Repository navigation
Description
Context
On 2026-10-18 the fork had 109 open issues, 79 of them unclaimed. Claims mostly go to the newest issues, so older ones get forgotten, duplicated, or stay open after they are fixed. Examples from tonight: #192 was fixed twice (#260, #772) but is still open because the Fixes keyword was lost in the squash; #401 was implemented by #791, which does not link it; #1042 has a maintainer comment "all four points look resolved" from 10-17.
Proposal
A read-only backlog review of the unclaimed open issues:
- When: as part of the coordinator's backlog retrospective (
.claude/skills/obs-coordinator/SKILL.md§6, #1420), or when there are more than about 60 unclaimed open issues. - What: for each unclaimed open issue, collect the evidence (merged and open PRs that close or mention it, master commits since 2026-10-08, the features ledger, links between issues, whether the files it links still exist, text similarity to all other issues, the last comment), then check the flagged ones by hand against master.
- Output: one comment or issue with a verdict per issue: A (likely done), B (goal met another way or superseded), C (partly done: rescope or refresh), D (claim state), E (waiting for information or hardware), F (still valid). The maintainer closes or rescopes. Nobody closes on heuristics.
- Pitfalls found in the first run: the fork's git log contains upstream history, whose old PR numbers collide with this repo's issue numbers (for example
Merge pull request #43 from fryshorts, 2014). Count only commits from 2026-10-08 on (this removed 24 false matches). The ifdef ratchet scans only C/C++, so CMake inventories must be checked by hand. "Mentioned in a merged PR" alone is weak evidence.
If you want this, I can put the scan script into the repo in a separate PR (stdlib Python, reads gh --json output).
First run (master 9907fc43a, 2026-10-18)
79 unclaimed open issues: A 9, B 6, C 22, D 6, E 6, F 30. No duplicates among the open issues (the high similarity scores point to closed predecessors that an issue replaced on purpose, for example #43 → #6).
A. Likely done: please close (or say why not)
| Issue | Evidence |
|---|---|
| #43 Phase 3 remaining ports | All ported: utf8.rs, lexer.rs, cf_tokenizer.rs (#49), text_lookup.rs, dstr.rs, bmem.rs (#68, #59), base.rs (#61). |
| #45 Phase 5 remaining | All port items landed (#148–#151 closed; #352, #184, #420, #745; parsers in obs-codec; Tier 4). The checklist was never updated (nathan-v, 10-15: "I can't edit this checklist"). One item left: "Document the aarch64 last-bit difference" (related open PR #1432). Close and file that item on its own? |
| #153 Phase 4 remaining | All done except platform.c (own issue #262) and the oracle stand-ins (#145, see C). |
| #4 meta: roadmap phases 2–5 | Done when #43, #45 and #153 close. |
#192 Flaky pipe_parity dead-pipe write |
Fixed by #260 (Fixes #192) and #772 (title "Fixes #192", lost in the squash). No new report since 10-14. |
| #401 KDE/GNOME: show a hotkey that runs on a different key | Implemented by #791: "▲ other key" state, a banner, test localized_desktop_key. |
| #1042 appimage bundle check after glibc 2.28 | Your comment, 10-17: "all four points look resolved". |
| #635 Release retrospective, betas 10-11 to 10-13 | Completion audit (10-14): all fixes #621–#633 merged. Only "hardware confirmation" left. |
| #1261 migrate settings from obs-studio to obs-studio-rust | #706 (merged 10-14) migrates the upstream settings on first run. The issue is only a link to a post. Probably an older build or an edge case. |
B. Goal met another way, or superseded: your decision
| Issue | Evidence |
|---|---|
| #448 valgrind recheck after clean hotkey shutdown | Superseded by #449 (completed 10-13): clean shutdown under ASan/LeakSanitizer, OBS-owned leaks fixed. |
| #852 facade: obs-nvenc, #859 facade: nv-filters | No entries in ci/platform-ifdef-baseline.json any more: since the unity build these files compile through platform-named groups, so path scoping works. The #ifs are still in the source (nvenc.c.h: 13). You noted this on #859 on 10-17. |
| #878 facade: libobs public headers (record) | Now kept in ci/platform-ifdef-exceptions.toml (#1051), with a reason per header. |
| #876 facade: libobs ported units (record) | "No work except removing the C files when the C originals are dropped." A record, not a task. |
| #96 Agent friendly CLI | The ledger lists obs-stream-rust (#30, #276, --json) for #96. The MCP part is not done: close, or rescope to MCP? |
C. Partly done: rescope or refresh
- #143: unsafe lint (#290) and sanitizers (#171) are done, Miri is not in CI → rescope to Miri.
- #145:
test_stubs.cis gone, #1094 compilesplatform.c, #430 checks drift.platform_host.candobs_properties_host.cremain → done, or keep for those two? - #870: 3 of 14 left (
obs-hotkey,media-remux,obs-audio-controls). #856: 2 of 9 left (both in the pch). - #888, #889: the CMake inventories name files that moved (
cmake/os-*.cmake,cmake/rust.cmake); the if-chains are still infrontend/CMakeLists.txt,libobs/CMakeLists.txtandfrontend/cmake/*.cmake. - Facade inventories with stale file names and line numbers after the unity renames (
.c→.c.h,.cpp→.cpp.hpp): #851, #853, #854, #862, #865, #867, #869, #871, #872, #873, #874, #875, #877, #879, #881, #882, #883, #884, #885, #887. Still valid; refresh the inventory from the baseline before work starts.
D. Claim state
- #1092: I claimed the Linux sandbox part on 10-17 in the design comment, but without the
Claiming:prefix, so the bot did not register it. Waiting for your design decision. - #143, #516: bot claims expired (10-15).
- #476: parked on purpose (your comment, 10-16).
- #266: unclaimed, but draft PR #493 exists.
E. Waiting for information or hardware
#107, #617, #891, #1028, #1349, #1395 (next "Scheduled" run, 10-19 04:35 UTC).
F. Still valid
#443, #444, #516, #525, #613, #697, #700, #752, #753, #755, #756 (the four proposals from 10-14 have no maintainer reply yet), #782, #847, #897, #962, #1027, #1273, #1275–#1279, #1295, #1300, #1344, #1345, #1359, #1390, #1427.
🤖 Generated with Claude Code
Metadata
Metadata
Assignees
Labels
Type
Projects
- StatusShow more project fieldsReady
Activity
zackees commentedon Oct 25, 2026
Coordinator: thanks, this is useful and I'm working through it. The coordinator's work list now runs the periodic part of this proposal every tick (
triagecovers each issue without a priority and size, oldest first). Done so far:Next ticks: section A (meta: Rust port unit-testing roadmap, phases 2–5 #4, Phase 3 remaining ports: utf8, lexers, text-lookup, dstr, base, bmem #43, Phase 5 remaining: graphics math, media-io, parsers, obs-websocket Tier 4 #45, meta: Phase 4 remaining ports (replaces #44) #153, Release retrospective: daily betas 2026-10-11 to 2026-10-13 #635). I'll check each against master and close it with evidence or say what's left. Then sections C and D. /ready
zackees commentedon Oct 25, 2026
/ready
zackees commentedon Oct 25, 2026
Worker sweep for one pitfall class: an issue a merged PR meant to close, but listed after the first keyword ("Closes #A and #B" closes only #A). I checked the last 400 merged PRs: for each closing keyword, the issue numbers in its clause, against GitHub's
closingIssuesReferencesand the open-issue list. One real case: #1294, from #2410, which I closed by hand earlier today. One false positive, #2317, a plan that #2358 only mentions ("step 2 of the #2317 plan"). None left open.zackees commentedon Oct 27, 2026
Claiming: #1436 — the section-A pass the coordinator listed on 10-25 ("Next ticks: section A (#4, #43, #45, #153, #635)"). Read-only evidence against master
0ef8117b8. No closes: the issue says the maintainer closes or rescopes, nobody on heuristics.#43 — the Rust units are all on master, but the C is still compiled.
utf8.rs,lexer.rs,cf_tokenizer.rs,text_lookup.rs,dstr.rs,bmem.rs,base.rsare inrust/obs-util/src/, each with itsffi/shim. Howeverlibobs/util/{utf8,lexer,cf-lexer,cf-tokenizer,text-lookup,dstr,bmem,base}.c.hare all still present, and naming.md §37 deletes a C component's sources only "when fully ported". So "ported" is true at the unit level (Rust + ffi exist) and false at the component level (the C is still built). Which one #43's checklist means is the call that keeps it open; I did not close it.#45 — the listed work is present too, with one doc item left.
quat.rs,matrix3.rs,matrix4.rs,plane.rs,bounds.rs,math_extra.rsare inrust/obs-graphics/src/(the checklist still shows them unticked, though rust: Port graphics/quat.c to obs-graphics #148–rust: Port media-io/video-frame.c to obs-media-io #151 are closed).audio_io.rs,video_frame.rs,video_io.rs,format_conversion.rsare inrust/obs-media-io/src/.vec2_normFMA contraction). No issue of its own.#153 — same picture.
profiler.c(#152) is closed andprofiler.rs/profiler_io.rs/profiler_snapshot.rsare on master;pipe.rs,threading.rs,task.rs,platform/(mod.rs,unix/,windows/,stdio.rs),platform_file.rs,platform_mkdirs.rs,platform_time.rsare present; the C originals (profiler.c.h,pipe.c.h,pipe-windows.c.h,platform-*.c.h) are still compiled. Open: the oracle stand-ins, #145.#635 — all fixes merged; only the hardware confirmation it names is left. The 10-14 completion audit verified #621–#633 merged and their merge commits ancestors of
a48a9a13d. Nothing release-blocking is open. Close, or rescope to the hardware confirmation.#4 — done when #43/#45/#153 close (same unit-present / C-present picture above).
Not checked here: the section-C inventories (
ci/platform-ifdef-*) against the.c→.c.hrenames, and each port's component-level C deletion.zackees commentedon Oct 30, 2026
Claiming: #1436 — section A of the coordinator's list (#4, #43, #45, #153, #635), read-only evidence against master
e2782b4ac.Plan: for each tracker, check every unchecked item against the tree (ported crate/module present, shim swapped) and the merged-PR history, then post one comment with a verdict per issue (A–F). No closes — the maintainer closes or rescopes.
2 remaining items
zackees commentedon Oct 30, 2026
Addendum to the section-A review above — the discriminator, since the 10-27 pass on this issue left exactly that question open.
That pass wrote: "the Rust units are all on master, but the C is still compiled …
libobs/util/{utf8,lexer,cf-lexer,...}.c.hare all still present, and naming.md §37 deletes a C component's sources only 'when fully ported'. So 'ported' is true at the unit level and false at the component level." The file-present argument does not decide it, because the presence of a.c.hsays nothing about whether it is built:libobs/CMakeLists.txt:33compiles the unported set (LIBOBS_SOURCES) intolibobs-objectsunconditionally.LIBOBS_PORTED_SOURCES,LIBOBS_WINDOWS_PORTED_SOURCES, …) go throughobs_libobs_add_c_sources(), and inlibobs/cmake/implementation.cmakethat adds them only toOBS_LIBOBS_C_TARGETS— which is empty unlessOBS_BUILD_C_REFERENCE=ON(the default backend isHYBRID, whereOBS_LIBOBS_C_TARGETSstays empty, line 17-23, and the comment at line 34 says it outright: "Without a C target (the default hybrid) nothing compiles these groups").So the machine-checkable discriminator for "is this port still shipping C" is which unity group the
.c.hsits in, not whether the file exists. Applying it:.c.hgroup on masterprofiler.c.hlibobs.ported-util-io+platform*.c.h(core, file, filename, mkdirs, time, conversions)libobs.ported-platform+,libobs.ported-windows-platform+platform-nix.c.hlibobs.ported-posix+threading-posix.c.h,threading-windows.c.hlibobs.ported-posix-threading+,libobs.ported-windows+pipe-windows.c.hlibobs.ported-windows-pipe+cf-parser.c.h,cf-tokenizer.c.hlibobs.ported-util+cf-lexer.c.hlibobs.util-cf+(unported)That is what makes #43 "one item left (#1499)" and #153 "six items done, #145 open" rather than "everything is still C", and it is also why
crc32.c.hcan still contain the original C table whilecalc_crc32comes fromrust/obs-util/src/ffi/crc32.rs(#[unsafe(no_mangle)]) in every shipped build: the C copy only exists forlibobs-c-reference.No closes, no edits — this is a note on the review above so the next tick does not have to re-derive the rule.
zackees commentedon Oct 30, 2026
Backlog review, sections C and D, read-only against master
93b911a67Scope: the coordinator's "Then sections C and D" (10-25 comment). Evidence is the tree at
origin/master(93b911a67) plus issue/PR state read now. Nothing was closed, rescoped or changed: verdicts only, as the issue asks.Section C — the inventory list is now almost entirely closed
The first run's section C is stale by closure, not by content:
So the ".c → .c.h stale inventory" refresh this section asked for is no longer pending work; every row that carried it has landed and closed. Two open items remain.
#143 "Miri, sanitizers and an unsafe-code lint" — C, two of the four work items remain.
#![deny(unsafe_code)]for the safe core (item 1): done. Every port crate states it insrc/lib.rs(forbidinrust/libobs-rust,rust/obs-build,rust/obs-frontend;denyplus a scoped#[allow(unsafe_code)]on theffimodule in the rest)..github/workflows/sanitizers.yamlruns ASan+UBSan, TSan and valgrind over the cmocka tests plus a timed headless OBS launch, andbuild-project.yamlcalls it asSanitize 🧪whenci_scopeselectsasan(bug: master's Sanitizers lane is red — asan finds 8 conversion sites the Tier 2 build never compiles, valgrind finds a v4l2 format mismatch #2923).clippy::undocumented_unsafe_blocks(item 2): not enabled.grep -rn undocumented_unsafe_blocksover the tree returns nothing.grep -rni miri .github ci rust/toolsreturns nothing, andrust-toolchain.toml/Cargo.tomlname nomiricomponent.Suggested action: tick items 1 and 4, and rescope the body to the two remaining items (or file one sub-issue each, as the 10-25 triage said).
#145 "Oracle stand-ins can diverge" — C, unchanged.
rust/obs-c-oracle/oracle/platform_host.candrust/obs-c-oracle/oracle/obs_properties_host.care still in the tree, andrust/obs-c-oracle/build.rsstill compiles everyoracle/*.c("Every oracle/*.c is compiled (#137)"), so both stand-ins are live, exactly as the 10-18 run recorded. Nothing to refresh: the two files are the whole remaining scope.Section D — claim state
no_sandbox): open, and now unclaimed — vpacher's claim lapsed and there is no assignee. Linux and macOS are done (obs-browser#3/Phase 5: beyond util — graphics math, media-io, AVC/HEVC/AV1 parsers, obs-websocket Tier 4 #8 and obs-browser: Update to 63fd678 (macOS sandbox, no hang on quit) #2402,6a7ef1742); Windows is the only platform left, and the code shows it:UseCefSandbox()has only__APPLE__and#elif !defined(_WIN32)arms (obs-browser-plugin.cpp:304-324), so on Windowscef_dirstays empty,ProbeSandbox()returns{}(browser-sandbox.cpp, the#elsearm),DecideSandboxtherefore cannot allow it, andsettings.no_sandbox = truestays atobs-browser-plugin.cpp:364. A rescope to "Windows CEF sandbox" is accurate; the issue headline is still true there.What is left of the 10-18 list after A–D
Of the first run's section F ("still valid"), 16 are now closed (#525, #613, #752, #753, #962, #1027, #1275–#1279, #1300, #1344, #1345, #1359, #1427). Still open and unclaimed: #443, #444, #516, #697, #700, #755, #756, #782, #847, #897, #1273, #1295, #1390 — and, from C/D, #143 and #145. Every facade/conditional inventory row is closed.
Suggested next pass: skip sections A–D (no open rows left in A/B; C and D are down to the five issues named above) and review only those open F rows when there are more than ~60 unclaimed open issues again.
zackees commentedon Oct 31, 2026
Claiming: #1436 — the section-F pass the 10-30 review named ("review only those open F rows").
Plan: read-only evidence against
origin/master, one verdict (A–F) per open first-run F row: #443, #444, #516, #697, #700, #755, #756, #782, #847, #897, #1273, #1295, #1390, plus #143 and #145 from the C/D pass. For each: does the tracker's checklist still match the tree (ported group vs still-compiled C, file present/moved), is there a merged PR or master commit since 2026-10-08 that finished an item, is it now duplicated or superseded, and is anything claimable as a real task. No closes, no edits — verdicts only, the maintainer closes or rescopes.zackees commentedon Oct 31, 2026
Backlog review, section F (the 10-30 list), read-only against master
3e17f5d0dScope: the open first-run F rows, plus the two C/D leftovers, as the 10-30 pass named: #443, #444, #516, #697, #700, #755, #756, #782, #847, #897, #1273, #1295, #1390, #143, #145. Evidence: the tree at
origin/master(3e17f5d0d), each tracker's sub-issue states, and merged PRs since 2026-10-08. Nothing was closed, rescoped, relabelled or edited — verdicts only, the maintainer closes or rescopes.Context for the "more than ~60 unclaimed" trigger in the proposal: 58 open issues, 41 unclaimed right now, so this pass is early relative to that rule; it is the named F follow-up, not a scheduled one.
Verdicts
3e17f5d0dbuild-project.yamlhasrust-tests-macos/-windows(§2062, §2210) andrust-tier2-macos/-windows(§2883) with per-jobskip_*gate outputs;ci/local_gate.pydefines themacos-native/windows-nativelanes (exit 75 off-host,§835-885);ci-attestations.ymlcarriesrust/aarch64-apple-darwin/test,cpp/macos-arm64/tier2,cpp/windows-x64/tier2→ those lanes and maps them to the jobs.docs/ci-tooling.md§371/§510-516 documents the table and §681 the audit sampling. It is assigned to me and unchanged since 10-31 01:39; only remaining step is one dev-cycle observation of a skip on real hardware.libobs/build/libobs.ported-data+.h.c("ported obs-data.c"),obs-properties.c.handobs-hotkey-name-map.c.hsit inlibobs.ported-media+;rust/obs-datahasdata.rs+ffi. What is left is the part the checklist never made claimable per unit: the privateobs-internal.hfield inventory, the adapter seams for remaining C users ofobs-source/scene/output/obs.clayouts, and theENABLE_RUST_LIBOBS=ONapp validation. Rescope to those three, or split them.docs/ci-tooling.md§206-229: per-alias cold cross-build cost measured 2026-10-31 (win-x64 32/41 s, mac-arm64 117/32 s, mac-x64 134/26 s against 12m25 / 8m01 / 7m50 native),OBS_CROSS_LINK=ONlinks a realobs.dll/obs.dylibper alias, PR CI runs theCross Compile 🧰job, and the report states signing and packaging stay native. The follow-up it asked for is filed and closed (#3059). Assigned to me; recommend closing and letting the remaining native-signing work stay on #2383/#2424.rust/tools/benchmarks/record.py, (2) a failing-test sketch for the GPU-encode stop/start race, (3) #742 is still open (the design it edits). Nothing for a worker until a Mac host records (1) and #742 lands.rust/tools/benchmarks/record.pyexists anddocs/rust-port/benchmarks/now carriesbaseline-2026-10-30-{linux-x86_64,macos-arm64,windows-x86_64}.json. No kernel work is claimed and the issue still asks for direction, so it stays a proposal; do not read the baselines as progress on the SIMD question.release-rust.yaml: Windows ccache set up andCCACHE_DIRpinned (§167-190, stats §255), macOS points Xcode's CAS atXCODE_CAS_PATH(§155), soldr S3 remote cache withsave-cache-remoteon master (§211-226). The missing half is exactly whatci.toml§109-114 records: phase A of #1801 (Decided 10-30) moved the six app-build families (includingwindows-ccache-x64/-arm64andxcode-x86_64/-arm64, 5.71 GiB, 9 of 42 restores) out of the Actions cache, and "the release-asset transport (#782) can bring them back outside this budget". So #782 is now the owner of that transport; it needs a rescope to that one deliverable, not a re-measure of the 10-14 numbers.ci/platform-ifdef-*are live). Recommend closing the tracker.#ifblocks), #1294. Open: #847 (A) and #700 (C). Close after those two./ready: the graph is stilltest/cmocka/build/test-graph.jsonwith 313 targets/tests,obs-buildvalidation::build_c_testsandtest/cmocka/native-test-graph.cmakeare its only consumers, and there is still no check that fails when a CMake-registered test is absent from the graph. Scope stands: evaluated CTest inventory per configuration, then the missing entries, then the two-variant export/run pass.deny(unsafe_code)) and 4 (ASan/UBSan/TSan/valgrind insanitizers.yaml) are done;clippy::undocumented_unsafe_blocksis still not enabled anywhere, and there is no Miri job (grep -rni miri .github ci rust/toolsis empty, no toolchain component). Rescope to those two, or file one sub-issue each.What this leaves as actually claimable work
Only five rows carry a task someone could take now, and none is a code change a worker can start blind:
undocumented_unsafe_blockslint) that would each make a clean sub-issue.Everything else is a tracker whose children are closed (#847, #1273), a parent waiting on its children (#700, #897), a follow-up whose deliverable has been narrowed by a decision (#782 → release-asset transport), a completed investigation (#697, #443), or a proposal awaiting numbers (#756). No duplicates were found among these rows.
Suggested next pass
Do not re-walk sections A–F again until the unclaimed count is back above ~60 (it is 41). When it is, start from the two trackers this pass found finished (#847, #1273), then the five items above; #143's two sub-issues and #1390's inventory are the only ones that need a worker rather than the maintainer.
wesleyhawes commentedon Nov 12, 2026
Claiming: #1436 — November 12 read-only backlog reconciliation pass.
1273c7ea21c9c6fae1ba7bb3ba7425429dfebc56(master push run 43221554857 passed).Prepared with AI assistance under Wesley's request to complete this contribution.
wesleyhawes commentedon Nov 12, 2026
Releasing: #1436 — the November 12 read-only reconciliation pass is complete; findings below. The recurring audit issue stays open.
What needs maintainer attention
Snapshot: 398 open issues, 371 unassigned before this claim; 15 open PRs at collection, 11 at the pre-publication recheck. Pinned master
1273c7ea2, with master push 43221554857 successful. This is a backlog audit, not a new code change or fresh runtime certification.18 issues are likely complete and worth reviewing for closure. Two are particularly misleading in the work queue: #4069 is still marked P0 although the pthread build fix is on master, and #4208 still advertises installer diagnostics that #4211 already supplies. The following table names the concrete evidence; a maintainer should close or retain each explicitly.
rust/obs-c-oracle/oracle/output_pause_queries.cpushes, undefines and restores the pthread macros around the probes. The PR records Windows-GNU RED/GREEN and eight native tests; current master push is green.rust/tools/sanitize/sanitize.pyreports installer exit code and the final 40 stderr lines, with failure/quiet-success coverage inci/test_sanitize_harness.py.test/cmocka/test_v4l2_busy_signal.cexecs--slow-holderbefore OBS startup and preserves the intended handler. The latest contributor already released the issue as resolved; PR reports five passing sanitizer/plain lanes.docs/ci-tooling.md; its follow-up #3059 is closed. Native signing/packaging are explicitly retained, not counted as cross-build completion.shared/file-updater/file-updater/file-updater.c.hcalls the completion callback on success/failure.timer-dispatch.c.hplus actual-library characterization; Rust follow-up #4557 also merged.For those code candidates I verified the reported PRs are merged into
master, their merge commits are ancestors of the pinned master, and the corresponding source/test artifacts remain present. Validation numbers above are the PR authors' recorded evidence, not tests rerun by this audit. Full native Windows/macOS or full-suite execution is not inferred from a focused local test or a green merge.Keep implementation and acceptance separate
Coordination and useful next work
clippy::undocumented_unsafe_blocksin the inspected workflow/tool configuration; split and triage those, instead of reclaiming the whole tracker. ci(sanitizers): about 150 tests report ThreadSanitizer races and nobody triages them #3369 needs a fresh nightly and one coordinated residual report group, not reuse of its Nov 8 counts. bug: image-source: an Image Slide Show plays a folder in filesystem order on Linux and macOS (upstream #10382) #3451 has a reported remaining slideshow-order bug, but its ordering/test scope and Ready state should be settled first.Coverage and verdict index
Collected every issue body and full comment pagination for the 371 initially unassigned issues; compared open PR references, 2,317 first-parent fork PR commits since 2026-10-08, current mentioned paths, source/test artifacts from 386 relevant merge commits, feature-ledger entries and similar open titles. Pre-fork history and bare cross-repository
owner/repo#Nnumbers were excluded from local-number matching. Exact-reference collection is not complete semantic discovery: absence of a match does not prove absence of work.94 rows received individual acceptance/coordination review. The remaining rows retain conservative scan dispositions and evidence leads; this is not a line-by-line acceptance audit or a runtime test of all 371 tasks. In particular, C does not mean its code is still missing. Only the 18 individually reviewed A rows above are proposed closure candidates. High-similarity examples such as #4691/#4692 and #4688/#4689 are deliberate extraction/port pairs; no duplicate closure is proposed from similarity alone.
Counts: A 18, B 2, C 208, D 93, E 28, F 22 (371 total).
All 371 issue dispositions and evidence leads
No issues were closed, relabeled or rescoped, and no code, workflows or other workers’ branches were changed. This report and release finish the accepted read-only audit pass; maintainers decide the resulting tracker changes. Prepared with AI assistance.