GNUmakefile hardcodes its own CFLAGS= assignment (patched by
0001-Build-so-lib.patch to add -fPIC for the shared-lib build). As a
plain makefile variable, that assignment overrides the CFLAGS bitbake
exports into the build environment -- including the -ffile-prefix-map
flags from DEBUG_PREFIX_MAP. Without those, -g debug info in the
compiled snowball/stemwords/libstemmer.so binaries embeds the absolute
build-tree path, tripping the buildpaths QA check on the -dbg package:
ERROR: libstemmer-3.1.1-r0 do_package_qa: QA Issue: File
/usr/bin/.debug/snowball in package libstemmer-dbg contains
reference to TMPDIR [buildpaths]
Pass CFLAGS via EXTRA_OEMAKE instead, which oe_runmake places on the
make command line and therefore does take precedence over the
makefile's own CFLAGS= assignment.
Verified via cleansstate rebuild: log.do_package_qa no longer contains
any buildpaths warnings, and `strings` on the rebuilt snowball debug
binary no longer contains any build/tmp/work host paths.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
glib-mkenums records the "enumerations from" comment using the
input file's full path (@filename@), embedding the build host's
TMPDIR/HOME path into the installed gladeui-enum-types.c/.h and
tripping the buildpaths QA check:
WARNING: glade-3.40.0-r0 do_package_qa: QA Issue: File
/usr/src/debug/glade/3.40.0/gladeui/gladeui-enum-types.c in
package glade-src contains reference to TMPDIR [buildpaths]
Switch both mkenums templates to @basename@ (supported since glib
2.22), the same fix already used for the analogous gcab-enums
issue. Verified with a clean rebuild: do_package_qa now reports no
buildpaths warnings for glade.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
restinio 0.7.x is a major architectural change from 0.6.x: the release
tarball no longer bundles any of its third-party dependencies
(http_parser, expected-lite, fmt, catch2, sobjectizer). They must now
be supplied externally and resolved through find_package()/system
linkage. This requires two new supporting recipes added in prior
commits: llhttp (replacement for the now-dropped http_parser) and
expected-lite (nonstd::expected backport, now required unconditionally
instead of restinio's old embedded fallback). asio, fmt and catch2
dependencies are already satisfied by existing recipes in the tree.
Recipe changes:
- Release tag format changed from "v.X.Y.Z" to "vX.Y.Z"; update
SRC_URI and SRC_URI[sha256sum] accordingly.
- LIC_FILES_CHKSUM md5 updated for the new tarball's root LICENSE file.
- DEPENDS: drop http-parser, add llhttp and expected-lite.
- EXTRA_OECMAKE: RESTINIO_FIND_DEPS and RESTINIO_USE_EXTERNAL_HTTP_PARSER
were removed upstream. RESTINIO_BENCH renamed to RESTINIO_BENCHMARK,
RESTINIO_ALLOW_SOBJECTIZER renamed to RESTINIO_WITH_SOBJECTIZER (kept
disabled). Added RESTINIO_DEP_STANDALONE_ASIO=find,
RESTINIO_DEP_FMT=find and RESTINIO_DEP_EXPECTED_LITE=find so restinio's
CMakeLists.txt resolves these via find_package() instead of trying to
add_subdirectory() vendored sources that no longer exist in the
tarball. RESTINIO_DEP_LLHTTP=system is used instead of "find" because
llhttp's own CMakeLists.txt only exports an "llhttp::llhttp_shared"
imported target (llhttp::llhttp is just a build-tree ALIAS, never
installed/exported), which restinio's dependency-resolution logic
does not recognize as a valid find_package() target name.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
restinio 0.7.x drops its bundled nonstd::expected backport and
instead requires expected-lite to be supplied externally via
find_package(). Add a new recipe pinned to v0.8.0, the exact
version restinio 0.7.9.1 pins in its externals.rb manifest.
Modeled on the existing span-lite recipe (same author/ecosystem,
same LICENSE.txt boilerplate).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
restinio 0.7.x drops its bundled http_parser and instead requires
llhttp to be supplied externally via find_package().
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Stay within the 4.23.x stable series (do not touch the sibling
talloc/tdb/tevent/ldb recipes). This picks up two intervening point
releases:
- 4.23.9: bug-fix release (use-after-free in ACL claims/conditions
handling, CTDB read-only record use-after-free/resource leak, RODC
auth fixes, among others).
- 4.23.10: security release fixing CVE-2026-6949 (TSIG/DNS OOB write
crash), CVE-2026-58216 (KDC kpasswd OOB read crash),
CVE-2026-58218 (DNS TKEY cache-flood DoS), CVE-2026-58221 (LDAP
privilege escalation via internal LDB special DNs),
CVE-2026-58222 (LDAP Compare filter used as protected-attribute
disclosure oracle) and CVE-2026-58224 (CTDB protocol
bounds-checking gaps).
- 4.23.11: further bug fixes (DRS memory leak, pthreadpool fork race,
RODC/NTLMv2 trust fixes, CephFS vfs crashes).
All 13 existing patches (including the musl-only pam/getpwent_r ones)
still apply cleanly against 4.23.11 with offset only, no fuzz, so none
needed to be reworked or dropped. LIC_FILES_CHKSUM for COPYING is
unchanged.
Build-verified with cleansstate + full build for qemux86-64.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bump to the latest snort3 release. Drop
0001-Fix-build-on-c23.patch: the once_flag/call_once C23 name-collision
fix it carried (submitted upstream as snort3 PR #431) is now merged
verbatim upstream (src/loggers/alert_fast.cc already uses
std::once_flag/std::call_once as of this tag), and the patch fails to
apply (reversed/already-applied) against 3.12.2.0.
Keep 0001-cmake-Check-for-HP-libunwind.patch (applies cleanly) and
0001-cmake-Pass-noline-flag-to-flex.patch (applies with offset only,
no fuzz) unchanged.
The ~180 commits between the old and new SRCREV include several
crash/security fixes (use-after-free, heap-buffer-overflow,
null-deref) that are picked up naturally by this version bump.
Build-verified with cleansstate + full build for qemux86-64.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bump to the 4.6.0 release tarball. As of 4.6.0 upstream introduces
CMake as the recommended build system, but the release tarball still
ships the Autotools build files (configure/Makefile.am) and Autotools
remains fully supported for building from release tarballs (only
deprecated for eventual removal in a future release), so the existing
autotools-brokensep-based recipe continues to work unchanged.
0001-libopts.m4-set-POSIX_SHELL-to-bin-sh.patch kept unchanged, still
needed, applies cleanly.
Build-verified: tcpreplay builds and packages successfully (existing
buildpaths QA warnings for libtcpreplay.pc/defines.h are pre-existing
and expected, WARN not ERROR on this distro).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bump to the upstream 2.66 orig tarball (now .orig.tar.gz instead of
.orig.tar.bz2) from Debian.
- ssmtp-bug584162-fix.patch dropped: the fix (dropping trailing '\r'
in standardise() and the header_parse() EOF/empty-header handling)
is already present verbatim in the 2.66 source.
- 0001-ssmtp-Correct-the-null-pointer-assignment-to-char-po.patch
dropped: upstream already initializes all the char pointers with
NULL instead of (char)NULL.
- 0001-Fix-incompatible-pointer-types-error-with-gcc-15.patch
rebased: the call site already casts to (sighandler_t)handler in
2.66, so only the handler() signature hunk is still needed; the
call-site hunk was dropped as it no longer applies and is
unnecessary once the signature takes an int.
- use-DESTDIR.patch rebased: upstream 2.66 already appends $(LDFLAGS)
in the ssmtp link rule, so that hunk was dropped; the DESTDIR
hunks for install/uninstall are still needed and were regenerated
against the new Makefile.in.
- build-ouside_srcdir.patch and 0001-include-libgen.h-for-basename.patch
kept unchanged, still needed, apply cleanly.
Build-verified: ssmtp builds and packages successfully.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bump to the v2.7.0 tag. Rebase the ptest ICMPv6 skip patch: upstream
added a "needs_root" tag to the "Sending and receiving an
ICMPv6EchoRequest" test's tag line in test/regression.uts, so the
"skipped_ptest" tag now needs to be appended after "needs_root"
rather than replacing the whole line. With this rebase both hunks
apply with offset only (no fuzz).
Build-verified: python3-scapy builds and packages (including
-ptest) successfully.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Drop 0001-Shim-for-renaming-of-GNUTLS_NO_EXTENSIONS-in-GnuTLS-.patch:
it was a workaround for GnuTLS >= 3.8.1 renaming GNUTLS_NO_EXTENSIONS
before a compatibility shim landed upstream; that compatibility shim
is long merged and this openconnect release range no longer needs
the local re-definition (build succeeds without it).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bump to the nm-1-58 stable branch. CVE-2026-10805 and the DHCPv4
out-of-bounds read fix are both already included in the 1.58.0 tag
(verified via git merge-base --is-ancestor against upstream), so no
backport patch is required.
- 0002-meson-fix-cross-compilation-issues.patch dropped: merged
upstream verbatim as commit e5d2c7c by the same author, and is an
ancestor of the 1.58.0 tag.
- 0001-linker-scripts-Do-not-export-_IO_stdin_used.patch kept, still
needed, applies with offset only.
- dhclient support was removed upstream (all: drop dhclient code and
build configuration); drop PACKAGECONFIG[dhclient] and update the
NETWORKMANAGER_DHCP_DEFAULT comment accordingly.
- Wireless Extensions (wext) meson option is deprecated and now
errors on -Dwext=true; drop it from PACKAGECONFIG[wifi], keeping
only -Dwifi.
- New CLAT feature defaults to enabled in meson_options.txt and pulls
in libbpf >= 1.3.0; add PACKAGECONFIG[clat] but leave it out of the
default PACKAGECONFIG set so it stays disabled and avoids a new
hard dependency.
- Package the new nm-initrd-generator.sh systemd generator directory
under ${PN}-daemon and make bash an unconditional RDEPENDS since
the generator script requires it regardless of the ifupdown
PACKAGECONFIG setting.
Build-verified: networkmanager and its openvpn/openconnect/
fortisslvpn plugin recipes all build successfully.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Update the Debian-snapshot SRC_URI to snapshot.debian.org's 1.238-1
timestamp (20260519T203124Z). LIC_FILES_CHKSUM is refreshed since
debian/copyright's md5 changed (license text/terms unchanged,
BSD-3-Clause).
0001-fix-netcat-openbsd-ftbfs-with-GCC-15.patch is dropped: Debian's
own quilt-patches/quit-timer.patch (applied before our OE patches via
netcat_do_patch) already changes quit() to take an int sig argument in
1.238-1, making our GCC-15 fix fully redundant.
0001-bundle-own-base64-encode-decode-functions.patch is kept (still
needed, no libc-provided b64_ntop/b64_pton) but rebased: the Makefile
hunk's context line changed from "LIBS=" to "LIBS ?=" and the netcat.c
hunk's context line changed from "static void quit();" to
"static void quit(int sig);" following the quit-timer.patch rebase
upstream; both are updated to apply without fuzz.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
0001-configure.ac-Do-not-emit-compiler-flags-into-object-.patch still
applies cleanly and is still needed: upstream 2.4.3's configure.ac
still adds -grecord-gcc-switches unconditionally under
enable_hardening, which embeds absolute sysroot paths into object
files and breaks reproducibility.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
0001-cmake-Allow-SYSTEMCTL_BIN-to-be-overridden-from-envi.patch still
applies (with a minor line-offset shift) and is still needed: upstream
v64.0's CMakeLists.txt still unconditionally calls find_program() for
SYSTEMCTL_BIN without honoring a pre-set value.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
fix-openssl-no-des.patch still applies cleanly (with line-offset shifts)
and is still needed: upstream 5.80's src/common.h and src/protocol.c do
not yet guard the DES includes/NTLM code with OPENSSL_NO_DES, so the
patch is retained unchanged.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Fixes CVE-2026-63649 (openvpnserv on Windows insufficiently
scrutinizes the command line passed in from the control socket,
allowing admin config-directory restrictions to be bypassed) and
CVE-2026-63650 (--x509-username-field could unintentionally permit
a certificate that should not have been accepted when built against
mbedTLS).
The upstream swupdate.openvpn.org/net release mirror has not yet
synced the 2.7.6 tarball (returns 404), so fetch the official signed
GitHub release asset instead, which is available immediately after
tagging.
The existing 0001-configure.ac-eliminate-build-path-from-openvpn-versi.patch
still applies cleanly (with line-offset only) against 2.7.6's
configure.ac; the CONFIGURE_DEFINES logic it patches is unchanged.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Same 1.5 release branch; bump SRCREV_opcua to the v1.5.6 tag's commit.
The mdnsd submodule dependency also moved forward to its current
master commit (upstream's .gitmodules for this repo now points at
troglobit/mdnsd, but Pro/mdnsd carries the same commit history so the
existing fetch URL still resolves it) - update SRCREV_mdnsd and its
LIC_FILES_CHKSUM md5 (BSD-3-Clause text unchanged in substance, just
reworded/re-wrapped with an added copyright holder name).
SRCREV_ua-nodeset and SRCREV_mqtt-c are unchanged.
Build-verified with a full cleansstate rebuild.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Fixes nbd-client failing to find the device index when given a device
name without the /dev/ prefix; also refactors nbd-client's argument
parsing to be testable and adds a test for it.
Upstream's 3.27.1 GitHub release carries no tarball assets (unlike
3.27.0), so switch SRC_URI to fetch via git protocol pinned at the
nbd-3.27.1 tag's commit, consistent with OE's "no unstable GitHub
archive tarball" QA policy.
No patches involved. Build-verified with a full cleansstate rebuild.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Upstream's vervest.org tarball archive no longer serves 2.0.x releases
(404) and htpdate's GitHub releases carry no tarball assets, so switch
SRC_URI to fetch via git protocol pinned at the v2.0.2 tag's commit,
consistent with OE's "no unstable GitHub archive tarball" QA policy.
License header block shifted by one line upstream (25-29 instead of
26-30); text and md5 are otherwise unchanged, updated LIC_FILES_CHKSUM
accordingly.
No patches involved. Build-verified with a full cleansstate rebuild.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
No patches involved. Build-verified with a full cleansstate rebuild.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
This release includes all fixes from 9.2.19 (silent data divergence
and corruption fixes for diskless primaries, false split-brain fix,
list corruption/AB-BA deadlock/use-after-free fixes during connection
teardown, RDMA transport bug fixes) plus new lb-tcp path add/remove
support and a TLS connection page-reference leak fix.
Drop check_existence_of_modules_before_installing.patch: it no longer
applies (confirmed it already fails to apply against the unmodified
9.3.2 source too, since SKIP_RECIPE normally prevents do_patch from
ever running for this recipe). The install-time "module not built"
problem it guarded against is now handled by upstream's own rewritten
drbd/Makefile install logic: the KDIR path now delegates to kbuild's
own modules_install target, and the non-KDIR fallback only globs
*.ko files that actually exist via find | cpio. Neither path can hit
the old "install: cannot stat" failure the patch was working around.
Verified do_fetch/do_unpack/do_patch succeed cleanly with the new
tarball and without the patch (SKIP_RECIPE temporarily lifted for
testing). do_compile still requires coccinelle/spatch or network
access to spaas.drbd.io, neither available in this sandbox, matching
the existing SKIP_RECIPE annotation - not a regression.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Release Notes: https://github.com/google/perfetto/releases/tag/v57.2
perfetto 57.2 added a new build-time dependency on RE2 (via
enable_perfetto_re2, unconditionally true for standalone builds) for trace
processor regex support, so a matching buildtools/re2 git fetch/SRCREV was
added to perfetto.bb, mirroring the existing buildtools/* pattern from
tools/install-build-deps. abseil-cpp is now fetched the same way for the
same reason.
Upstream also dropped the amalgamated sdk/perfetto.{h,cc} from the git
release branches (now shipped only as a GitHub release .zip asset), so
libperfetto.bb now fetches perfetto-cpp-sdk-src.zip and unpacks it into
sdk/, which the meson build requires.
Build-verified for qemux86-64: both perfetto and libperfetto build
through do_package_qa cleanly.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move to the tagged libusbgx-v0.3.0 release instead of tracking an
untagged master SHA. Both OE-local patches
(0001-fix-stack-buffer-overflow-in-usbg_f_foo_attr_val-pro.patch and
0001-libusbgx-Add-interface-name-for-NCM-Feature-Descript.patch) are
already merged upstream by this tag, so drop them.
Build-verified for qemux86-64 (do_package_qa green, no patch fuzz).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
wx-config gained --baselib from 0005-wx-config-fix-libdir-for-multilib.patch
in the wxwidgets recipe. The 3.2.11 upgrade dropped that patch, since
upstream reworked the install paths to derive from CMAKE_INSTALL_LIBDIR
and no longer needs it. wx-config therefore no longer accepts the option
and the build fails:
| *** Error: Unrecognised option: '--baselib=lib'
| Use wx-config --help for information on command line options.
Nothing else needs it: wxwidgets' do_compile:append already rewrites
libdir in wx-config to an absolute ${STAGING_LIBDIR}, so the lib
component never comes from the prefix here.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Upstream refactored install.cmake/config.cmake/webview's CMakeLists.txt
to derive install paths from a new library_dir variable (=
CMAKE_INSTALL_LIBDIR, from GNUInstallDirs) and a wxCONFIG_DIR helper,
both already multilib-aware on their own:
- Drop 0003-create-links-with-relative-path.patch: wx_install_symlink()
now does "ln -s --relative" natively.
- Drop 0005-wx-config-fix-libdir-for-multilib.patch: superseded by the
library_dir/wx_get_install_platform_dir refactor.
- Reduce 0002-fix-libdir-for-multilib.patch to its one remaining hunk
(wxOUTPUT_DIR in CMakeLists.txt, a build-dir path untouched by the
refactor).
- Rebase 0004-don-not-append-system-name-to-lib-name.patch (context
shift only).
- Drop the now-unneeded wxBUILD_INSTALL_{RUNTIME,LIBRARY,PLATFORM_SUBDIR}
EXTRA_OECMAKE overrides: forcing an absolute libdir into
wxBUILD_INSTALL_RUNTIME_DIR broke the new install.cmake's wx-config
symlink construction (double /usr prefix, installed-vs-shipped QA
failure); GNUInstallDirs-based defaults are multilib-correct without
them.
- do_install:append: guard the bindir rmdir (bindir no longer created
at all, not just left empty) and repoint the host-contamination sed
at the new versioned cmake/wxWidgets-3.2/ install path.
Build-verified for qemux86-64 (do_package_qa green).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
8.10.0's default "all"/"build" goal also builds the bundled modules
(redisbloom, redisearch, redisjson, redistimeseries), which pull in
their own vendored build systems and a python3 interpreter this
recipe has never needed. Restrict do_compile to the core server only,
the way scripts/build.sh does it internally (oe_runmake -C src all).
Rebase 0004-src-Do-not-reset-FINAL_LIBS.patch for the shifted context.
Build-verified for qemux86-64 (do_package_qa green).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Minor release. Dropped 0001-setup.py-Disable-autodection-of-modules.patch:
2.13's setup.py already passes both packages=[] and py_modules=[] itself, so
re-adding py_modules=[] produced a "keyword argument repeated" SyntaxError.
Rebased 0001-unattended-upgrade-Remove-distro_info-usage-to-check.patch onto
the new import block and the shifted run() body.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Major release. Three patches rebased onto 3.0.2's shifted line numbers:
0001-make-Respect-variables-from-environement.patch (upstream added an
alternate -L$(ZPATH) -lz definition, commented out, above our target lines;
same changes, offset only), 0001-png-fix-IDAT-windowsize-...patch and
10-pngsplit-format-strings.patch (upstream renamed the usage-string macros
png_fix_IDAT_windowsize_usage/PNGSPLIT_HDR/PNGSPLIT_USAGE to their current
names; same %s-format fix, offset only).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. Two patches needed rebasing:
0009-Using-native-binaries.patch's context shifted because upstream added
its own '<@(emulator)' gyp variable (defaults to empty in common.gypi) to a
couple of the same action arrays this patch touches; this recipe never sets
GYP_DEFINES emulator=..., so it stays empty and our v8-qemu-wrapper.sh
insertion is unaffected either way. 0007-positional-args.patch's single-line
shebang fix (bash -> sh) was already applied on top of an unrelated wording
change upstream made to the same 3-line fixture file, just offset by one
line (context-only).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Minor release. MZ_PPMD (new default-on feature since 4.2.x) git-clones
ip7z/7zip over the network at configure time via CMake FetchContent;
disabled it, there's no network access in the build sandbox and nothing in
this recipe needs it. do_install:append's post-install sed also needed
retargeting: 4.2.x renamed the generated cmake config file from
minizip.cmake to minizip-config.cmake.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. Dropped 0001-cmake-Do-not-use-mrelax-all-with-clang-on-RISCV64.patch:
its actual fix (the CMAKE_SYSTEM_PROCESSOR MATCHES "riscv" guard around
-mrelax-all) is already present upstream in 0.12.0's CMakeLists.txt; only the
CONTRIBUTORS name-credit hunk was left, not worth carrying as a patch on its own.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Major release. 10.0.1 installs a new terminfo entry
(${datadir}/terminfo/k/kmscon) that FILES:${PN} didn't claim; added.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Minor release. Two recipe-side fixes needed for the new meson.build:
3.40.0 renamed the "docs" option to "gtk_doc" and dropped "introspection"
entirely (introspection is no longer optional), so gtk-doc.bbclass and
gobject-introspection.bbclass are repointed via GTKDOC_MESON_OPTION and
GIR_MESON_OPTION rather than guessing at PACKAGECONFIG. 3.40.0 also ships
new /usr/share/help (translated user docs) and /usr/share/gettext/its
(catalog ITS rules) content that the old FILES:${PN} didn't claim, added.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Tracks the postgresql 18 major release this driver targets. Rebased
psqlodbc-fix-for-ptest-support.patch onto 18.00.0002: upstream now generates
a second (ANSI) DSN block alongside the original Unicode one, so the
localhost/postgres/5432 test defaults are applied to both; upstream also
added a server_version parameter to runtest() and native TAP "ok"/"not ok"
progress lines, neither of which this patch touches -- it continues to
reformat the final per-test line to PASS:/FAIL:, which is what
ptest-runner's log parser matches on.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Major release. Dropped 0001-Add-missing-include-in-Cluster.pm.patch, which is
no longer needed / no longer applies against 18.4's Cluster.pm. Rebased
0003-configure.ac-bypass-autoconf-2.69-version-check.patch and
0005-postgresql-fix-ptest-failure-of-sysviews.patch onto 18.4's shifted line
numbers (context-only changes; same single-line edits as before).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. SRC_URI[sha256sum] filled in for the new tarball. Refreshed
0001-Provide-strtod_l-signature-on-non-glibc-linux-musl-systems.patch: 2.8.144
widened the same guard lines with its own opt-in WITH_MUSL/WITH_POSIX_*_R
macros (which this recipe never defines), so the patch's automatic __GLIBC__
detection is still needed on top and was rebased onto the new line numbers
and macro list.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. LIC_FILES_CHKSUM md5 updated for LICENSE (copyright year bumped, license text unchanged).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. LIC_FILES_CHKSUM md5 updated for COPYING and COPYING.tools (upstream re-encoding, license text unchanged).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release from oe-core gstreamer1.0 1.28.x. SRC_URI[sha256sum] filled in for the new tarball.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. SRC_URI[archive.sha256sum] filled in for the new tarball.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. LIC_FILES_CHKSUM md5 updated for the LICENSE file whose copyright header now says 2017-2026 instead of 2017-2025; license text unchanged.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Six minor releases spanning ~15 months of development on top of the 0.11.0
snapshot the recipe was pinned to: 0.71/0.72 added man-page and import fixes,
0.73 broadened -p time-format support and added Hurd CLOCK_BOOTTIME
compatibility, 0.74 fixed the varlink interface name, and 0.75 uses an empty
memory table instead of failing on empty-file reads. SRCREV points at the
v0.75.0 tag; PV matches the tag so the check-version-mismatch class sees a
clean release.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Trevor Gamblin
Khem Raj