AVA2 new_switchbot ported as port_switchbot.{h,cpp}: SWITCHBOT item window (7),
CG/GC 171 and GG 31 with AVA2 subheaders and field order, pack(1) structs pinned
by static_assert, 0.1 s roll, P2P hand-over on warp, boot ALTER of
player.item.window. Switchers come from the inventory; current checks and
ItemLog are kept. Clients sending "mt20261016" or newer as the version
timestamp get the switchbot; older clients are left alone.
The v1 /mt_switchbot chat command, its lock verb and query capability bit 64
are removed (cap mask 1023 -> 959).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7.0 KiB
PORT attribute lock
AVA2 ENABLE_LOCK_ATTR, rebuilt for this server's item layout. Up to two of the five normal attribute slots keep their type and value when the item is rerolled. Source: src/server/game/src/port_attr_lock.{h,cpp}, CItem::ChangeAttribute / CopyAttributeTo (item_attribute.cpp), the cache fields in item.h/item.cpp, and the boot hook in main.cpp. The lock verb of the v1 mt_switchbot command was removed with the AVA2 switchbot port (FEATURE-SWITCHBOT.md). Locks are now set only by using 50348 (lock) / 50349 (unlock) on the item. UseItemEx passes the material to PortAttrLockUse, which reads the 1-based slot from the material's socket 0 and consumes one material on success.
Storage
AVA2 keeps up to two 1-based slot indices as attribute values at indices 7 and 8 of its 15-slot attribute array. This server has ITEM_ATTRIBUTE_MAX_NUM = 7 (5 normal + 2 rare) and ITEM_SOCKET_MAX_NUM = 3, which metin stones use. The item table does have socket3..socket5, but TPlayerItem and the client packets only carry 3 sockets, so using them would change both the db↔game and client wire formats. The item row is also rewritten with REPLACE INTO, so a column written behind the db cache would be lost.
The lock therefore lives in its own table, which no packet or db-cache structure carries:
CREATE TABLE IF NOT EXISTS mt_item_attr_lock (item_id INT UNSIGNED NOT NULL PRIMARY KEY, mask TINYINT UNSIGNED NOT NULL) ENGINE=MyISAM
- Each game core creates it at boot (
PortAttrLockBoot, auth cores skip it). No manual migration is needed. - Bit i of
maskis normal slot i (0..4). A mask of 0 deletes the row. PortAttrLockGetreads the mask once perCItemwith a synchronousDBManager::DirectQueryby primary key and caches it on the item (m_portAttrLock,m_portAttrLockLoaded). Exchanges on the same core keep the object and its cache. A warp or relog creates a new object, which reads the row again.PortAttrLockSetwrites synchronously, so the next core always sees the new value, and writes anATTR_LOCKitem log (hint = new mask).- A failed lookup logs a syserr, returns 0 and is retried next time. A failed write leaves the cache unchanged and returns the
dberror. CopyAttributeTo(refine, upgrade, over9 refine, quest item swaps) copies the lock to the new item ID together with the attributes.- Rows of destroyed items stay behind. They are harmless because item IDs are not reused.
Rules
-
Only the five normal slots can be locked; the rare slots 5/6 are not affected. The item must sit in the inventory, not be equipped or in an exchange, not be a costume, and have an attribute set. The slot must hold an attribute.
-
At most 2 lock groups, and at least one filled slot must stay unlocked. On items with an addon type (
sAddonType, weapons whose slots 0/1 are the skill / normal hit bonus pair), slots 0 and 1 lock and unlock together and count as one group. -
Locking is free. AVA2 consumes 50348/50349 per lock or unlock; those vnums do not exist in 203
item_proto. -
CItem::ChangeAttribute(scroll rerolls, the switchbot, the GM command):- take the effective mask: stored mask ∩ filled slots, with the addon pair kept whole;
- clear the normal slots;
- apply the addon only if the addon pair is not locked;
- put back the locked slots at their indices;
- reroll the remaining count with the original
PutAttribute. It writes to the first empty slot and skips types already present, so the gaps fill in order without duplicating a locked type.
With no lock, the effective mask is 0 and the sequence equals the original. The attribute count is kept, and the rare slots are untouched.
-
Adding an attribute (71085) and rare-attribute scrolls do not look at the lock.
Protocol
No dedicated packet. The v1 /mt_switchbot 1 lock CELL SLOT / MTSB1 lock lines were removed together with that command. What remains is the item-use path above. AVA2's UI for picking the slot and showing the lock in the tooltip is still open (metin2-client docs/AVA2-PARITY-PLAN.md).
Verification
tests/features/test_attr_lock.pycuts the productionCItem::ChangeAttributebody out of item_attribute.cpp and compiles it against a stub item. It checks that:- with no lock every slot changes;
- locked slots 1+3 are kept in place and the gaps are refilled;
- a locked type is not rolled again;
- a lock on an empty slot is ignored;
- a locked addon pair skips
ApplyAddon, while an unlocked one is rerolled; - the rare slot is untouched and the count is kept;
- there is no DB lookup when there are no attributes;
- every toggle, effective-mask and group-count rule holds.
- It also checks the boot hook, the Makefile, the cache initialisation and that
CopyAttributeTocopies the lock. tests/features/test_switchbot.pychecks that the v1 command (and with it thelockverb) is gone.- FreeBSD: full rebuild of game in
/root/metin2-features-dev(logattrlock-build.log;item.hchanged, so every .cpp was touched). - Live on 2026-10-15, channel-5 test core (13040) with the isolated binary, macOS client, [SA]Admin, armour 11209 (item 20000178, 5 attributes), scroll 71084:
- the first boot created the table;
- locking slot 1 → mask 2, then slot 3 → mask 10, then slot 0 →
error lock max; - 10 switchbot rolls (
limit, scrolls 143 → 133): (23,10) in slot 1 and (39,10) in slot 3 stayed the same on every roll, slots 0/2/4 changed every roll, and there were no duplicate types; - the row read
20000178 10; - a new client session showed
lock 19 11209 10(reloaded from the table); - unlocking → 8 → 0, and the row was deleted;
loghas fourATTR_LOCKrows (hints 2, 10, 8, 0).
- Not done: Android device test.
Production deployment
- 2026-10-15 16:08, approved by the user. All 203 live cores now run main
dc71593(gamesha256a8556bdc…, the exact binary the channel-5 core was tested with, built in/root/metin2-features-dev). The db binary is unchanged. 0 players were online. - Package
/root/deploy-features/20261015-attrlock(deploy.sh check|go|rollback,SHA256SUMS). The previousgame/db(boss deploy,400e44f) are in/root/backup-bin/20261015-160813. The 8 changed files were synced into/usr/metin2/src, and the replaced ones are insrc-before/. Becauseitem.hchanged, a laterrebuild.sh buildmust touch every game .cpp.sh deploy.sh rollbackrestores both; themt_item_attr_locktable can stay, since old binaries ignore it. - CH2–CH4 and the channel-5 core were started by hand after
start.sh 1, giving 14 game cores plus db and auth. - Live acceptance on CH1 (13000), same armour 11209 (item 20000178), scroll 71084:
- lock slot 1 → 2, slot 3 → 10, slot 0 →
error lock max; - 10 switchbot rolls (scrolls 133 → 123): (23,10) and (39,10) unchanged on every roll, the other slots changed;
- the row read
20000178 10; a new session showedlock 19 11209 10; - unlocking → 8 → 0 deleted the row, and four new
ATTR_LOCKlog rows were written; - no
PortAttrLocksyserr entries.
- lock slot 1 → 2, slot 3 → 10, slot 0 →