Update 0001-Avoid-bashisms-in-init-scripts.patch as cluster mirror log
daemon is dropped[1]
Changelog:
- Wipe stale signatures before formatting a VDO data LV on thin storage.
- Warn on non-power-of-2 thin pool chunk size; prompt when --chunksize is used.
- Fix lvmpolld shutdown when reaping a finished client thread.
- Fix dmeventd raid repair after partial replacement or resync completion.
- Retry failed dmeventd snapshot extension with exponential backoff.
- Fix dmeventd snapshot threshold advance when lvextend failed.
- Fix dmeventd snapshot unmount on Invalid/Overflow, not 100% usage alone.
- Reject undersized --pooldatavdo conversion before allocating pool metadata.
- Allow lvchange --refresh and properties on hidden thin and cache pool sub LVs.
- Allow 4K stripe size smaller than page size (except on DAX and MD RAID).
- Set recursion limit to 1024 LVs when processing device tree.
- Remove cmirrord and clustered mirror log daemon support.
- Remove legacy vgimportclone.sh script (superseded by native vgimportclone).
- Respect /run/mdadm/creating-<md_dev_name> in 69-dm-lvm.rules.
- Add `lvmconfig --edit Section/Field=Value` to set an lvm.conf setting.
- Add `lvmconfig --edit Section/Field=-` to remove an lvm.conf setting.
- Fix use-after-free of host tag list on toolcontext refresh.
- Fix memleak of config tree list nodes on toolcontext refresh.
- Fix lvmconfig output of merged and unregistered config nodes.
- Recurse into subsections when merging config trees.
- Make lvmpersist reject non-block devices and unsupported device types.
- Make lvmpersist validate `--vg`/`--device` and key options before use.
- Make lvmpersist check for required PR tools (sg_persist, nvme, mpathpersist, jq).
- Make lvmpersist clean up registered keys when interrupted during start/takeover.
- Make lvmpersist report failures from start, stop, clear, remove and query commands.
- Fix lvmpersist `--vg` device list splitting to avoid pathname expansion.
[1]https://sourceware.org/git/?p=lvm2.git;a=commit;h=f999dcc03828d097d0cd7288c224efdc07c44ac7
Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move the protocol=https into the SRC_URI so that both mirrors use it
instead of the git protocol.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move the protocol=https into the SRC_URI so that both mirrors use it
instead of the git protocol.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move the protocol=https into the SRC_URI so that both mirrors use it
instead of the git protocol.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move the protocol=https into the SRC_URI so that both mirrors use it
instead of the git protocol.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
ibus_input_context_convert_text() replaced the attribute list of an
IBusText with a new, floating list without sinking it. gjs sinks the
floating reference when gnome-shell calls get_attributes(), so gjs and
the IBusText end up sharing a single reference, and whichever releases
it second unrefs a freed object. gnome-shell crashes with SIGSEGV in
ibus_text_destroy(), typically while typing with dead keys, which ends
the session.
The same crash is reported in Debian:
https://bugs.debian.org/1146664
AI-Generated: Uses Claude Code (Claude Opus 5.5)
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
There are two routes to building jsonapi.o, and one of them adds CPPFLAGS
that then break the build. Ensure the correct route is always taken by
explicitly building it first.
This has been reported upstream and more details are in the linked mail.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Update cxx crate from 1.0.197 to 1.0.202.
Changes in this update:
v1.0.198:
- Update to syn 3
v1.0.199:
- Suppress clone_on_copy clippy lint in generated code
v1.0.200:
- Do not classify references as guaranteed POD
v1.0.201:
- Specialize enable_borrowed_range and enable_view for rust::Slice<T>
- Register bazel toolchains as dev dependencies
v1.0.202:
- Support build environments that limit build-script filesystem writes
outside of OUT_DIR
Assisted-by: Kiro CLI 2.12.0 (AI assistant)
Signed-off-by: Deepesh Varatharajan <Deepesh.Varatharajan@windriver.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Update nushell from 0.114.1 to 0.116.0.
Highlights across v0.115.0 - v0.116.0:
Breaking changes:
- YAML support reworked on serde-saphyr/granit-parser: defaults to YAML 1.2
(use --spec 1.1 for old behavior), proper tag support, multi-document
streams, anchors/aliases/merge keys, stricter mapping keys
- Parser keywords can no longer be shadowed (e.g. `def def [] {}` errors)
- The `nu` binary no longer ships with `--testbin`
- Removed `idx import`/`idx export`; `idx init` is now the single way to
build the in-memory index
- Unified custom completions: one shared input/output model across
parameter, command-wide, external completers and menu sources; existing
custom completers/menu sources may need updating. Removed the global
`background-completions` option (use the `@interactive` attribute)
- `to yaml`/`to yml` `--compact-list-indent` replaced with `--list-indent`
Features:
- New `$ans` reserved variable holding the last REPL result (output,
duration, exit code, command); `$ans.last` opt-in via
$env.config.max_last_result_size
- New Helix/Kakoune-style `helix` edit mode
- New `matrix` custom value type with 21 subcommands (ndarray-backed)
- New `tui` command family to build interactive terminal UIs from a pipeline
- `nu --dap` starts a Debug Adapter Protocol server for debugging scripts
- KDL v1/v2 support and JSON-in-KDL output for `from kdl`/`to kdl`
- Comparison operators and loose parsing for `semver` values
- More interactive `tutor` lessons (shortcuts, $env, custom commands,
pipeline input, if/match/where)
- `any`/`all` accept row conditions like `where`; `--include` for
`take while`/`take until`; `hash sha512`; `mkdir --fail-if-exists`;
`save --force` creates parent dirs; `update cells --recursive`;
named-flag null omit/pass-through and record spread into named flags
Bugfixes:
- Numerous fixes to try/catch/finally ordering, completions, YAML float
token parsing, format date locales, tables, and more
Release notes:
- https://www.nushell.sh/blog/2026-08-15-nushell_v0_115_0.html
- https://www.nushell.sh/blog/2026-09-26-nushell_v0_116_0.html
Assisted-by: Kiro CLI 2.12.0 (AI assistant)
Signed-off-by: Deepesh Varatharajan <Deepesh.Varatharajan@windriver.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Ptyxis is a GTK 4 terminal for GNOME with first-class support for
containers. It is the default terminal in Fedora Workstation and
Ubuntu.
Build it from the release tarball.
AI-Generated: Uses Claude Code (Claude Opus 5.5)
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Refreshed patches for the new version.
Behaviour changes:
- client contexts created from the SSLManager configuration verify the host
name by default
- FTPSClientSession no longer falls back to plaintext when the server
refuses TLS, unless allowPlaintextFallback() is called
- HTTP requests with duplicate or conflicting framing headers are rejected
- URIStreamOpener and HTTPSStreamFactory follow redirects only to http and
https.
Changelog:
https://github.com/pocoproject/poco/releases/tag/poco-1.15.4-release
Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Upgrade considerations:
- If you're parsing the nodename as received by connect_cb,
stats_cb, ssl_cert_verify_cb or throttle_cb, make sure you're
correctly parsing it following RFC 3986. Metadata API calls
are unaffected: they return the host and port as separate
fields.
- Make sure the brokers' certificates carry the IP in an
iPAddress entry, not a dNSName entry. Previously the
certificate was validated against dNSName (falling back to the
subject CN) whenever the address could not be parsed as an IP
literal: with OpenSSL < 3.0 or BoringSSL for any IP address,
and on all OpenSSL versions for an IPv6 address configured in
brackets in bootstrap.servers ([::1]:9092), which never
verified. A scoped address (fe80::1%eth0) was likewise
dNSName-matched on every version and now has the zone
stripped before matching. IPv6 addresses ending in :: could
not be connected to at all, as the nodename became host:::port
and failed name resolution. In the remaining cases, a bare
IPv4 or IPv6 literal on OpenSSL >= 3.0, the iPAddress entry
was already used, and behaviour is unchanged.
Changelog:
https://github.com/confluentinc/librdkafka/releases/tag/v2.15.1
Signed-off-by: Ankur Tyagi <ankur.tyagi85@gmail.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
The recipe installs /etc/polkit-1/rules.d as polkitd:root 0700, but
polkit's own tmpfiles.d entry declares it as root:polkitd 0750:
d /etc/polkit-1/rules.d 0750 root polkitd - -
As a result, systemd-tmpfiles changes the ownership and mode of the
directory at every boot, and on a read-only root filesystem it fails
with:
systemd-tmpfiles[241]: fchownat() of /etc/polkit-1/rules.d failed: Read-only file system
The polkitd:root 0700 ownership was added to avoid rpm conflicts with the
polkit-group-rule-* packages, which no longer install into /etc since
commit d5e90541f8.
Match upstream instead. This is also the stricter setting, as polkitd
can read the rules but cannot modify them.
Signed-off-by: Sergio Prado <sergio.prado@e-labworks.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Writing an image whose size is not a multiple of the bmap block size
failed at its last block with "Image ended before the block map did",
because every range was taken to end at (endBlock + 1) * BlockSize.
Add a patch that reads ImageSize from the bmap and clamps the ranges to
it.
Tested by writing a wic image whose size ends 2048 bytes into its
last 4 KiB block to an NVMe disk.
AI-Generated: Uses Claude Code (Claude Fable 5.1)
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Cast socket addresses to struct sockaddr * for bind() and sendto(), since
musl has no transparent union for them like glibc, and pin threads with
pthread_setaffinity_np() after creation where
pthread_attr_setaffinity_np() is not available.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
configure.ac's AC_SEARCH_LIBS([backtrace], [execinfo], ...) fails
with 'libexecinfo not found' on musl, which lacks glibc's
backtrace()/backtrace_symbols(); depend on the libexecinfo compat
library there.
Once configure succeeds, ticks.h's timeval2nsec() needs struct
timeval, which was only pulled in via <sys/time.h> under __FreeBSD__,
and vma-xlio-redirect.cpp calls sscanf()/sprintf() without including
<cstdio>; fix both for musl too.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Several independent musl gaps surfaced one after another as each was
fixed and the build progressed further:
- android-base/off64_t.h only worked around __APPLE__ lacking off64_t;
extend the same typedef fallback to musl.
- sysdeps.h's adb_pread/adb_pwrite/adb_lseek only fell back to the
plain (non-LFS64) functions on __APPLE__; musl needs the same
fallback for pread64/pwrite64/lseek64.
- daemon/auth.cpp calls b64_pton() from <resolv.h>, which the Linux
host build gets via -lresolv and bionic provides natively, but musl
has neither; vendor the standard BSD implementation for it.
- android-base/endian.h and threads.h already gate a musl-aware branch
behind ANDROID_HOST_MUSL, matching upstream's own expectation, but
nothing defined that macro for this recipe's build; define it via
CFLAGS/CXXFLAGS for the musl toolchain.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
netdata's cmake build unconditionally requires systemd for the
'Systemd units' plugin unless explicitly disabled. Tie
ENABLE_PLUGIN_SYSTEMD_UNITS to the same DISTRO_FEATURES systemd check
already used for the journal option, so sysvinit builds configure
cleanly.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
The bundled liblog/libbase compat shim defines a getprogname()
fallback macro and then immediately redefines program_invocation_name
using it in a way that collides with musl's own declaration. Fix the
ordering/guard so the two definitions don't conflict.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
backtrace.cpp sets sCrashCallback but only reads it back inside a
glibc-only execinfo backtrace path, so on musl (no backtrace()) clang
flags it as -Wunused-but-set-variable under -Werror. Reference it in
the non-glibc branch too.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
spdm_emu_mctp_kernel.c uses AF_MCTP, a very recent Linux socket
family constant that musl's <sys/socket.h> doesn't define yet. Add a
guarded fallback definition.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
modbus-rtu.c uses the Linux-specific TCGETS2/TCSETS2 ioctls, which
glibc's <asm/termbits.h>/<asm/ioctls.h> chain declares transitively via
<termios.h>, but musl's don't.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
secure_execution_linux.c includes both <sys/prctl.h> and
<linux/prctl.h>; musl's <sys/prctl.h> already declares everything
needed, and including the raw uapi header afterwards redefines struct
prctl_mm_map.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Same struct ethhdr / __UAPI_DEF_ETHHDR ordering issue as xdp-tools:
common.h/txring.c pull in musl's <netinet/if_ether.h> after the raw
<linux/if_ether.h> uapi header has already defined struct ethhdr.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
musl's <netinet/if_ether.h> and the kernel's raw <linux/if_ether.h>
both define struct ethhdr, and only avoid colliding if
<netinet/if_ether.h> (or something that pulls it in) is processed
first. lib/util/params.{c,h} and xdpsock.c, xdp-filter.c and
xdp-dump/xdpdump.c all got this backwards in various ways; fix the
include order/choice in each.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
atop calls the wide-character ncurses function waddwstr(), which
needs NCURSES_WIDECHAR defined; get that (and the right include path)
from the ncursesw pkg-config module's own Cflags instead of hardcoding
plain ncurses flags.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
A test program uses pthread_setname_np(), a GNU extension that needs
_GNU_SOURCE defined before <pthread.h> is included to be declared.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
The camel-provider-imapx test uses strcasestr(), a GNU extension that
needs _GNU_SOURCE defined before <string.h> is included to be
declared.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
fs-monitor.c uses __s32/__u32 without <linux/types.h>, and rw_hint.c
uses RWH_WRITE_LIFE_NOT_SET which musl's headers don't define. Add the
missing include and a guarded fallback define.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
lib/cap-cli-main.c uses htole32()/le32toh() from <endian.h> without
including it directly; pull it in via the project's own endian
compat header.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
ec.h unconditionally #defines _GNU_SOURCE without checking if it is
already defined, which trips -Werror on musl toolchains that predefine
it via the driver. Guard the define.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
dvd_filesystem.h unconditionally uses the glibc LFS64 off64_t name.
musl's off_t is already 64-bit, so typedef it the same way the other
non-glibc branches already do. This also fixes libdvdnav, which
consumes this staged header.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
bit7z's internal code uses off64_t/stat64/lstat64, which musl doesn't
provide since its off_t is already 64-bit. Use the plain POSIX names.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
sofprobeclient uses the bare 'uint' typedef, which is a glibc/BSD
extension musl doesn't provide. Use 'unsigned int' instead.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
StreamCallback.h uses int64_t without including <cstdint>, relying on
a transitive include that only exists on glibc.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
args.h uses int64_t without including <cstdint>; glibc headers pull it
in transitively via other includes, musl's don't.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
canvenient_socketcan.c uses struct timeval without pulling in a header
that declares it; glibc gets it transitively, musl doesn't.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Ankur Tyagi
Ryan Eatmon
Markus Volk
Ross Burton
Deepesh Varatharajan
Sergio Prado
Khem Raj