udp-broadcast-relay-redux: Add package back (re-pointed at new fork location) [squash before merge will resolve guidelines issues]

[accwebs]

📦 Package Details

Maintainer: @accwebs

Description:
This brings back the udp-broadcast-relay-redux package that was dropped in #29452. @accwebs has forked the now-archived upstream and is re-pointed at the new fork location in this PR.

I am happy to squash the commit history if that's desirable before we merge. For now I deliberately have tried to keep the revert changes, directory renames, and whitespace changes as their own commits. I personally hate trying to review diffs where those things are all included in the same commit; you can't focus on the actual changes.

Details on changes:
Please simply advise whether you would want it squashed all down to 1 commit or if you want something kept split out.

Commit 1: Revert the commit that removed the package.

Commit 2: Rename package directory to remove -openwrt suffix

• I somehow goofed when I created the package years ago.

Commit 3: Switch package to new fork location

• Point at new location https://github.com/accwebs/udp-broadcast-relay-redux
• Update to latest commit hash

Commit 4: Modernize Makefile. Add test-version.sh

• Add PKG_MAINTAINER
• Add PKG_LICENSE_FILES
• Fix PKG_LICENSE to GPL-2.0-or-later
• Add TARGET_CPPFLAGS and TARGET_LDFLAGS
• Switch to $(INSTALL_BIN) for /usr/sbin file
• Fix whitespace
• Add test-version.sh since the program does not implement --version flag

Commit 5: Whitespace in .init & .config file

• Switch to tabs in both .init and .config file

Commit 6: init file improvements]

• Set SIGKILL timeout to 1s; jail/PID1 rules prevent SIGTERM from stopping it
• Enable stdout/stderr routing to log
• Remove unnecessary jail arg cgroupsns

---

🧪 Run Testing Details

• OpenWrt Version: 25.12.5-ish, r33143-d78c0fd16d (buildroot SNAPSHOT from Aug 10th, 2026 on openwrt-25.12 branch)
• OpenWrt Target/Subtarget: mediatek/filogic glinet_gl-mt6000
• OpenWrt Device: GL-MT6000

---

✅ Formalities

• I have reviewed the CONTRIBUTING.md file for detailed contributing guidelines.

[openwrt]

Formality Check: Failed

We checked this pull request against the contribution guidelines. Here is what needs your attention:

🛑 CRITICAL ERRORS

Commit 47c306d - udp-broadcast-relay-redux: revert "udp-broadcast-relay-redux: remove package":

• New OpenWrt package is missing the mandatory parameter: 'PKG_MAINTAINER'
• New OpenWrt package is missing the mandatory parameter: 'PKG_LICENSE_FILES'
• Makefile line 'This program listens for packets on a specified UDP broadcast port.' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'When a packet is received, it sends that packet to all specified interfaces but' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'the one it came from as though it originated from the original sender.' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'The primary purpose of this is to allow games on machines on separated' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'local networks (Ethernet, WLAN) that use udp broadcasts to find each other to do so.' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'It also works on ppp links, so you can log in from windows boxes (e.g. using pptp)' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'and play LAN-based games together. Currently, you have to care about upcoming or' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces
• Makefile line 'downgoing interfaces yourself.' inside 'Package/udp-broadcast-relay-redux/description' must be indented with at least 2 spaces

⚠️ STYLISTIC WARNINGS & SUGGESTIONS

Commit 47c306d - udp-broadcast-relay-redux: revert "udp-broadcast-relay-redux: remove package":
PKG_LICENSE value 'GPL-2.0' is deprecated by SPDX. Use 'GPL-2.0-only' or 'GPL-2.0-or-later' instead.

Tip

Do not close this pull request to make corrections. Instead, modify your existing commits (e.g. git commit --amend) and update the branch using git push --force-with-lease --force-if-includes. The checks will re-run automatically.

---

_{Something broken? Consider reporting an issue.
Running version 3eeda58 deployed on 2026-08-18 11:06:29 CEST}

[openwrt-ai]

Reviewed 4 new commits.

Verified out-of-band: the fork accwebs/udp-broadcast-relay-redux exists, aa9adef316824e88a28b0e19a4a7a29ababe92fc is the current master HEAD, LICENSE is present at that commit (GPLv2 text), and main.c is byte-identical to the previously packaged commit 6713729, so all of --id/--port/--dev/-s/-t/--multicast used by the init script still parse. The reverted SHA 90b98c1 in fde50d4 is correct and the revert is faithful.

Commit checks

• cbf1189 "udp-broadcast-relay-redux: switch package to new fork location" — missing blank line between subject and body, so the bullet list is folded into the subject. git log --oneline renders it as one 150-character subject: udp-broadcast-relay-redux: switch package to new fork location * Point at new location https://github.com/accwebs/udp-broadcast-relay-redux * Update to latest commit hash. Insert a blank line after the subject.
• afbadf9 "udp-broadcast-relay-redux: update package Makefile to modern conventions" — same problem; the three * Add ... bullets are part of the subject line rather than the body.

---

Generated by Claude Code

[openwrt-ai]

Reviewed 2 new commits (ef8a406, d2c6058) since my last review at afbadf9; the branch was force-pushed, so dcdb1a1/ef8a406 are the rewritten versions of cbf1189/afbadf9.

The commit-message folding I flagged last time is fixed — both rewritten commits now have a proper blank line between subject and body, and FormalityCheck / Git & Commits is green. Both new commit messages match their diffs, so no commit checks this round.

Verified out-of-band against accwebs/udp-broadcast-relay-redux@aa9adef3: LICENSE is present and is the GPLv2 text, and main.c's header reads "either version 2 of the License, or (at your option) any later version", so the PKG_LICENSE:=GPL-2.0-or-later change is correct. PKG_SOURCE_DATE:=2026-07-26 also matches that commit's date.

Inline: the build is red on all 10 targets for a fixable reason (stale PKG_MIRROR_HASH), the Makefile formality check is red for a series-hygiene reason, and the tab conversion in the init script flattened two levels of nesting.

---

Generated by Claude Code

[openwrt-ai]

Reviewed the force-pushed series (5 commits) since my last review at d2c6058. All three previously flagged items are fixed:

• PKG_MIRROR_HASH is now 5462b1d1…, and every target gets past download and compiles.
• The init script is consistently hard-tab indented; the two flattened blocks (|| { … } and the for net loop) now nest correctly, and the uci_validate_section continuation lines were indented as well.
• The new commit 0481de6 renames net/udp-broadcast-relay-redux-openwrt/ → net/udp-broadcast-relay-redux/, aligning the directory with PKG_NAME. Verified content-neutral (R100 on all three files), and there are no leftover references to the old path anywhere in the feed.

Verified locally against the base checkout:

• a8d4687 is a faithful revert — the restored tree is byte-identical to 90b98c14^:net/udp-broadcast-relay-redux-openwrt, and 90b98c14 is indeed the removal commit.
• PKG_VERSION is intentionally absent; include/download.mk derives it from PKG_SOURCE_DATE/PKG_SOURCE_VERSION, which matches the 2026.07.26~aa9adef3-r1 the CI installs. PKG_RELEASE reset to 1 in 7a358be is correct for a source bump.
• The init script is functionally sound: procd_open_instance with no name auto-numbers via _procd_open_instance, so the multi-section case doesn't collide, and cgroupsns is a valid procd_add_jail flag. conffiles is unindented with a correct absolute path.

All five commit messages match their diffs, so no commit checks this round.

One new inline finding: the five targets that run the runtime suite now fail on the generic version check rather than on the build. That is the documented test-version.sh case, not a defect in the code.

FormalityCheck / OpenWrt Makefiles is still red on a8d4687 for the introduce-then-fix reason I raised on the two previous rounds — not repeating it inline since the existing thread still anchors to the current Makefile. Worth folding before undrafting if you want the not following guidelines label to clear, but it is a series-shape question rather than a problem with the end-state tree, which looks correct.

---

Generated by Claude Code

[openwrt-ai]

Reviewed the force-pushed series (5 commits) since my last review at 8387db5; no new issues found.

The branch was rebased onto current master and the commits rewritten, but the only change to the package itself since the last reviewed head is the new net/udp-broadcast-relay-redux/test-version.sh — git diff 8387db5..89d5e51 -- 'net/udp-broadcast-relay-redux*' is exactly that one added file. It matches the in-tree convention (#!/bin/sh, # shellcheck shell=busybox, case "$PKG_NAME" with an explicit *) failure arm, mode 0755, no grep -q), and it is folded into 173233c rather than added as a separate fix commit, so the series shape did not grow.

That closed out the last inline finding: all 10 "Feeds Package Test Build" jobs are now green on 89d5e51, including the five that run the runtime suite and previously failed the generic version probe with Unknown arg: -?.

All five commit messages match their diffs — including 173233c, whose seven body bullets each correspond to a hunk (PKG_MAINTAINER, PKG_LICENSE_FILES, GPL-2.0-or-later, TARGET_CPPFLAGS/TARGET_LDFLAGS, $(INSTALL_BIN), description re-indent, test-version.sh) — so no commit checks this round.

One item carries over unchanged: FormalityCheck / OpenWrt Makefiles is still red, now on e174336, for the introduce-then-fix reason raised on the previous three rounds (the revert restores the pre-2019 Makefile without PKG_MAINTAINER/PKG_LICENSE_FILES and with tab-indented description lines; 173233c adds them two commits later, and the check evaluates each commit individually). The two threads about it are marked resolved, so noting it here rather than opening a third: the end-state tree is correct, and this only matters if you want the check green and the not following guidelines label cleared before undrafting.

---

Generated by Claude Code

[accwebs]

Okay I think I've got this right now. Hopefully. Kinda a noob at this.

[openwrt-ai]

Reviewed 1 new commit (647de88) since my last review at 89d5e51. The branch was rebased onto current master and the earlier five commits rewritten, but git diff 89d5e51..647de88 -- 'net/udp-broadcast-relay-redux*' is confined to the init script — Makefile, test-version.sh and the .config file are byte-identical to the last reviewed head.

Verified against procd.sh: term_timeout is a valid procd_set_param key (line 260), as are stdout/stderr, and log is a valid procd_add_jail flag. All 10 "Feeds Package Test Build" jobs are green on 647de88, including the five that run the runtime suite.

One inline finding on the jail flags.

Commit checks

• 647de88 "udp-broadcast-relay-redux: init file improvements" — the third bullet says "Remove unnecessary jail arg cgroupsns", but the diff does not just remove it: procd_add_jail ubr-${PIDCOUNT} cgroupsns becomes procd_add_jail ubr-${PIDCOUNT} log signals. log is arguably covered by the stdout/stderr bullet, but signals is a new flag that no bullet mentions (and, per the inline comment, has no effect). Worth either dropping it or describing the swap in the body.

The only red check remains FormalityCheck / OpenWrt Makefiles, still failing on 48adbce for the introduce-then-fix reason covered in the earlier (now resolved) threads — the bot comment above has the details. Unchanged in substance from previous rounds; the end-state tree is correct and this is only about series shape.

---

Generated by Claude Code

[openwrt-ai]

Reviewed 1 new commit (34673a3) since my last review at 647de88; no new issues found.

The branch was rebased onto current master and all six commits rewritten, but the only change to the package tree since the last reviewed head is the one line I flagged last round — git diff 647de88..34673a3 -- 'net/udp-broadcast-relay-redux*' is exactly procd_add_jail ubr-${PIDCOUNT} log signals → procd_add_jail ubr-${PIDCOUNT} log. Makefile, test-version.sh, the .config file and the rest of the init script are byte-identical to 647de88.

Verified against _procd_add_jail in procd.sh: log is in the case list, and signals — which was being silently dropped — appears nowhere in procd.sh. term_timeout and stdout/stderr remain valid procd_set_param keys, so the rest of 34673a3 stands as reviewed.

That also closes the commit check I raised last round: with signals gone, the three body bullets of 34673a3 account for the whole diff (term_timeout 1, stdout/stderr, and the cgroupsns → log jail flag, covered by the log-routing bullet). All six commit messages match their diffs, so no commit checks this round.

All 10 "Feeds Package Test Build" jobs are green on 34673a3, including the five that run the runtime suite.

The only red check is FormalityCheck / OpenWrt Makefiles, still failing on the first commit 0d52e2c for the introduce-then-fix reason the bot comment above spells out (the revert restores the pre-2019 Makefile without PKG_MAINTAINER/PKG_LICENSE_FILES and with tab-indented description lines; b312b4f adds them three commits later, and the check evaluates each commit individually). Unchanged in substance from the previous four rounds and already covered by resolved threads, so not re-opening it inline — the end-state tree is correct, and this only matters if you want the check green and the not following guidelines label cleared. Given your question in the PR body: squashing the series (or at least folding b312b4f's Makefile hunks into 0d52e2c) is what would resolve it.

---

Generated by Claude Code

[accwebs]

@BKPepe sorry for the ping; following up on #29452. This re-adds udp-broadcast-relay-redux pointed at a maintained fork (I'll try to maintain it for now; I also might later re-point if another 'better' fork emerges). The automated review rounds are resolved to the best of my knowledge; the one remaining red check is a per-commit artifact of the revert commit being deliberately separated out to aid with reviewing (seeing what I've changed from the prior package's implementation). LMK how you want the final commits structured (all in one? revert separate?) and I'll squash once reviews are done.

[accwebs]

Oops, github rebase (to update onto latest target commits) didn't seem to work. Will do it locally shortly.

[accwebs]

Fixed. Ready again.

[openwrt-ai]

Reviewed 6 new commits since my last review at 34673a3; no new issues found.

All six commits were rewritten by the rebase onto current master (a583d91), but the package tree is unchanged: git diff 34673a3..d5a78ac -- 'net/udp-broadcast-relay-redux*' is empty, so the Makefile, the init script, the .config file and test-version.sh are byte-identical to the last reviewed head. The per-commit shape is also unchanged (revert → directory rename with R100 on all three files → fork re-point → Makefile modernization + test-version.sh → whitespace → init improvements), and each message still matches its diff, so no commit checks this round.

All 10 "Feeds Package Test Build" jobs are green on d5a78ac, including the five that run the runtime suite, and FormalityCheck / Git & Commits and Code Patches are green.

The only red check remains FormalityCheck / OpenWrt Makefiles, still failing on the first commit 47c306d for the introduce-then-fix reason the bot comment above spells out — the revert restores the pre-2019 Makefile without PKG_MAINTAINER/PKG_LICENSE_FILES and with tab-indented description lines, and c38747f adds them three commits later, while the check evaluates each commit individually. Unchanged in substance from the previous five rounds and already covered by resolved threads, so not re-opening it inline: the end-state tree is correct. To answer the question in your PR body — squashing the series, or at minimum folding c38747f's Makefile hunks into 47c306d, is what clears this check and the not following guidelines label.

---

Generated by Claude Code