergoth> it didn't end up being used that way
< kergoth> but there's really never a reason to raise it yourself
FuncFailed.__init__ takes a 'name' argument rather than a 'msg'
argument, which also shows that the original purpose got lost.
(From OE-Core rev: 9635af9785509a39c1ac2509740d46276119a0ca)
Signed-off-by: Ulf Magnusson <ulfalizer@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
