tcl:tcl is a better CPE ID than tcl_tk:tcl_tk as this CPE ID has the
latest CVE (whereas tcl_tk:tcl_tk only has CVEs up to 2008):
https://nvd.nist.gov/products/cpe/search/results?keyword=cpe:2.3🅰️tcl:tcl
Fixes: 299e5b0a9b (treewide: add PKG_CPE_ID for better cvescanner coverage)
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
After reinstalling the packages with the preserved configuration files
after a sysupgrade, the reinstalled package config files overwrite what
is on disk rather than being placed as conf-opkg. Defining these config
files will preserve them appropriately.
Signed-off-by: Joel Low <joel@joelsplace.sg>
This reverts commit 5313dd9be from PR #25324 , related to #25323 and
https://github.com/openwrt/openwrt/pull/16899
The additional symlink libpng16.so -> libpng16d.so is now unnecessary as
OpenWrt main repo PR https://github.com/openwrt/openwrt/pull/18709
commit 703e7d2d5b
changed the cmake build type from 'Debug' to 'RelWithDebInfo', which
causes libpng .so to be compiled with the normal name libpng16.so
(instead of the debug-styled libpng16d.so).
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Update to latest release and change URL to official upstream mirror.
Removed upstreamed patch: 130-musl-svcgssd-sysconf.patch
Added new patch to correct host build error as we do not build with
gss enabled anyway: 100-fix-host-build.patch
Build system: x86/64
Build-tested: bcm27xx/bcm2712
Run-tested: bcm27xx/bcm2712
Signed-off-by: John Audia <therealgraysky@proton.me>
- chore(lib): expose model methods to obtain progress
- feat(gui): explanation to options enabled or disabled per folder type
- fix(gui): validate device ID in canonical form
- fix(config): remove discontinued primary STUN server
- fix(stun): better error handling
- chore(config): remove discontinued secondary STUN servers
- chore(fs): speed up case normalization
- build(deps): update dependencies
- feat(fs, config): add support for custom filesystem type construction
- build: replace underscore in Debian version
- chore(model): add metric for total number of conflicts
- fix(config): properly apply defaults when reading folder configuration
- fix(config): zero filesystemtype is "basic"
- build: push artifacts to Azure
- chore(config): resolve primary STUN servers via SRV record
- chore(fs): changes to allow Filesystem to be implemented externally
- fix(strings): differentiate setup(n) and set(v) up
- fix(gui): mark unseen disconnected devices as inactive
- fix(syncthing): use separate lock file instead of locking the
certificate
- feat(api, gui): allow authentication bypass for metrics
- chore: add missing copyright in new files from infra branch
- fix(osutil): give threads same I/O priority on Linux
- chore(syncthing): remove support for TLS 1.2 sync connections
- chore(gui): update dependency copyrights, add script for periodic
maintenance
- chore(api): log X-Forwarded-For
- feat(config): add option for audit file
- chore(gui): use go list --deps for dependency list
- fix(strelaysrv): make the session limiter session-dependent
Changelog: https://github.com/syncthing/syncthing/compare/v1.29.5...v1.29.6
Signed-off-by: George Sapkin <george@sapk.in>
We have received reports of builds of perl occasionally failing when
building with many parallel jobs, with a log like the following:
LD_LIBRARY_PATH=[...]/perl/perl-5.40.0 ./miniperl -Ilib make_ext.pl \
dist/constant/pm_to_blib MAKE="make" LIBPERL_A=libperl.so
File/Path.pm did not return a true value at [...]/hostpkg/usr/lib/perl5/5.40.0/ExtUtils/MakeMaker.pm line 13.
BEGIN failed--compilation aborted at [...]/hostpkg/usr/lib/perl5/5.40.0/ExtUtils/MakeMaker.pm line 13.
Compilation failed in require at Makefile.PL line 3.
BEGIN failed--compilation aborted at Makefile.PL line 3.
Unsuccessful Makefile.PL(dist/constant): code=65280 at make_ext.pl line 532.
The failing extension (dist/constant in the above log) would differ
between runs.
The cause of the issue is the `-Ilib` in the command line of miniperl.
In the host build, `./miniperl -I lib` will use the following include
path:
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/AutoLoader/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/Carp/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/PathTools
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/PathTools/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/ExtUtils-Install/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/ExtUtils-MakeMaker/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/ExtUtils-Manifest/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/File-Path/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/ext/re
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/Term-ReadLine/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/Exporter/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/ext/File-Find/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/Text-Tabs/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/dist/constant/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/version/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/Getopt-Long/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/Text-ParseWords/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/cpan/ExtUtils-PL2Bat/lib
[..]/build_dir/hostpkg/perl/perl-5.40.0/lib
.
Various dependencies of the extension build scripts (Makefile.PL) -
including File-Path, which failed to be loaded in the error log - are
included in the path by buildcustomize.pl, as these extensions are only
installed to `lib` as the build proceeds.
However, in a target build, miniperl is just a symlink to the previously
built host perl. As the host perl does not implicitly load
`buildcustomize.pl`, we get the following include path for
`./miniperl -Ilib`:
lib
[..]/staging_dir/hostpkg/usr/lib/perl5/site_perl/5.40.0/x86_64-linux
[..]/staging_dir/hostpkg/usr/lib/perl5/site_perl/5.40.0
[..]/staging_dir/hostpkg/usr/lib/perl5/5.40.0/x86_64-linux
[..]/staging_dir/hostpkg/usr/lib/perl5/5.40.0
The host perl's install location is used as the default include path
which provides File-Path etc. for the target build; however, as more
and more libraries get installed into `lib` during the extension build,
they may get loaded from there instead, as `lib` is at the beginning of
the include path. When multiple extensions are built in parallel, a
Makefile.PL may attempt to load File/Path from `lib` after the file has
been created, but before its contents have been written fully, resulting
in the build to fail.
In fact, we should not load anything from `lib` during the target build,
as it is the staging directory for the target, including native
extensions built for the target architecture - with one exception: The
build scripts expect to find target information in the `Config` module,
so simply removing `lib` from the include path completely would break
the build.
Solve the issue by creating an alternative lib directory `lib_build`,
symlinking `Config.pm` and its dependencies in it, and replacing the
`-Ilib` argument with `-Ilib_build` using a wrapper script around the
host perl executable. This is similar to the approach seen in perl's own
obsolete/broken cross compile scripts (`Cross/Makefile`).
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
The patch was introduced in commit 4c57844f0f ("lang/perl: Add hack to
make perl always use miniperl during build"), but it is not actually
necessary. By setting $perl to a non-empty value (using 'perl' as is
common on desktop distros), the logic works as intended and selects the
correct perl binary for host and target builds.
As miniperl just symlinks to host perl for target builds, the main
effect of this change is not unconditionally passing `-Ilib -I.`
anymore. This seems like a good thing; host libraries should be used
with host perl by default.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
The patch was first introduced in commit 4a94479f96 ("perl: update to
5.26.1") to fix the target build when the host perl has
default_inc_excludes_dot enabled. It just added back the `-I`. to every
call of miniperl; this solution is questionable however, as it adds `.` to
the beginning of the search path, not as a final fallback like perl did
before default_inc_excludes_dot (and like miniperl does).
It is also not necessary - only two scripts, write_buildcustomize.pl and
configpm, expect to be able to include a file from `.` (in both cases a
file the script just generated). Just fix the two scripts instead.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
With no SONAME set, when linking against the full library path, that
path will be used. But if SONAME is set, it will be used instead.
Set --soname=legacy to add a SONAME to the library to allow projects
that use full path to link correctly.
Link: https://sqlite.org/src/forumpost/5a3b44f510df8ded
Fixes: https://github.com/openwrt/packages/issues/26449
Signed-off-by: George Sapkin <george@sapk.in>
Description: fix in full compile a firmware
pkg_hash_check_unresolved: cannot find dependency aria2-openssl for aria2
pkg_hash_fetch_best_installation_candidate: Packages for aria2 found, but incompatible with the architectures configured
satisfy_dependencies_for: Cannot satisfy the following dependencies for luci-app-aria2:
aria2-openssl
opkg_install_cmd: Cannot install package luci-app-aria2.
Signed-off-by: Lunatic Kochiya <125438787@qq.com>
Changed source URL to github (faster/geo-redundancy).
build: x86_64
run tested: x86_64
```
# radiusd -v
radiusd: FreeRADIUS Version 3.2.7, for host x86_64-openwrt-linux-gnu, built on Apr 18 2025 at 00:10:48
FreeRADIUS Version 3.2.7
```
Signed-off-by: Paul Donald <newtwen+github@gmail.com>
Trying to compile with GCC14 will fail on compiler sanity check with:
configure:1056:1: error: return type defaults to 'int' [-Wimplicit-int]
1056 | main(){return(0);}
| ^~~~
This is due to GCC14 not allowing implicit integer types anymore[1].
So, patch configure to avoid this and make it compile with GCC14.
Proper fix would be to use autoreconf to rebuild configure but configure.in
is completely outdated and would likely be more broken when regenerated.
[1] https://gcc.gnu.org/gcc-14/porting_to.html#implicit-int
Signed-off-by: Robert Marko <robimarko@gmail.com>
Tianling Shen