slide-switch is my software, I choose to continue to be sole maintainer.
(This was also the case in #28429.)
Signed-off-by: Jeffery To <jeffery.to@gmail.com>
Dropped patch: 0001-Make-GEMM3M-parameters-available-on-32bit-X86-GENERI.patch
Part of upstream: https://github.com/OpenMathLib/OpenBLAS/pull/5418
Changelog:
- Revert a 0.3.30 optimization that could cause race conditions
and invalid results in GEMM
- Fix thread lockup with Python 3.9 and NumPy
- Fix deadlock in multithreaded code after fork()
- Add bfloat16 extensions (BGEMM, BGEMV) and basic FP16 infrastructure
- Add batch GEMM operations with strided variants
- Add multithreaded LAPACK SLAED3/DLAED3 for improved eigensolvers
- Add Apple M4 and Intel Lunar Lake support
- Add initial POWER11 architecture support
- Improve GEMM performance on A64FX and ARM processors
Full release notes:
https://github.com/OpenMathLib/OpenBLAS/releases/tag/v0.3.31
Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
With fortify sources libutp fails to compile because the fortify sources
for musl use the GNU extension include_next. Do not fail when the
compiler issues a warning.
Fixes the following compile error:
```
In file included from libutp-2023.02.14~c95738b1/utp_utils.cpp:23:
/include/fortify/stdlib.h:22:2: error: #include_next is a GCC extension [-Werror]
22 | #include_next <stdlib.h>
| ^~~~~~~~~~~~
```
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Update modbus to version 3.1.12
https://github.com/stephane/libmodbus/releases/tag/v3.1.12
Fix FD_SET overflow when socket fd >= FD_SETSIZE.
Check dest pointer not null and nb in read functions.
NULL check for src and nb < 1 validation in write functions.
modbus_reply: don't compute address for FC 0x07/0x11.
Use O_NONBLOCK instead of deprecated O_NDELAY.
Explicit cast for Coverity CID 416366.
Document required buffer size of modbus_receive.
Document macros for error codes corresponding to Modbus exceptions.
Fix example of modbus_rtu_set_serial_mode.
Test filesystem provides symlink in autogen.sh.
Sync API signatures with the documentation.
Many documentation fixes and typo corrections.
Add coverage target and helper script.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Remove no longer used CONFIGURE_ARGS since the conversion to CMake.
Remove patented stuff. Red Hat Legal seems to have approved its
inclusion into Fedora without any flags disabling SBR. If it's good
enough for Red Hat it's good enough for OpenWrt.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
libtatsu is a dependency only for libimobiledevice-utils
the library itself does not use it during build, and is a
core component for iPhone tethering, while the utils are optional
move the dependency to the utils, to reduce the build size:
libtatsu depends on libcurl, which is compiled with a TLS library,
so users of prebuilt packages are forced to install both
OpenSSL and mbed TLS. This patch removes the unnecessary dependency.
Fixes: https://github.com/openwrt/packages/issues/28427
Signed-off-by: Georgi Valkov <gvalkov@gmail.com>
This software seems no longer maintained by upstream.
The latest upstream release is 16 years ago,
and no package depends on this.
Signed-off-by: Yanase Yuki <dev@zpc.st>
This release is needed in order to build against the 6.18 kernel.
Removed upstreamed: 020-gcc15.patch
Signed-off-by: John Audia <therealgraysky@proton.me>
We are pleased to announce the release of Unicode® ICU 78.2. It updates to CLDR 48.1. These are maintenance releases for ICU 78 and CLDR 48, with limited sets of bug fixes and no API or structural changes.
ICU 78.2 also includes a small number of bug fixes, as well as a minor update for time zone data (tzdata) version 2025c (2025-dec) ICU-23296.
Signed-off-by: Hirokazu MORIKAWA <morikw2@gmail.com>
Bump to 9.2.0 and swtich from git build to tagged tarball.
Added the following patches:
900-Fix-Compilation-on-kernel-6.17.patch (backport)
901-pf_ring-fix-virt_to_page-type-mismatch.patch (pending backport)[1,2]
Removed all upstreamed patches and removed MAINTAINER as no one is
maintaining this pkg. Note that this version is needed in order to
build against linux 6.18.
1. Suggested by CI co-pilot to prevent failure builds for ARM targets
including: arm_cortex9_vfpv3-d16, arm_cortex-a15_neon-vfpv4,
powerpc_464fp, and powerpc_8548.
2. https://github.com/ntop/PF_RING/pull/1010
Signed-off-by: John Audia <therealgraysky@proton.me>
Obsolete use of $(SDK) in configure conditionals can result in
dependency errors when building a subset of packages for packages which
have multiple sub-packages.
The reason it causes dependency issues is that (using libdbi-drivers as
an example) lines like:
ifneq ($(SDK)$(CONFIG_PACKAGE_libdbd-sqlite3),)
always evaluate to true if you are compiling in the SDK. So for a user
compiling from the SDK, the configure arguments are always added to the package build.
In the case of libdbi-drivers:
CONFIGURE_ARGS += \
--with-sqlite3 \
--with-sqlite3-incdir=$(STAGING_DIR)/usr/include \
-with-sqlite3-libdir=$(STAGING_DIR)/usr/lib
is always added even if PACKAGE_libdbd-sqlite3 is deselected. When
libdbd-sqlite3 is deselected, this dependency:
DEPENDS:=libdbi +libsqlite3
is not present, so when configure tries to find sqlite3 it fails.
Closes#28173 "tree-wide: obsolete $(SDK) in conditionals"
See also:
* "include: remove SDK exception from package install targets"
openwrt/openwrt@28f44a4
Performed tree-wide to ease revert if necessary, per:
https://github.com/openwrt/packages/issues/28173#issuecomment-3694615980
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
$(INSTALL_BIN) follows symlinks, causing the .so to be copied multiple times.
Use $(CP) instead to preserve symlinks and cut package size by ~2/3.
OpenWrt libraries don’t need to be executable, so $(INSTALL_BIN) isn’t required.
Signed-off-by: Jordan Ngako <jordanfalken@gmx.de>
fixes float endianness issues in 3.1.8.
I use a small, custom application to convert data from my
electricity meter into Modbus format for my inverter and wallbox.
I developed this program against an earlier version of libmodbus
on x86, and it ran flawlessly on my ath79 router (big endian).
After migrating to a Mediatek router (little endian), the program
only outputted huge, meaningless values. However, it ran perfectly
fine on my x86 system (little endian), which is running Fedora
with libmodbus v3.1.11.
I then found several bug reports and the changelogs for libmodbus 3.1.11
that described and resolved my problem. So update openwrt to 3.1.11.
Signed-off-by: Stefan Hellermann <stefan@the2masters.de>
Without it the build fails for me with:
/usr/lib/libltdl.so: file not recognized: file format not recognized
Signed-off-by: Aleksey Vasilenko <aleksey.vasilenko@gmail.com>
This is a security release.
Security:
* CVE-2025-31498. A use-after-free bug has been uncovered in read_answers() that was introduced in v1.32.3. Please see GHSA-6hxc-62jh-p29v
* CVE-2025-62408. A use-after-free bug has been uncovered in read_answers() that was introduced in v1.32.3. Please see GHSA-jq53-42q6-pqr5
Signed-off-by: Hirokazu MORIKAWA <morikw2@gmail.com>
Fixes pgsql-server: the setup fails for any folder
Fixes#27228
Sets postgresql-specific configure flags that configure cannot run-test
to determine their value. This fixes improperly linked files that
prevent database initialization (at least) from working on the device.
Signed-off-by: Daniel F. Dickinson <dfdpublic@wildtechgarden.ca>
