Upstream Mining Ledger

Upstream Mining Ledger

Claude-owned ledger (source-cited). Records the 2026-06-23 upstream-mining loop: a self-paced pass that mined the upstream parent Miksuu/a2waspwarfare for features genuinely missing from rayswaynl/a2waspwarfare and delivered the worth-doing slice as draft, human-merge-gated code PRs. It complements Upstream Miksuu commit intel (which catalogs upstream commit lineage) by recording the triage verdicts and PR outcomes of one bounded mining batch. All source claims here were verified against origin/master and the named upstream branch tips.

Goal & method

rayswaynl/a2waspwarfare is a heavily-diverged fork — roughly 557 commits ahead of its upstream parent Miksuu/a2waspwarfare. The goal of this loop was to find any upstream feature that is still absent from our master and is worth porting, then deliver each worth-doing candidate as a draft PR for human review (never an auto-merge).

Method (funnel):

1. Fetch all 490 branches from the upstream Miksuu remote.
2. Diff each against our master → 93 branches carry at least one unique commit not already absorbed into our fork.
3. Workflow triage of the 93 (type, value, risk, effort, present-on-master guess, OA-safety) → see triage_table.tsv.
4. Adversarial source verification of the promising rows — confirm each feature is actually missing on origin/master (not just renamed/absorbed), check OA 1.64 command safety, MP/JIP/locality/server-authority hazards, and whether the upstream diff even applies against our diverged tree → see verified_table.tsv and actionable_detail.txt.
5. Rank into a PR shortlist; open drafts for the genuinely-missing, portable, OA-safe candidates; flag the rest with owner/security notes.

Headline finding

Our fork has already absorbed nearly all high-value upstream work; the remaining surface is thin. The funnel collapsed hard:

Stage	Count
Upstream branches fetched	490
Branches with unique commits vs our master	93
Deep-verified — pass 1	18
Re-verified — pass 2	30
Net draft PRs opened	6
Flagged findings (no PR)	3

Most of the 93 unique-commit branches turned out to be debug/test scratch branches, stale merge tombstones (empty diff vs merge-base), version-snapshot bundles whose payloads are already on master, reverts, or A3-port experiments. Of the small set that were genuinely missing and worth shipping, several upstream implementations were broken as written and had to be reimplemented cleanly rather than cherry-picked.

Draft PRs opened

All six are DRAFT and human-merge-gated. None has had in-engine smoke yet — engine validation (OA 1.64 hosted/dedicated) is pending and is a merge precondition. Where the upstream diff did not apply against our diverged tree (the common case), the change was hand-reapplied / cleanly reimplemented rather than cherry-picked.

PR	Branch	Summary	Notes
#54	claude/upstream-at-naming-order	AT-soldier buy-menu display names + NLAW barracks tier	RU/TK AT → Rifleman (RPG-7 VL); BAF AT → AT Specialist (NLAW); NLAW gated to barracks L2 (Dragon parity, verified); NLAW reordered below SMAW. Pure data-array literals, OA-safe, lowest-risk. Source: UnitNamingAndOrderChanges (only ready row in pass 1).
#55	claude/upstream-blinking-mapicons	Blink the mounted player's own soldier marker	When player is gunner/commander of a vehicle whose LFTB flag is active, also push the player's soldier unit into BLINKING_UNITS_*. Client-local, OA-safe. Hand-ported into our diverged Client_BookkeepBlinkingIcons.sqf. Source: BlinkingMapIconsV2.
#56	claude/upstream-an2-climb-boost	AN-2 high-climb boost, capped below cruise	Adds the Valhalla climbing-assist low-gear action to An2_TK_EP1, with a deliberately chosen boost target so it doesn't become a velocity exploit. Rewritten against our assist-only LowGear model (upstream's _min=90/_max=250 was meaningless against our rewrite). Source: An-2FastLiftVehicleWithHalo (HALO half already shipped on master; only climb-boost was missing).
#57	claude/upstream-hq-repair-price	Show the next MHQ repair price in the action menu	Clean reimplementation of upstream's broken version. Upstream's localize calls referenced strings/keys with no %1 placeholders, so the price never rendered, and it dropped the cap-reached guard. This PR adds proper %1 placeholders to both stringtables, fixes the format calls, keeps the Repairs_Used exit guard, and applies to both Chernarus + Takistan. Source: HQRepairPriceVisibileOnTheActionMenu.
#58	claude/upstream-town-defense-diag-sync	Town-defense group-count diagnostics + RptTownDefenseAnalyzer tool	Group-saturation diagnostics for town-defense HC groups plus a PowerShell RPT analyzer, syncing a 3-behind diagnostics line. Tooling/observability, not gameplay logic. Source: Marty_stress_test_town_defenses.
#59	claude/upstream-barracks-4th-level	4th Barracks level + sweeping infantry tier rebalance	Adds a 4th barracks level and increments many infantry units' required level by +1 (raises the bar for HAT/AAR/Spotter etc.), plus MAAWS HEDP→HEAT and the AT renames. Needs balance sign-off. The two upstream [REVERT LATER] test hacks (research TIMES=1s, supply default 76800) were stripped; the new CZ_Soldier_AT_Wdl_ACR classname must be confirmed in-engine before merge. Source: BarracksInfantryRework.

Flagged findings (no PR — owner decisions / security)

These three are genuinely missing from master but are not safe to port as-is. Each needs an owner/design decision; one is a security regression.

(a) AdditionalHQBuying — security: would WORSEN a known HIGH-severity exploit

Porting this would worsen an already-flagged HIGH-severity client-authority hole, not just add a feature.

• It removes the only duplicate-rebuy guard (cashrepaired) on the WASP depot RECOVER-HQ action and adds no server-side gate.
• The price escalation is a client-local missionNamespace setVariable with no publicVariable/broadcast — so the "+50k each rebuy" never escalates globally and is trivially reset by reconnect (the feature's core promise is broken in MP).
• The existing depot rebuy is already client-authoritative: the client unilaterally debits funds, sends RequestMHQRepair (server spawns a fresh MHQ with zero validation), and broadcasts a town-supply reset that zeroes all own-side town supply. Removing the one-time lock turns a one-time client-side exploit into a repeatable own-side town-supply drain — a griefing/desync exploit.

Cross-reference the existing entries that already call this out: Deep-Review Findings (DR-55 / :1145), Server Authority Migration Map (:110), WASP Overlay (:69). Recommendation: server-side redesign only — move price escalation + rebuy-count + funds debit + town-supply reset into Server_MHQRepair.sqf / RequestMHQRepair.sqf with publicVariable broadcast and server-validated debit. Do not ship the raw branch.

(b) AirRework_GunDamageAdjustedToPlanes — inverted intent + dead ammo classnames

• The per-plane HandleDamage scaling uses _p=99, which yields _dam*0.01 — planes take only 1% of cannon damage (near gun-immune), the opposite of the stated "more lethal dogfights" intent. Owner must confirm the actual balance direction.
• 4 of 6 ammo classnames per plane (B_30mmA10_AP, B_20mm_AP, B_23mm_APHE, B_77x56_Ball) match zero real OA rounds — most case arms never fire. Real rounds are B_30mm_AA/B_23mm_AA/B_20mm_AA/B_25mm_HEI.
• Only the Chernarus diff is portable (the branch also patches tasmania2010 and a takistan folder that don't exist on our master). Mechanically trivial to port one file, but it is effectively a no-op until the design defects are fixed. Recommendation: template only — keep as a starting point, drop the dead folders, then have the owner set the intended _p direction and rewrite ammo cases to real OA classnames.

(c) Tournament_SideSpeakerWIP / TournamentFeaturesImplementation — caster/spectator system, needs-design

• A tournament caster/spectator system (civilian "side speaker" broadcast, full spectator/spectating module, custom chat-color mod, authorized spectator-slot IDs) that is genuinely absent from master.
• But it is unfinished WIP (134 / 132 commits) and, by design, the side-speaker broadcast leaks both-team intel to all clients (the tip commit itself admits the channel scope is unresolved).
• Recommendation: needs-design, not a port. If wanted, it requires a server-side audience-gating design so spectator/caster intel does not leak to active players — not a raw branch import.

Reusable artifacts

The full scan/triage/verify tables live under the mining lane at C:/Users/Chill/a2ww-upstream-mining/_mining/:

• triage_table.tsv — the 93 unique-commit branches with type/value/risk/effort/present/oa columns + one-line descriptions.
• verified_table.tsv — the deep-verified rows with ready verdicts (ready / needs-design / skip).
• actionable_detail.txt — per-candidate verdicts: FEATURE / MISSING? / SELF-CONTAINED / PORTABILITY / OA-SAFE+HAZARDS / TEST PLAN / RECOMMENDATION, with path:line source citations.

These are the audit trail for why each branch did or did not become a PR, and are the starting point if the loop is rerun against a fresh upstream fetch.

Pass 3 — Trello board (Miksuu's Warfare)

A second, independent mining pass that does not look at upstream branches at all — it mines the Miksuu's Warfare Trello board (the community's feature/bug backlog) for forward work worth doing in rayswaynl/a2waspwarfare, and delivers the worth-doing slice as grouped draft, human-merge-gated PRs. Passes 1–2 (above) mined the upstream parent's git branches; this pass mines the idea backlog.

Source & funnel

• Board: Miksuu's Warfare — https://trello.com/b/Si4okJLd/miksuus-warfare. Exported via curl -L https://trello.com/b/Si4okJLd.json (public board JSON).
• Raw size: 104 lists / 712 open cards. Most lists are version changelogs (work already shipped in past releases) or the Low Priority (Won't) bucket — neither is forward work.
• Tier-1 forward cards (118): the cards in the actionable lists — Must, Should, Next, Needs-triage, Suggested, Draft, plus the perf and cleaning lists.

Stage	Count
Open cards on the board	712
Tier-1 forward cards (Must/Should/Next/Needs-triage/Suggested/Draft/perf/cleaning)	118
Deep-verified against origin/master source	22
Ready → grouped draft PRs	8 cards / 6 PRs
Needs-design (owner/vote/art decision, no PR)	8
Skip (false premise / not worth it)	6

The same discipline as passes 1–2 applies: each "ready" card was verified against origin/master source (feature actually missing, OA 1.64-safe, no MP/JIP/locality/server-authority regression) before a PR was opened. All PRs are DRAFT and human-merge-gated; in-engine smoke is pending and is a merge precondition.

6 grouped draft PRs opened (8 cards)

Cards are grouped where they share a file/subsystem so review stays single-purpose.

PR	Summary	Trello card(s)
#60	Show vehicle ammo % in the service menu	#99
#61	Action to toggle automatic IR smoke off/on	#38
#62	Flashing warning when FAB-250 / Mk82 is selected above the bomb altitude limit (a "Must")	#106
#63	Anti-spam guards for building (#71) and vehicle (#74) repair — two exploit-fixes	#71, #74
#64	Block engineer salvage when a friendly salvage truck is in range	#15
#65	AAR upgrade enhancements — tiered detection height (#65) + new-contact warning (#66)	#65, #66

8 needs-design (owner / vote / art decisions — no PR)

Each is a real card but blocked on a decision that is not a coding call:

• #105 paratroop kill reward — root cause was never pinned upstream; needs source-level diagnosis before a fix can be trusted.
• #26 first-blood bonus — the reward numbers are a vote decision and the SFX is an art decision.
• #27 OPFOR Hind re-tier — balance/owner call on which tier the Hind belongs in.
• #29 town-depot supply trucks @10x — economy-tuning owner decision.
• #41 radiation damages factories — scope/design owner decision.
• #111 free supply truck at spawn — economy/owner decision.
• #113 all-arty-to-gunner key — control-scheme owner decision.
• #91 IR-smoke "cooldown ready" cue — false premise: there is no player-facing IR-smoke cooldown in current source for a cue to track, so the card as written has nothing to bind to.

~34 deferred small codeable wins (later pass)

Genuine, small, codeable wins held for a later batch to keep this pass focused — examples: #87 airlift teleporting onto HQ, #96 rearm-all-AI-at-cannon, #85 max-cities visual bug, #84 arty reload-ready sound, #76 dead WF-menu HUD button, an audio-cue cluster, #98 Linebacker/Tunguska markers.

Tier-2 — not yet triaged

The large lower-priority lists are untouched this pass: the Could list (149), the Backlog list (71), and the voting list. They are the starting point if the Trello pass is rerun.

Cross-pass total

Across all three passes this loop has opened 12 draft PRs — #54–#65 — all DRAFT and human-merge-gated, all awaiting OA 1.64 in-engine smoke before any merge.

Pass 4 — deep backlog (deferred wins + needs-design + Tier-2)

The follow-through batch that drained the backlog Passes 1–3 had parked. It completed Ray's a/b/c asks: deep-verified the ~34 deferred small codeable wins held back in Pass 3, re-examined the needs-design cards (the Warfare a + c items), and triaged the ~220-card Tier-2 long tail (the Could (149) + Backlog (71) lists). The verified, genuinely-ready wins were built into 10 more grouped draft PRs. As with every prior pass, all PRs are DRAFT and human-merge-gated; OA 1.64 in-engine smoke is pending and is a merge precondition.

10 new draft PRs

Each is grouped so review stays single-purpose; the source card(s) are listed for traceability.

PR	What	Source card(s)
#66	Bypass the anti-TK satchel & building scripts while in debug mode	#2, #3
#67	Audio cues: paradrop-drop sound + IR-smoke-ready sound	#90, #91
#68	Base-integrity bugfixes: airlift drop-on-HQ + nuke vs in-progress construction	#87, #97
#69	Fix gear/ammo display for special AT soldier units in the buy menu	—
#70	EASA aircraft loadout role categories (AA / AG / multirole) — completed half-wired code	#59
#71	Higher bounty for killing players on a killstreak (server-authoritative, tunable coefs)	—
#72	Restrict building static defenses near base when enemies are near (anti tank-trap, tunable consts)	#104
#73	One-click crew-all-artillery button	#113
#74	Show the player's artillery cooldown on the RHUD	#219
#75	Clearer info text: upgrade result value + new-HQ cost in the transfer menu	#107, #204

Known caveat (PR #68)

PR #68 surfaced a pre-existing Takistan bug: Zeta_Hook.sqf calls Zeta_Unhook without the [_vehicle] args array that the Chernarus B66 fix added. The airlift drop-on-HQ fix depends on that arg being passed, so on Takistan the fix needs the args folding-in (or a separate Takistan fix). Flagged in the PR; not silently shipped.

~37 needs-design items (no PR — owner / vote / art decisions)

These are real items but blocked on a decision that is not a coding call (a balance/economy owner decision, a community vote, or an art/SFX asset). They are catalogued in the mining artifacts; the notable ones:

• #105 — paratroop kill reward fix
• #26 — first-blood bonus
• #27 — OPFOR Hind re-tier
• #29 — town-depot supply trucks @10×
• #41 — radiation damages factories
• #63 — captured-town vehicle reward
• #64 — redeployment "barracks-on-wheels" truck
• #215 — sell-vehicles feature
• #120 — anti-illegal-loadout sanitiser
• #35 — static-count-by-barracks-level

Deferred / skip backlog

The remaining lower-value Tier-2 cards plus the big Could / Backlog long tail are recorded but not actioned — they fell below the worth-doing bar for this loop and are the starting point if the backlog is re-mined. One small flagged bug noted along the way: Core_MVD.sqf's [Done] log line mislabels itself "Core_RU".

Cross-pass total

Across all four passes this loop has opened 22 draft PRs — #54–#75 — all DRAFT and human-merge-gated, all awaiting OA 1.64 in-engine smoke before any merge. The initial three-source sweep (upstream branches + Trello Tier-1 + Tier-2 long tail) is now complete; the loop from here is delta-only (new upstream commits / new board cards as they appear).

Pass 5 — deferred Tier-2 tail (closeout)

The closeout batch. Pass 4 deep-verified the Tier-2 long tail but capped out before finishing every codeable+missing candidate; 42 were triaged worth-a-deep-look but deferred. This pass deep-verified all 42 of that deferred tail, shipped the genuinely-ready slice as grouped draft PRs, and confirmed there is no remaining un-swept mineable surface. As with every prior pass, all PRs are DRAFT and human-merge-gated; OA 1.64 in-engine smoke is pending and is a merge precondition.

Funnel

Stage	Count
Deferred Tier-2 candidates (codeable+missing, capped out of Pass 4)	42
Deep-verified against origin/master source	42
Ready → grouped draft PRs	7 cards / 4 PRs
Needs-design (owner/vote/art decision, no PR)	22

4 grouped draft PRs opened (7 cards)

Cards are grouped where they share a file/subsystem so review stays single-purpose; the source card(s) are listed for traceability.

PR	What	Source card(s)
#76	Auto-deploy bipod when going prone with a bipod MG	#210
#77	Show upgrade icons in the upgrade menu	#164
#78	Map radius circles: HQ build area + ambulance redeploy range	#80, #76
#79	Artillery UX: per-type range in menu, gun count in the "artillery called" message, and fix arty marker leaking when the caller disconnects mid-barrage	#115, #116, #171

22 needs-design items (no PR — owner / vote / art decisions)

These are real items but blocked on a decision that is not a coding call (a balance/economy owner decision, a community vote, or an art/SFX asset). They are catalogued in the mining artifacts; the notable ones:

• #216 — Federal Reserve passive-income endgame building
• #143 — ICBM impact countdown shown to both teams
• #178 — automatic minefield markers
• #174 — more-frequent house-wreck cleanup (perf)
• #165 — LoadoutManager data-driven missile-list refactor

Coverage confirmation (surface is exhausted)

• Miksuu GitHub Issues = 0 — the community runs its backlog on Trello, not Issues, so there is no second issue tracker to mine.
• Forks = 1 (ours) — there are no hidden contributor forks carrying un-mined work.
• The only remaining un-swept Trello surface is low-yield: the Low Priority (Won't) list (53 cards) plus 202 archived cards — both explicitly out of the worth-doing bar.

Final cross-pass total

Across all five passes this loop has opened 26 draft PRs — #54–#79 — all DRAFT and human-merge-gated, all awaiting OA 1.64 in-engine smoke before any merge. The full mineable surface — 490 upstream branches + Trello Tier-1 + Tier-2 (including the deferred tail) — is now EXHAUSTED; the loop from here is delta-only (new upstream commits / new board cards as they appear).

Round 6 — last surface swept (closeout, exhaustion confirmed)

The final sweep. Pass 5 confirmed every high-value surface was covered but left two explicitly-low-yield Trello buckets un-mined: the Low Priority (Won't) list (53 cards) and the archived-from-planning cards (110). This round swept that last surface — 163 cards — to formally close the mining loop. (The first run was rate-limited on 13/15 chunks — server throttle from too many back-to-back workflows — then resumed to completion.)

Result

• The archived-from-planning cards were all already-shipped (archived = done), as expected; they yielded no new work.
• Only the Low Priority (Won't) list produced genuine wins — 2, both landed in one PR:

PR	What	Source card(s)
#80	Commander tie-vote now picks a random tied player instead of silently falling to the AI commander (card "Tie vote → random commander"); and block calling paratroopers when already over the player AI unit cap (reuses the BuyUnits _realSize formula; server backstop; tunable WFBE_C_PARA_RESPECT_CAP)	"Tie vote → random commander"

Exhaustion confirmed

Every mineable surface is now covered — all 490 upstream branches, all Trello cards (Tier-1 + Tier-2 + deferred + Won't + archived), GitHub Issues (0), and forks (only ours). There is no further surface to mine; the loop is delta-only (new upstream commits / new Trello cards) from here.

FINAL cross-pass total

27 draft PRs open — #54–#80 — all DRAFT / human-merge-gated. Remaining work is human: review/merge the PRs and decide the ~61 needs-design items catalogued across the passes above.

2026-07-03 BE 2073 AICOM Fork-Delta Addendum

• Upstream BE 2073 AICOM fork-delta classification records the strict-path comparison between refs/remotes/upstream/v24042025 and Build84/cmdcon36 for Server/AI/Commander.
• The strict path has no upstream files and 18 WASP-added Build84 files (8157 insertions(+)), so future AICOM mining should port upstream concepts into WASP modules instead of looking for direct file cherry-picks.

Continue Reading

• Upstream Miksuu commit intel
• Upstream BE 2073 AICOM fork-delta classification
• Abandoned feature revival review
• Feature status register