| Age | Commit message (Collapse) | Author | Files |
|---|
|
During spec generation, ideally directories should not be auto
packaged under the %file section of rpm packages but take ownership of
specific directories.
* packages only empty directories or explict directory.
See:
- http://www.rpm.org/max-rpm/s1-rpm-inside-files-list-directives.html
- "The %dir Directive"
* This will prevent the overlapping of security permission.
For example, in Tizen the directory /etc have smack label 'System::Shared'
So Only one package should own and set the label of /etc to prevent
the overwriting of the smack label.
Existing behaviour is maintained if DIRFILES is not set. If it is set,
the modified behaviour is used. If can be set to an empty value by
core recipes to trigger the modified behaviour.
[RP: Modified to allow optional usage of DIRFILES]
Signed-off-by: Ronan Le Martret <ronan@fridu.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This allows its usage in other RPM macros so files in ${S} can be found.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Add hooks to allow customisation of the rpm spec files. Since python functions
aren't visible in the data store, one variable is used to trigger the call to
a separately named function. A dummy function is not provided since this then
triggers various class ordering complexities which are best avoided.
Ultimately this will be replaced by a refactor of the code to generate the
spec file using a python class.
This allows the tizen layer to add hooks for the security manifests for
example.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When building target packages, HOST_OS and TARGET_OS are the same, as is the
VENDOR field. However, when building an SDK this is not true. The patch
corrects the oversight and switches to using the 'HOST' version and resolves
the issue of meta-mingw not working w/ the rpm packaging.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The meta-mingw layer attempts to change the SDK Suffix, but the rpm
packaging had a hard coded reference to _nativesdk.
I did a quick scan for other hard coded entries and did not fine any
more.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The package_write task was previously removed. Remove a remaining superfluous
reference to it.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The archiver.bbclass will put the sources to ARCHIVER_OUTDIR according
to configuration, then the rpmbuild -bs will create the srpm.
[YOCTO #4986]
[YOCTO #5113]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
|
|
The reasons this task was introduced are lost in the mists of time. It
allowed for the a single "package_write" task instead of spelling out
the explicit package backends, however in all but one case we do that
anyway.
As such as might as well give in and delete the task, converting that
single reference into explicit dependencies.
This gives bitbake a bit less work to to when processing the runqueue
since there are less tasks (but more dependencies in some cases).
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Package indexing is done in python and package-index.bb uses the new
routines.
Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The default value for HOMEPAGE of "unknown" has been in place since the
early OE-Classic days, but it doesn't really make sense - "unknown" is
not a valid URL and it just means we have to explicitly check for this
hardcoded string if we're displaying the value in some form of UI, such
as Toaster.
This has required some changes to the packaging classes as they
previously did not expect the value to be blank.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This commit cleans up the functions that were ported to python.
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
|
|
This causes issues when postinstalls have ERROR keywords
its interpreted as error and image build is cancelled
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
Additional metadata from user-defined variable is written into
control/spec file of binary package.
Three variables are searched for adiitional package metadata:
* PACKAGE_ADD_METADATA_<PKGTYPE>_<PN>
* PACKAGE_ADD_METADATA_<PKGTYPE>
* PACKAGE_ADD_METADATA
First found variable with defined value wins.
<PN> is a package name. <PKGTYPE> is a distinct name of specific
package type:
* IPK for .ipk packages
* DEB for .deb packages
* RPM for .rpm packages
Variable can contain multiple [one-line] metadata fields separated by
literal sequence '\n'. Separator can be redefined through variable flag
'separator'. In package control/spec file separator is replaced by
newline character.
Signed-off-by: Leonid Borisenko <ive.found@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
We have the odd situation where the CONTROL/DEBIAN directory can be removed
in the middle of the walk, the isdir() test would then fail and the walk code
would assume its a file hence we check for the names in files too.
This resolves the autobuilder failure:
error: File not found: /home/pokybuild/yocto-autobuilder/yocto-slave/nightly-fsl-arm/build/build/tmp/work/armv7a-vfp-neon-poky-linux-gnueabi/xinit/1_1.3.3-r0/package/DEBIAN
RPM build errors:
File not found: /home/pokybuild/yocto-autobuilder/yocto-slave/nightly-fsl-arm/build/build/tmp/work/armv7a-vfp-neon-poky-linux-gnueabi/xinit/1_1.3.3-r0/package/DEBIAN
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When originally developed, it was thought a task may have more than one associated
sstate archive. The way the code has grown that idea is now not possible or needed.
We can therefore assume one sstate archive per task and drop the crazy name
mapping code. Simpler is better in this case.
The downside is that various sstate archives will change name so this forces a cache
rebuild. Given the other sstate changes going in at this time, this isn't really
a bad thing as things would rebuild anyway.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
On a multilib system when one of the multibs has a different OS then
other multilibs a failure can occur during the install process because
RPM assumes all systems have the same OS.
When an n32 platform is selected as an alternative multilib, it shows
up as mips64_n32-.*-linux-gnun32 in /etc/rpm/platform. This causes
problems when the smart tool tries to add a channel for the multilib.
RPM archScore call always returns zero for arch "mips64_n32" -
after appending default vendor and os, it finds "mips64_n32-wrs-linux"
doesn't match any predefined platforms. Fix this by removing the
restriction of -gnun32 suffix in platform file.
Signed-off-by: Lei Liu <lei.liu2@windriver.com>
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Jeff Polk <jeff.polk@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Currently if multiple package tasks are running against a recipe, package_rpm
is restricted to the slowest speed of them due to the locking. This patch
explicitly ignores the opkg/debian artefacts and hence allows a speedup.
It also removes an issue were a Ctrl+C interrupting a deb.ipk packaging task
would end up with CONTROL/DEBIAN files in the spec file resulting in a build
failure.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The postinst files were being generated using purely the name of the package,
this unfortunately meant the run order would be based on the name of the
package and not the order in which it was installed on the filesystem.
If package A requires package Z to be fully installed, this causes a problem.
Note:
rpm - as the rpm based install proceeds the order is defined and captured.
so the problem is resolved there.
ipk - this unfortunately does not appear to solve the problem for ipk, as
the status file is not ordered in any appreciable way. This does not
cause any regressions however and sets the stage for a proper fix.
deb - this -may- fix the deb install. Early testing indicates at least some
ordering to the status file. But it's unclear if it completely resolves
the issue.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
[YOCTO #5313]
When performing an attemptonly install, we should skip an errors and
continue to install everything else. However, there is a case where two
packages can conflict, and cause a hard failure.
This workaround, ignores this and allows the image to be constructed.
Note: Some items in the failed transaction may not get installed.
To fix this properly we need to find the issue in smart, and make it ignore
or at least attempt to resolve these kinds of conflicts.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The function that "bitbake package-index" relies upon when using the RPM
package backend (package_update_index_rpm()) uses MULTILIB_PREFIX_LIST
to get the list of package architectures to be indexed, but that
variable is only set when populate_sdk_rpm or rootfs_rpm are inherited,
which is not the case for the package-index recipe. Until we're able to
refactor this properly, for minimal impact just use the value of
ALL_MULTILIB_PACKAGE_ARCHS if MULTILIB_PREFIX_LIST does not give us any
architectures (the equivalent function in the ipk backend uses the
former variable).
Having "bitbake package-index" working is important because it's the
only practical way of indexing RPM packages for use as a feed; host
versions of createrepo won't work properly because they won't support
indexing recommends relationships.
Stopgap fix for [YOCTO #5278].
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
directory
Currently we have a hierarchy of pkgdata directories and the code has to put together
a search path and look through each in turn until it finds the data it needs.
This has lead to a number of hardcoded paths and file globing which
is unpredictable and undesirable. Worse, certain tricks that should be
easy like a GL specific package architecture become problematic with the
curretn search paths.
With the modern sstate code, we can do better and construct a single pkgdata
directory for each machine in just the same way as we do for the sysroot. This
is already tried and well tested. With such a single directory, all the code that
iterated through multiple pkgdata directories and simply be removed and give
a significant simplification of the code. Even existing build directories adapt
to the change well since the package contents doesn't change, just the location
they're installed to and the stamp for them.
The only complication is the we need a different shlibs directory for each
multilib. These are only used by package.bbclass and the simple fix is to
add MLPREFIX to the shlib directory name. This means the multilib packages will
repackage and the sstate checksum will change but an existing build directory
will adapt to the changes safely.
It is close to release however I believe the benefits this patch give us
are worth consideration for inclusion and give us more options for dealing
with problems like the GL one. It also sets the ground work well for
shlibs improvements in 1.6.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Not only outdir had the wrong value, it wasn't used actually used in that function.
Signed-off-by: Stefan Stanacar <stefanx.stanacar@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
[ YOCTO #3723 ]
Add a mode to smart that will allow an installation to continue, instead of
failure in the case that one or more items is uninstallable.
Uninstallable packages are simply ignored, and no error is generated.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When the code was refactored to address review comments, the wrong version
was sent to the community. Replace the $1 with ${target_rootfs}
Fix identified by: Yue Tao <yue.tao@windriver.com>
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
These have been deprecated for a long time, convert the remaining
references to the correct modules and prepare for removal of the
compatibility support from bitbake.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Add NO_RECOMMENDATIONS support. A way to disable all recommended
packages from being installed. This will help shrink the size of
the resulting filesystem.
Add documentation on NO_RECOMMENDATIONS and BAD_RECOMMENDATIONS.
Note, using NO_RECOMMENDATIONS has side effects such that kernel-modules may
not have been installed. A user will need to manually add to their image
any kernel-modules required to be on the image for functionality.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
Using the new smart exclude mechanism an error will be generated in the
excluded package is required for the image to be generated.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
* usually it's more important to know how much space will each
package take on target device then size of compressed package
* example for libewebkit0 with 4 different architectures, interesting
that om_gta02 .ipk is bigger but it's smaller when installed
before:
MACHINE DEFAULTTUNE SIZE (.ipk file)
om_gta04 cortexa8t-neon 15996 KiB libewebkit0
qemux86_64 x86-64 16992 KiB libewebkit0
spitz xscale 16148 KiB libewebkit0
om_gta02 arm920t 16260 KiB libewebkit0
after:
MACHINE DEFAULTTUNE SIZE (installed)
om_gta04 cortexa8t-neon 60544 KiB libewebkit0
qemux86_64 x86-64 63720 KiB libewebkit0
spitz xscale 60588 KiB libewebkit0
om_gta02 arm920t 56268 KiB libewebkit0
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Add support for the BAD_RECOMMENDATIONS variable that can be used to
prevent specific packages from being installed via an RRECOMMENDS
relationship when using the RPM backend. (Previously this
functionality was only available when using ipk packaging.)
In the process this moves the defaulting of BAD_RECOMMENDATIONS (as
empty) to bitbake.conf since it is no longer specific to the ipk
backend, as well as unifying some of the code that creates the
configuration for smart for use on the host and target.
Fixes [YOCTO #3916].
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
If the spec file already exist, and has not been stored into pseudo's files.db,
it maybe cause rpmbuild src.rpm fail, so remove it before doing rpmbuild src.rpm.
Signed-off-by: Bian Naimeng <biannm@cn.fujitsu.com>
|
|
The recipe's DESCRIPTION is wrapped automatically by textwrap, make it
support newline ("\n") to let the user can wrap it manually, e.g.:
DESCRIPTION = "Foo1\nFoo2"
In the past, it would be:
Foo1\nFoo2
Now:
Foo1
Foo2
[YOCTO #4348]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
Currently the scriptlet_wrapper is designed to handle only pre/post
install scriptlets. This patch will slightly change the wrapper script
to handle also pre/post remove scriptlets when we want to remove
packages from the final image.
Signed-off-by: Laurentiu Palcu <laurentiu.palcu@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
Back in 2010, I added these in commit: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=36f1ae42fe13dae174b7fb5eb85dc49d7d7b516b
however the package_write tasks sstate only consists of package files and no
fakeroot privileges are needed to write these out, only originally create them.
We can therefore drop these for some small performance gains and a less
convoluted depenency chain.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The bb.fatal() is defined as:
def fatal(*args):
logger.critical(''.join(args))
sys.exit(1)
So anything after bb.fatal() in the same code block doesn't have any
effect, e.g.:
bb.fatal("%s_%s: %s" % (var, pkg, e))
raise e
The "raise e" should be removed.
I searched all the files which use bb.fatal(), only the following 3
classes have this issues:
insane.bbclass
package.bbclass
package_rpm.bbclass
[YOCTO #4461]
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
There are various bits of cruft that have built up around our file accesses. This patch
cleans some of them up, specifically:
* Remove pointless "from __builtin__ import file"
* Use open(), not file()
* Wrap file usage in a with container to ensure files are closed
* Add missing .close() calls in some cases
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
1, In bitbake.conf
PKGR ?= "${PR}${EXTENDPRAUTO}"
EXTENDPKGV ?= "${EXTENDPKGEVER}${PKGV}-${PKGR}"
RDEPENDS_${PN}-dev = "${PN} (= ${EXTENDPKGV})"
2, When PR Server is enabled, EXTENDPRAUTO is not none which means PKGR and PR
don't have the same value.
3, When multilib is enabled, RDEPENDS_${PN}-dev is not expanded correctly
which uses PR rather than PKGR in the versioned dependency string.
4, Make sure PKGR rather than PR in version string when do_package_rpm.
[YOCTO #4050]
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This fixes build failures introduced with "classes/buildhistory: implement history
collection for SDK" by moving the functions to files where only the specific image
type which is enabled is inherited. The failures occured when multiple PACKAGE_CLASSES
were enabled.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
SDKs are constructed in a similar manner to images, and the contents can
be influenced by a number of different factors, thus tracking the
contents of produced SDKs when buildhistory is enabled can help detect
the same kinds of issues as with images.
This required adding POPULATE_SDK_POST_HOST_COMMAND and
SDK_POSTPROCESS_COMMAND variables so that data collection functions can
be injected at the appropriate points in the SDK construction process,
as well as moving the list_installed_packages and
rootfs_list_installed_depends functions from the rootfs_{rpm,ipk,deb} to
the package_{rpm,ipk,deb} classes so they can also be called during
do_populate_sdk as well as do_rootfs.
Implements [YOCTO #3964].
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The tclibc file for uclibc already empties IMAGE_LINGUAS, so there's no point
to this conditional as far as I can tell, and it can cause issues for certain
values of TARGET_OS.
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The previous change to this function fixed one manipuation at
the expense of the other, depending on ow expanded the versioned
dependency string was when the AUTOINC substitution was made.
This update ensures we cover both cases and ensures the classextended
version works as as well as the normal case.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
If a recipe has versioned dependencies on another package within
the same recipe, there are potentially races where the version
remapping may not happen correctly.
This issue triggered with neard in multilib builds since it
uses a "-" character in its PV which is illegal in an rpm version
field. The remapping to "+" was not occuring.
It only triggers in the multilib case since in this case, expansion
of the datastore happens at slightly different points.
The correct fix is to search for PV, not PKGV but substitute the
PV value.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Currently the process for injecting the per file rpm dependencies into
rpmbuild is painfully slow. Its done through the repeated execution of
a script which has to return the correct value in each case. This continual
execution means the CPU usage of rpmbuild is low.
This patch allows the option of collapsing the per file dependencies to
a per package basis and injecting them through the .spec file. This removes
the execution overhead and allows rpmbuild to run at 100% of cpu.
Ultimately it would be nice to inject the per file dependencies through
the .spec file however that is not currently possible.
Since few people use the per file dependency information, this patch
goes for the faster approach. It can be enabled if anyone needs it although
I'd mention that its being used to us as this code may well go away in
the future if nobody complains.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When the OE arch is of the format "foo_bar-foobar" the previous
comparison routine did not selectively translate the '-' causing
a failed comparison.
In order to work around this issue, we -always- compare the
RPM translated version of the package architectures.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When write rpm spec file, one description line
"# pkgname - script_type"
is added to pre/post scriptlets for base package but no such line
writted for subpackage.
Add similiar line to subpackage to facilitate handling the pre/post
scriptlets.
Signed-off-by: Kang Kai <kai.kang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When create rootfs, some post install scripts need be run on device.
They are saved under directory /etc/rpm-postinst and named with numbers
such as 100, 101 etc.
Update to name the postinst script file with its package name instead of
just number. That may be more easy to debug when there is a error.
[YOCTO #3218]
Signed-off-by: Kang Kai <kai.kang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
[ CQID: WIND00404082 ]
It is possible for the TARGET_OS to change dependending on the multilib used
for a given package build. mips64 has two potential TARGET_OS values: linux
and linux-gnun32. The RPM and Smart setup needs to distinguish between the
two otherwise packages of the "non-default" TARGET_OS may be ignored as
incompatible.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The package_xxx_install functions date from a different era and are not used by
anything. In the rpm case, they're simply unimplemented, in the tar case they're
using broken whitespace and deprecated functions. We might as well clean
out the old broken unused code.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
package index
If for whatever reason the package directory is empty of packages, it
makes sense to error early rather than later in what become much
more obtuse errors. This adds in a sanity check to each of the packaging
backends. It also removes the duplicate createrepo call since the
core index creation function now uses this directly after the switch
to smart.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
rerunning task
If we don't do this, stale files can build up, particularly with the PR
server.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Currently, do_rootfs has a dependency on all the do_package output being present
due to its usage of the pkgdata directories. This means that if you run:
bitbake xxxx-image -c rootfs
you end up having to fetch and unpack all the do_package data which is usually
large and inefficient. It also means rm_work has to leave all the do_package
data lying around so rootfs works.
This patch splits the actual creation of the pkgdata directory off into a separate
task, "packagedata" which happens immediately after do_package. We can then remap
the dependencies so this task is depended upon, not do_package. Sstate can then be
programmed not to require do_package at the appropriate times.
Whilst this patch doesn't do so, it opens the possibility of rm_work wiping
out the do_package output from WORKDIR as long as it also removed the do_package
stamp (both normal and setscene variants) and allowing more space savings
with rm_work which has been regularly requested.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
