GNU Octave changed its detection of readline library in [1]. This
commit was first included in version 8.1.0.
GNU Octave was updated to 8.1.0 in Buildroot in commit b36e4b10f3
"package/octave: bump to version 8.1.0".
Since this commit, Octave can fail to find readline automatically in
some specific situations. For example, when host system is Fedora 39
and the host "readline-devel" package is installed (see detailed
explanation below).
Octave is now using a m4 macro from gnulib to detect readline.
See [2].
This macro is calling AC_LIB_LINKFLAGS_BODY([readline]). Note that
this macro will look into $libdir and $includedir by default. See [3].
Buildroot is calling target autotools configure command with
--prefix=/usr and --exec-prefix=/usr arguments. See [4].
Autotools derives libdir='${exec_prefix}/lib' and
includedir='${prefix}/include'.
Finally, gnulib will also search automatically into alternate library
directories (i.e. lib32, lib64). See [5].
All of this will make the configure script searching the readline
library by default (i.e. if the library prefix is not provided) into
the host "/usr/lib", "/usr/lib32" and "/usr/lib64", when configuring
for target.
This issue is not happening on the Buildroot docker reference image,
because the package "libreadline-dev" is not present in this image.
Even if the package "libreadline-dev" is installed on a Debian based
host systems, the issue is still not happening because libraries are
installed in the path "/usr/lib/x86_64-linux-gnu", which is not
searched by gnulib macros.
On host systems which installs libraries into one of the
"/usr/lib{,32,64}" directories, the Octave configuration script will
fail, because it will detect the host library and try to link against
it with target architecture and compilation flags and will fail. Since
the --enable-readline configure option is present, the configuration
script will fail because it cannot find a working readline library.
This can be seen in the octave configuration log, in file:
output/build/octave-8.4.0/config.log
configure:73671: checking for readline
configure:73705: /buildroot/output/host/bin/aarch64-none-linux-gnu-gcc -o conftest -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -g0 -D_FORTIFY_SOURCE=1 -pthread -fopenmp -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 conftest.c -lpthread -lm /usr/lib64/libreadline.so >&5
/buildroot/output/host/opt/ext-toolchain/bin/../lib/gcc/aarch64-none-linux-gnu/13.2.1/../../../../aarch64-none-linux-gnu/bin/ld: /usr/lib64/libreadline.so: error adding symbols: file in wrong format
collect2: error: ld returned 1 exit status
This situation can be reproduced on a Fedora 39 x86_64 host system,
with the "readline-devel" package installed. Note: uninstalling the
"readline-devel" will work around the issue.
The issue can be reproduced with a Buildroot configuration such as:
cat > .config <<EOF
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_PACKAGE_OCTAVE=y
BR2_PACKAGE_READLINE=y
EOF
make olddefconfig
make
In order to avoid those host/target readline detection mix-ups, the
readline search prefix need to be explicitly passed during octave
configuration. This commit adds this search prefix to fix this build
issue.
Fixes:
checking for readline... (cached) no
checking for readline/readline.h... (cached) yes
checking for readline/history.h... (cached) yes
configure: WARNING: I need GNU Readline 4.2 or later
configure: error: this is fatal unless you specify --disable-readline
[1] 3645c78658
[2] https://git.savannah.gnu.org/cgit/gnulib.git/tree/m4/readline.m4?id=2cdc1bafb20b187ad067056e090fcb4396ed9099
[3] https://git.savannah.gnu.org/cgit/gnulib.git/tree/m4/lib-link.m4?id=2cdc1bafb20b187ad067056e090fcb4396ed9099#n190
[4] https://gitlab.com/buildroot.org/buildroot/-/blob/2023.11/package/pkg-autotools.mk#L175
[5] https://git.savannah.gnu.org/cgit/gnulib.git/tree/m4/lib-prefix.m4?id=2cdc1bafb20b187ad067056e090fcb4396ed9099#n276
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
(cherry picked from commit 02df283415)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
strace needs headers >= 5.0 and
9966a05c7b
to avoid the following build failure on musl and aarch64 due to headers
conflict raised at least since bump to version 6.0 in commit
544806bfd8:
In file included from /home/autobuild/autobuild/instance-7/output-1/host/aarch64_be-buildroot-linux-musl/sysroot/usr/include/signal.h:48,
from strace.c:17:
/home/autobuild/autobuild/instance-7/output-1/host/aarch64_be-buildroot-linux-musl/sysroot/usr/include/bits/signal.h:18:16: error: redefinition of 'struct sigcontext'
18 | typedef struct sigcontext {
| ^~~~~~~~~~
In file included from /home/autobuild/autobuild/instance-7/output-1/host/aarch64_be-buildroot-linux-musl/sysroot/usr/include/asm/ptrace.h:26,
from /home/autobuild/autobuild/instance-7/output-1/host/aarch64_be-buildroot-linux-musl/sysroot/usr/include/linux/ptrace.h:107,
from ptrace.h:33,
from strace.c:16:
/home/autobuild/autobuild/instance-7/output-1/host/aarch64_be-buildroot-linux-musl/sysroot/usr/include/asm/sigcontext.h:28:8: note: originally defined here
28 | struct sigcontext {
| ^~~~~~~~~~
Fixes:
- http://autobuild.buildroot.org/results/74a480aa76970b36dcd890d9bd7a9df1d49e8e16
- http://autobuild.buildroot.org/results/79ee8ae5fb9712fd874b56e836eca1b997c50cd9
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
[Peter: fix architecture conditional]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit e2e8e655fe)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Group _CONF_{ENV,OPTS} together after the autreconf ones.
Keep AUTORECONF=YES close to AUTORECONF_OPTS
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit edde351cb8)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
A vulnerability was found in obgm libcoap 4.3.4. It has been rated as
critical. Affected by this issue is the function get_split_entry of the
file src/coap_oscore.c of the component Configuration File Handler. The
manipulation leads to stack-based buffer overflow. The attack may be
launched remotely. The exploit has been disclosed to the public and may
be used. It is recommended to apply a patch to fix this issue.
VDB-252206 is the identifier assigned to this vulnerability.
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 9002b818be)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fix CVE-2024-22667: Vim before 9.0.2142 has a stack-based buffer
overflow because did_set_langmap in map.c calls sprintf to write to the
error buffer that is passed down to the option callback functions.
Update hash of README.txt (version number updated with
b4ddc6c11e)
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit ef65c4f90b)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fix CVE-2024-25062: An issue was discovered in libxml2 before 2.11.7 and
2.12.x before 2.12.5. When using the XML Reader interface with DTD
validation and XInclude expansion enabled, processing crafted XML
documents can lead to an xmlValidatePopElement use-after-free.
https://gitlab.gnome.org/GNOME/libxml2/-/blob/v2.11.7/NEWS
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
libheif raises the following build failure with libjpeg since bump to
version 1.17.5 in commit 5714126eda and
ebd13a20b8
/home/buildroot/autobuild/instance-3/output-1/build/libheif-1.17.5/libheif/plugins/encoder_jpeg.cc: In function 'heif_error jpeg_encode_image(void*, const heif_image*, heif_image_input_class)':
/home/buildroot/autobuild/instance-3/output-1/build/libheif-1.17.5/libheif/plugins/encoder_jpeg.cc:366:37: error: invalid conversion from 'long unsigned int*' to 'size_t*' {aka 'unsigned int*'} [-fpermissive]
366 | jpeg_mem_dest(&cinfo, &outbuffer, &outlength);
| ^~~~~~~~~~
| |
| long unsigned int*
Fixes:
- http://autobuild.buildroot.org/results/8ca909564c8dabe28ad08c96ebbc04b25592e727
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit bd5d1f5b6a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit 2c5a82a29c (package/openssh: select linux-pam if refpolicy
upstream is selected) did not account for the linux-pam dependencies
before selecting it, causing unmet dependencies warnings (unfortunately,
not errors), such as:
$ KCONFIG_SEED=0xCF227CF4 make randconfig
WARNING: unmet direct dependencies detected for BR2_PACKAGE_LINUX_PAM
Depends on [n]: BR2_ENABLE_LOCALE [=n] && BR2_USE_WCHAR [=n] && !BR2_STATIC_LIBS [=n] && BR2_USE_MMU [=y] && BR2_TOOLCHAIN_GCC_AT_LEAST_4_9 [=y]
Selected by [y]:
- BR2_PACKAGE_OPENSSH [=y] && BR2_USE_MMU [=y] && BR2_PACKAGE_REFPOLICY_UPSTREAM_VERSION [=y]
2c5a82a29c made the choice of having openssl bear the responsibility
to select linux-pam when the upstream refpolicy version was enabled.
Semantically however, the responsibility really lies within refpolicy
itself, since that's what imposes linux-pam to openssh.
Move the select to refpolicy and drop it from openssh. Then, ensure that
linux-pam is only selected when it is available.
That means that one may get an openssh that is not linked against
linux-pam, when the linux-pam dependencies are not met; refpolicy (by
way of libsepol) also has a more stringent requirement on gcc version
than linux-pam, so most probably the missing dependencies would be
locale, wchar, or a static build. We consider that situation to be a
corner case that we do not want to address.
In the future, we may have more similar situations, whereby refpolicy
would impose other packages be linked with otherwise optional
dependencies. If (when) that were (will be) the case, then the proposed
mechanism would quickly become ugly; we could then re-assess a nicer way
to do that. Until then, this is good ebough.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Adam Duskett <adam.duskett@amarulasolutions.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit c20b2ae4ec)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Currently, when the version string is "too long", it is arbitrarily
truncated.
This works well for commit hashes, because usually the truncation is
long enough to provide a short hash that is still unique in the
upstream VCS.
However, there are non-hash-like versions strings that get truncated
and wihch the discriminant part is toward the end.
Yet, adapting the version cell to the widest versions string (most
probably a git hash) is not very interesting; the table is already very
large.
Make the cell with the version string scrollable: we get to keep the
best of both worlds: a narrow version cell, and a full-length version
string that can be copy-pasted if needed.
Signed-off-by: Sen Hastings <sen@hastings.org>
[yann.morin.1998@free.fr: reword commit log]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 4533e42622)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
redis-plus-plus unconditonally uses mutex and so needs threads since its
addition in commit c46df9de21 and
1fa8f6bbfb:
In file included from /home/autobuild/autobuild/instance-5/output-1/build/redis-plus-plus-1.3.10/src/sw/redis++/connection_pool.h:27,
from /home/autobuild/autobuild/instance-5/output-1/build/redis-plus-plus-1.3.10/src/sw/redis++/connection_pool.cpp:17:
/home/autobuild/autobuild/instance-5/output-1/build/redis-plus-plus-1.3.10/src/sw/redis++/sentinel.h:95:10: error: 'mutex' in namespace 'std' does not name a type
95 | std::mutex _mutex;
| ^~~~~
Fixes: c46df9de21
- http://autobuild.buildroot.org/results/cc4c54134e8a522b4c028aab4266f47cf862e2a9
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 044752f563)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Shortly before GCC 10.4.0 was dropped in d37a8f3a2, commit 4ce0dacb6 was
merged and introduced a patch to 10.4.0 that was not accounted for by
the patch that dropped 10.4.0 support.
Fixes: d37a8f3a2 ("package/gcc: remove gcc 10.x")
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit fbedcfa221)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit b9e89b340e (sudo: bump version) introduced the explicit build
of mksigname and mksiglist as host tools, as they were required to run
on the build machine, to generate C code then used to build the target
program.
This is now failing to build since the bump to sudo 1.9.15p5 in commit
c87746afef (package/sudo: security bump to version 1.9.15p5)
(lines manually wrapped and slightly elided for ease of reviewing):
/usr/bin/cpp \
-I/home/ymorin/dev/buildroot/O/master/per-package/sudo/host/include \
-I../../include \
-I../.. \
./sys_signame.h \
| /usr/bin/sed -e '1,/^int sudo_end_of_headers;/d' -e '/^#/d' > mksigname.h
In file included from /usr/include/features.h:394,
from /usr/include/sys/types.h:25,
from ./sys_signame.h:4:
/usr/include/features-time64.h:26:5: error: #error "_TIME_BITS=64 is allowed only with _FILE_OFFSET_BITS=64"
26 | # error "_TIME_BITS=64 is allowed only with _FILE_OFFSET_BITS=64"
| ^~~~~
/usr/bin/gcc -I../../include -I../.. -I. -I. \
-D_PATH_SUDO_CONF=\"/etc/sudo.conf\"
-I/home/ymorin/dev/buildroot/O/master/per-package/sudo/host/include \
-DDEFAULT_TEXT_DOMAIN=\"sudo\" \
-O2 \
-I/home/ymorin/dev/buildroot/O/master/per-package/sudo/host/include \
./mksigname.c -o mksigname
In file included from /usr/include/features.h:394,
from /usr/include/bits/libc-header-start.h:33,
from /usr/include/stdlib.h:26,
from ./mksigname.c:27:
/usr/include/features-time64.h:26:5: error: #error "_TIME_BITS=64 is allowed only with _FILE_OFFSET_BITS=64"
26 | # error "_TIME_BITS=64 is allowed only with _FILE_OFFSET_BITS=64"
| ^~~~~
make[2]: *** [Makefile:263: mksigname] Error 1
The core of the issue has not been really identified, but it turns out
that neither mksiglist nor mksignames is used during the build. This has
been tested with a minimal sudo with no option, and with a sudo with all
options enabled (linux-pam, zlib, opensldap, and openssl), with the
three types of C libraries (glibc, musl, and uClibc-ng).
Digging in the sudo buildsystem did not reveal an obvious reason when
those would be needed either.
Drop the hook now it seems it is no longer used and is atually breaking
the build.
Fixes: http://autobuild.buildroot.org/results/72f/72ff18fb9b41394a29006f881ee1fbea67a66a09/
Note that there is a second issue in there: the call to the host cpp
fails, but since it is on the LHS of a pipe, the error is lost, as the
RHS of the pipe (the sed call) succeeds; a fix for that will be sent
in a separate patch.
Reported-by: Christian Stewart <christian@aperture.us>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Peter Korsgaard <peter@korsgaard.com>
Reviewed-by: Christian Stewart <christian@aperture.us>
(cherry picked from commit 5aade5ced5)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issues:
CVE-2023-40546 mok: fix LogError() invocation
CVE-2023-40547 - avoid incorrectly trusting HTTP headers
CVE-2023-40548 Fix integer overflow on SBAT section size on 32-bit system
CVE-2023-40549 Authenticode: verify that the signature header is in bounds.
CVE-2023-40550 pe: Fix an out-of-bound read in verify_buffer_sbat()
CVE-2023-40551: pe-relocate: Fix bounds check for MZ binaries
https://github.com/rhboot/shim/tree/15.8
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit f42004a046)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Update patch to compile correctly with newer versions of GCC, which
has gotten stricter about the placement of the alignas() attribute.
Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
[yann.morin.1998@free.fr: update .checkpackageignore]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 9a7a7f3d13)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issues:
https://wpewebkit.org/security/WSA-2024-0001.html
- CVE-2024-23222: Processing maliciously crafted web content may lead to
arbitrary code execution. Apple is aware of a report that this issue
may have been exploited. Description: A type confusion issue was
addressed with improved checks.
- CVE-2024-23206: A maliciously crafted webpage may be able to
fingerprint the user. Description: An access issue was addressed with
improved access restrictions.
- CVE-2024-23213: Processing web content may lead to arbitrary code
execution. Description: The issue was addressed with improved memory
handling.
Add an upstream post-2.42.5 patch to fix an issue with an invalid
backport causing a build issue.
Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 5275e14125)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
