meson will by default try to detect the presence of ccache, and if
found, will use it unconditionally.
However, using a system-wide ccache, which would be using our own cache
directory, may very well conflict with our own ccache.
But there is no option to disable that meson behaviour. The only
workaround that is even the official documented way to do so, is to
actually pass environment variables that point to the compiler:
https://mesonbuild.com/Feature-autodetection.html#ccache
For the host variants, we pass $(HOST_CONFIGURE_OPTS) in the environment,
and this contains correct settings for CC and CXX, so meson does not try
and detect ccache; it uses exactly what we tell it to use.
For the target variant, the settings for the cross-compiler are defined
in the cross-compilation file, and so meson just abides by our will. But
for the compiler-for-build, there is no way to specify the CC_FOR_BUILD
or CXX_FOR_BUILD via a cross-compilation file:
https://mesonbuild.com/Machine-files.htmlhttps://mesonbuild.com/Cross-compilation.html
We could pass the full TARGET_CONFIGURE_OPTS in the environment, like we
do for the host variant, but this contains a lot more variables that are
supposed to be covered by the cross-compilation file.
So, we stay safe and just provide the exact two variables that meson
will use to avoid detecting ccache.
If the current configuration defines the use of ccache, then these two
variables will be properly setup to use our own ccache.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Gleb Mazovetskiy <glex.spb@gmail.com>
Cc: James Hilliard <james.hilliard1@gmail.com>
Cc: Norbert Lange <nolange79@gmail.com>
(cherry picked from commit f4a61d1ae2)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
As reported on IRC by sephthir, the qemu_sparc_ss10_defconfig doesn't
work as expected: the system generated when booted under Qemu produces
illegal instruction messages.
gcc 8.3, 9.2 are the latest working gcc version. git bisect between
gcc 8.3 and 8.4 allowed to identify the commit that introcuced the
regression.
Reverting this patch allowed to produce a working rootfs.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/786589934
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 4d16e6f532)
[Peter: drop GCC 10.x patch]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since commit dfcc18f84b cmake-package
_INSTALL_STAGING_OPTS/_INSTALL_TARGET_OPTS use 'install/fast'
instead of 'install', adjust documentation accordingly.
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 0c80a0da9a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
- Fixes CVE-2021-23239, a potential information leak in sudoedit that
could be used to test for the existence of directories not normally
accessible to the user in certain circumstances. When creating a new
file, sudoedit checks to make sure the parent directory of the new
file exists before running the editor. However, a race condition
exists if the invoking user can replace (or create) the parent
directory. If a symbolic link is created in place of the parent
directory, sudoedit will run the editor as long as the target of the
link exists. If the target of the link does not exist, an error
message will be displayed. The race condition can be used to test for
the existence of an arbitrary directory. However, it cannot be used to
write to an arbitrary location.
- Fixes CVE-2021-23240, a flaw in the temporary file handling of
sudoedit's SELinux RBAC support. On systems where SELinux is enabled,
a user with sudoedit permissions may be able to set the owner of an
arbitrary file to the user-ID of the target user. On Linux kernels
that support protected symlinks setting
/proc/sys/fs/protected_symlinks to 1 will prevent the bug from being
exploited. For more information, see Symbolic link attack in
SELinux-enabled sudoedit.
- Update license hash:
- copyright of python bindings added with
6c1b155fed
- a few other files (ISC licenced) added with
d4b2db9078
- year updated with
9e111eae57
- Update indentation in hash file (two spaces)
https://www.sudo.ws/stable.html#1.9.5p1
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 1b14e99666)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Using absolute paths within getdeveloperlib isn't very sensible, it
makes a lot more sense to handle everything as relative paths from the
top-level Buildroot source directory.
parse_developers() is changed to no longer take the base path as
argument: it is automatically calculated based on the location of
utils/getdeveloperlib.py. Then, the rest of the logic is adjusted to
use relative paths, and prepend them with the base "brpath" when
needed.
This commit allows pkg-stats to report correct developers information
even when executed from an out of tree directory.
Before this patch:
$ ~/buildroot/support/scripts/pkg-stats -p ipmitool --json out.json
$ cat out.json | jq '.packages.ipmitool.developers'
[]
$ cat out.json | jq '.defconfigs.stm32f469_disco'
{
"name": "stm32f469_disco",
"path": "configs/stm32f469_disco_defconfig",
"developers": []
}
After this patch:
$ ~/buildroot/support/scripts/pkg-stats -p ipmitool --json out.json
$ cat out.json | jq '.packages.ipmitool.developers'
[
"Floris Bos <bos@je-eigen-domein.nl>",
"Heiko Thiery <heiko.thiery@gmail.com>"
]
$ cat out.json | jq '.defconfigs.stm32f469_disco'
{
"name": "stm32f469_disco",
"path": "configs/stm32f469_disco_defconfig",
"developers": [
"Christophe Priouzeau <christophe.priouzeau@st.com>"
]
}
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Tested-by: Heiko Thiery <heiko.thiery@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 40bb37bd70)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
With Python 3.8, the following deprecation warnings are emitted:
/home/thomas/projets/buildroot/./support/scripts/pkg-stats:418: DeprecationWarning: The explicit passing of coroutine objects to asyncio.wait() is deprecated since Python 3.8, and scheduled for removal in Python 3.11.
/home/thomas/projets/buildroot/./support/scripts/pkg-stats:536: DeprecationWarning: The explicit passing of coroutine objects to asyncio.wait() is deprecated since Python 3.8, and scheduled for removal in Python 3.11.
The correct way to pass coroutines is to use asyncio.create_task(),
but this is rather new method (Python 3.7), and using it breaks
compatibility with older Python versions. As suggested at
https://docs.python.org/3/library/asyncio-task.html#asyncio.create_task,
use the more cryptic, but also more compatible asyncio.ensure_future()
method.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit ffb2620405)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The cve module needs ijson, which may not be installed. Since cve
matching is only enabled when --nvd-path is passed, it is a bit silly
to error out about ijson being missing if it's not used.
So instead of unconditionally importing the cve module, only do it
conditionally.
However, instead of doing it right at the point where it is used, we
do it at the beginning of the main() function. Indeed, if the cve
module is needed but cannot be imported, we want to error out
immediately rather than doing a whole bunch of things, and failing on
the user later on in the middle of the pkg-stats execution.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Tested-by: Heiko Thiery <heiko.thiery@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 824032d168)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When a package has both a target and a host variant, and uses git
submodules, and the host variant is downloaded before the target one, we
end up with the generated archive missing the submodules.
This happens in exactly one package in our tree: c-capnproto.
This issue was not caught before because after a few days, the full
sources are added to sources.buildroot.net. So when the hash check
fails, the full tarball is simply downloaded from there.
Propagate the git submodule setting from the target variant to the host
variant, unless the host variant explicitly opted-out.
Fixes:
http://autobuild.buildroot.org/results/2de9c6c8ce83569d18cc7140ebc60d6fe1aadcbf/
Reported-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
(cherry picked from commit 4d03923b6a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The LLVM project has switched to using a monorepo to host all their
components. The separate, individual repositories have been closed
late 2020 / early 2021. The libclc repository is no longer.
Switch to using the libclc source from the llvm legacy and frozen
mirror.
Even though we could switch over to using the github helper, we just
keep using the git download method: it is a small repository, and it
will not impact people that were already using it.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Romain Naour <romain.naour@gmail.com>
Cc: Valentin Korenblit <valentinkorenblit@gmail.com>
Cc: Michael Opdenacker <michael.opdenacker@bootlin.com>
Acked-by: Romain Naour <romain.naour@gmail.com>
(cherry picked from commit 971d1ea7ac)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The following commits:
- 7868289fd5 package/zic: bump version to 2020f
- c99374ecbb package/tzdata: bump version to 2020f
bumped the tzdata from version 2020a to 2020f. However, in 2020b, the
zic option '-y' was removed, and so was the yearistype.sh script [0].
This now spews annoying warnings:
warning: -y ignored
Fortunately, it still consumes its argument, so the missing yearistype.sh
is simply ignored.
Drop that option.
[0] https://mm.icann.org/pipermail/tz-announce/2020-October/000059.html
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Bernd Kuhls <bernd.kuhls@t-online.de>
(cherry picked from commit 7f78eef767)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issues:
- CVE-2020-8265: use-after-free in TLSWrap (High) Affected Node.js versions
are vulnerable to a use-after-free bug in its TLS implementation. When
writing to a TLS enabled socket, node::StreamBase::Write calls
node::TLSWrap::DoWrite with a freshly allocated WriteWrap object as first
argument. If the DoWrite method does not return an error, this object is
passed back to the caller as part of a StreamWriteResult structure. This
may be exploited to corrupt memory leading to a Denial of Service or
potentially other exploits
- CVE-2020-8287: HTTP Request Smuggling in nodejs Affected versions of
Node.js allow two copies of a header field in a http request. For
example, two Transfer-Encoding header fields. In this case Node.js
identifies the first header field and ignores the second. This can lead
to HTTP Request Smuggling
- CVE-2020-1971: OpenSSL - EDIPARTYNAME NULL pointer de-reference (High)
This is a vulnerability in OpenSSL which may be exploited through Node.js.
You can read more about it in
https://www.openssl.org/news/secadv/20201208.txt
Update the license hash for the addition of the (MIT licensed)
cjs-module-lexer module:
9eb1fa1924
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit ecc8f0fe84)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes build error
output/host/opt/ext-toolchain/bin/../lib/gcc/aarch64-amd-linux-gnu/4.9.1/../../../../aarch64-amd-linux-gnu/bin/ld:
cannot find -latomic
using this defconfig
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_CODESOURCERY_AARCH64=y
BR2_PACKAGE_OPENSSL=y
libopenssl is only used here as an example: all packages adding -latomic
if BR2_TOOLCHAIN_HAS_LIBATOMIC=y are broken, like dav1d, ffmpeg, gnutls,
kodi and vlc.
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit a530fd4a42)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issue:
- CVE-2020-35512: On Unix, avoid a use-after-free if two usernames have the
same numeric uid. In older versions this could lead to a crash (denial of
service) or other undefined behaviour, possibly including incorrect
authorization decisions if <policy group=...> is used. Like Unix
filesystems, D-Bus' model of identity cannot distinguish between users of
different names with the same numeric uid, so this configuration is not
advisable on systems where D-Bus will be used.
Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit c80989aa9d)
[Peter: mention security fix]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit d82da39d55 forgot to move
BR2_PACKAGE_CA_CERTIFICATES condition under BR2_PACKAGE_LIBTASN1 to
avoid the following build failure:
checking if trust module is enabled... configure: WARNING: --with-libtasn1 is needed in order to build the trust module, disabling
no
checking for trust module paths... configure: error: need --enable-trust-module in order to specify trust module paths.
Fixes:
- http://autobuild.buildroot.org/results/e7f68205e1b776f9af34e6017f6eb17f46aa2f19
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 0ad12b3a6a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since its introduction in Buildroot in 2013 with commit 07203d78c2
(trace-cmd: new package), trace-cmd has declared installing in staging.
But trace-cmd is a generic-package, and has never, ever provided any
commands for staging installation.
Drop this declaration.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 896b93310a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
