mini-snmpd: update to v2.0 - #30227
Conversation
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Verified against upstream: the dropped 0001-make-MAX_NR_INTERFACES-configurable.patch is indeed upstream now — AC_ARG_WITH(interfaces, ...) plus AC_DEFINE_UNQUOTED(MAX_NR_INTERFACES, ...) are in 2.0's configure.ac, so the mini_snmpd-16/mini_snmpd-52 variants' --with-interfaces= still works. No new library dependencies in 2.0 (the new netlink monitor uses plain AF_NETLINK sockets, no libmnl/libnl), --with-config, --enable-ethtool and --disable-ipv6 all still exist, and PKG_RELEASE reset to 1 is right for a version bump.
Two things worth a look, neither of which blocks the build: the pinned source commit is one commit past the v2.0 tag (and a release tarball exists), and 2.0's dual-stack-by-default change makes the init script's option ipv6 0 no longer mean IPv4-only. Details inline.
Generated by Claude Code
fda69d1 to
941b128
Compare
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Re-reviewed the two changes since fda69d1.
PKG_SOURCE_VERSION:=dba5d0ce... is confirmed to be exactly the v2.0 tag (git describe → v2.0, no commits past it), so the pinned source now matches PKG_VERSION:=2.0. That resolves the earlier point about the pin; keeping PKG_SOURCE_PROTO:=git over the release tarball is your call and doesn't block anything.
The init-script change does restore 1.6's IPv4-only behaviour for option ipv6 0 on IPv6-enabled builds, which was the goal. One problem with it, inline: -4 is compiled out under --disable-ipv6, which this Makefile passes when CONFIG_IPV6 is unset, so on those builds the daemon would now fail to start with the shipped default config. That one is worth fixing before merge. The second inline comment is a pre-existing -v/--version mix-up in the neighbouring line — optional.
Generated by Claude Code
1e75046 to
ef2d5fa
Compare
|
@marcin1j |
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Re-reviewed the change since 941b128.
Both earlier points are addressed. The -4 capability probe is correct: --use-ipv4 is emitted under the same #ifdef CONFIG_ENABLE_IPV6 that gates the option itself, so it can't false-positive on an --disable-ipv6 build, and -h returns immediately without touching config or sockets. -l debug is the right replacement for -v — the log level is checked at runtime, so it works in this package's build, which does not pass --enable-debug. All 12 build targets and the formality checks are green.
One thing left, inline: removing the old -6 line leaves the mirror image of the bug just fixed. option ipv6 1 meant IPv6-only in 1.6; with no flag passed it is now dual-stack, so those installs gain an IPv4 listener on upgrade. Worth confirming that's intended before merge — a suggestion restoring 1.6 semantics on both settings (using a single probe) is inline.
Generated by Claude Code
Changelog: https://github.com/troglobit/mini-snmpd/blob/master/doc/ChangeLog.md Removed upstramed patches: - 0001-make-MAX_NR_INTERFACES-configurable.patch The startup script has been modified due to a change in the default listener behavior. Also fix param to enable debug level Fixes: openwrt#29943 Signed-off-by: Ivan Pavlov <AuthorReflex@gmail.com>
dccdd24 to
20421f7
Compare
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 1 new commit; no new issues found.
The amended commit folds the -6 line back in and drives both flags from a single --use-ipv4 probe, which closes the last open point: option ipv6 0 restricts to IPv4 and option ipv6 1 restricts to IPv6 again, as in 1.6, while an --disable-ipv6 build (CONFIG_IPV6 unset) passes neither flag instead of dying on a getopt_long() usage error. ipv6 1 on an IPv4-only build is now also better off than on master, where the unconditional -6 sent the daemon into a respawn loop. The ipv6 branch stays consistent with the network_get_ipaddrs6 selection at mini_snmpd.init:156, and PROG=/usr/sbin/mini-snmpd matches the binary the install recipe copies, so the probe resolves.
Commit message still matches the diff (version bump, dropped upstreamed patch, init changes), and all 12 build targets plus the three formality checks are green on 20421f7.
Generated by Claude Code
3 participants
📦 Package Details
Maintainer: @bmork @hnyman @marcin1j
Description:
Changelog: https://github.com/troglobit/mini-snmpd/blob/master/doc/ChangeLog.md
Removed upstramed patches:
The startup script has been modified due to a change in the default listener behavior. Also fix param to enable debug level
Fixes: #29943
🧪 Run Testing Details
✅ Formalities