RF455-127 RF 4.55: the 18 cash-shop armour exchange coupons can never be redeemed, their combine rows have no output

Game: RF 4.55 (Dragonborn), server v0.5.0, Windows build (stock gameserver.exe). Client: the supported Taiwan build, RF_Online.bin SHA-256 46528aac…2bbc558 (2025-07-16), Language=Taiwan. Suggested component: data (game data and scripts).

This is the same shape of problem as RF455-084Released item 3 (the premium reward coupons), which you fixed in v0.4.0 by filling in the recipes' outputs. Eighteen more coupons — these ones sold for cash points — still have it.

Summary

The 黃金黑 / 聖金黑 ("Golden Black" / "Saint Golden Black") armour exchange coupons are cash-shop items whose own description says to combine them at the Hero NPC for a 5-piece armour set. The combine always fails with "Item combination materials are incorrect".

The cause is in the data, not in the player's hands: each coupon has exactly one recipe row, and every one of that row's 24 output slots is -1. loadCombineRecipe discards a recipe with an empty output list, so CombineRecipeFor reports "not found" and the request is refused before the civil mask, the fee and the materials are even looked at.

They are the only coupons in the game that are both sold in the cash shop and impossible to redeem.

How to reproduce

1. Buy (or GM-grant) iyetw32, 聖金黑防具兌換券[阿克雷提亞發射器型] — bootyitem 7140, cash shop row 352 (cash353), 682 CP, m_bView 1 — on an Accretia character.
2. Talk to a Hero NPC (storelist trade type 26: 030F1 "Bo Hammer", 040F1 "Ziz Oadasha", 050F1 "Lothan the 3rd") and open the Item combiner.
3. Put the coupon in the centre slot. The window quotes the row's fee, "You need 0's money for combining fees".
4. Confirm → "Item combination materials are incorrect", and the server logs:

[2026-10-26T19:55:17] INFO combine refused {"session": 2, "err": 1, "reason": "unknown manual index"}

Only iyetw32 was tried live; the other 17 rows are identical in shape (section 3), so the same follows.

1. The recipe exists but produces nothing (read, gamedata_rf455.db)

iyetw32 is a material in exactly one row of the three combine tables — combinetable2 30145, m_strCode cocsa90521 (so combine index 60376 + 30145 = 90521, which matches the code):

combinetable2 row 30145  (combine index 90521)
  code=cocsa90521  civil=00001000  exist=1  fee=0
  material iyetw32   x1    聖金黑防具兌換券[阿克雷提亞發射器型]
  outputs  NONE

The row is otherwise complete and looks intentional: m_bCombineExist 1, m_nOutNum 1, m_dwEffectType 1, m_dwResultEffectMsgCode 3605, m_nOperationCount 1, m_strCivil 00001000 (Accretia, which matches the coupon's name). Only m_itmPdOutput … m_itmPdOutput_24 are all -1.

For comparison, the Armor Set Coupons that do work are built exactly the same way but with their outputs filled in:

combinetable2 row 10095  (combine index 70471)
  code=cocsa70471  civil=11111000  exist=1  fee=1000
  material iygco44   x1    Accretia Launcher Armor Set Coupon [Level 53]
  output   ihafb53   x1    weight 2000   Intense COS Head
  output   iuafb53   x1    weight 2000   Intense COS Core
  output   igafb53   x1    weight 2000   Intense COS Arm
  output   ilafb53   x1    weight 2000   Intense COS Leg
  output   isafb53   x1    weight 2000   Intense COS Landing

(The five weights sum to 10000, so the set is always granted.)

2. Why the server refuses (read, rf455-v0.5.0\server\windows\gameserver.exe)

• loadCombineRecipe 0x14051e120 picks the table: cmp ebx, 0xebd7 at 0x14051e160 — index < 60375 reads combinetable, index ≥ 0xEBD8 reads combinetable2 at index − 0xEBD8 (lea edx,[rbx-0xebd8] 0x14051e194).
• Its tail at 0x14051fcd3 is the gate: cmp qword [rax+0x50], 0; je (the material slice's length) then cmp qword [rax+0x68], 0; jne (the output slice's length). With either list empty it returns nil (0x14051fce2). (+0x48/+0x50 is the material slice — combineMatchMaterials 0x14088a4c0 loads it at 0x14088a5a4 and walks it with stride 0x20; +0x60/+0x68 is the output slice — combineDraw 0x14088bec0 loads both at 0x14088bff6.)
• CombineRecipeFor 0x14051df40 returns found = (pointer ≠ nil), setne bl at 0x14051e0c7.
• handleCombineExItem 0x140887ca0: test bl,bl; je 0x14088811f at 0x140887ee0 → combineRefuse 0x140887a60 with code 1 and the reason "unknown manual index" (0x140cdb6d4).
• The civil check (code 2, 0x1408880f8), the dalant check (3, 0x1408880d1) and combineMatchMaterials (0x23, 0x140888050) all sit after that branch, so nothing the player did is ever inspected. The client's "materials are incorrect" wording is misleading here.

The client cannot warn about it: DataTable\ItemCombine.edf carries each recipe's index, fee and civil mask but no outputs (outputs are server-side only), so the window opens and quotes the 0 fee whatever will happen next.

3. It is all 18 coupons

Every row below has m_bCombineExist 1, one material (the coupon, ×1), fee 0, and no output. The last column is the set each coupon's own name asks for; all 90 item codes exist in the item tables.

The coupon → set column is inferred, from three things that agree: the coupon's own name (race, class and the 黃金黑 / 聖金黑 grade word), the armour's name (iuafg65 is "Saint Golden Black Launcher Core", iuawg55 "Golden Black Eloi Core"), and the row's m_strCivil, which is read and already matches the race in the coupon's name. Everything else in this report is read from the data or the binary.

Cash shop rows are read from cashshop (m_strCsItemCode, m_nCsPrice, m_bView); the server prices orders from that same table (priceCashOrder 0x1408378e0, which logs "no cashshop row %d").

4. The same thing was fixed once already

RF455-084Released item 3 reported that the premium reward coupon iyevn10 combined into nothing. Comparing the three server versions we have:

So the fix for these 18 is the one you already applied there: fill the rows' outputs. Row 30145 is byte-for-byte unchanged across v0.3.1, v0.4.0 and v0.5.0.

Related, for the same pass

combinetable2 has 19,464 rows with no output and 18,030 of those still carry m_bCombineExist 1; combinetable has 126, all flagged. Among them, 110 take an iyetw* item as their first material, including:

• iyetw34–iyetw62, the "self-select vouchers" (55B/60B/65B/70B/75B weapon, 5-piece armour, shield, cloak);
• iyetw01–iyetw13, the Halloween recipes.

And nine set coupons named exactly like the working ones have no recipe row at all: iygco90–iygco98, the "… Armor Set Coupon [Level 50]" family (the Level 53/57/63/65 ones, iygco18–iygco53, all work).

Local workaround

None. The coupon cannot be redeemed, and it cannot be traded away either (m_bExchange 0); an NPC will take it (m_bSell 1) but pays nothing for it (m_nMoney 0). A GM can hand out the five armour pieces directly instead.