The text-proto parser (get_Item_LimitType_Value in ProtoReader.cpp)
matches LIMIT_TYPE columns against a positional table with the
"LIMIT_" prefix stripped ("TIMER_BASED_ON_WEAR", not
"LIMIT_TIMER_BASED_ON_WEAR" as previously set, which was rejected
outright and crash-looped the db process). This is purely a data-file
correction; the char_item.cpp/item.cpp logic from the previous commit
was already correct and untouched by the emergency revert.
The item_proto.txt CSV parser (Set_Proto_Item_Table) rejects this limit
type string outright ("Invalid value... col 16"), crash-looping the db
process on boot. The enum exists in item_length.h and the game-side
code fully supports it, but the text-format parser's string table
apparently never had it wired up. Reverting to LIMIT_NONE to restore
service; the wear-based timer fix in char_item.cpp/item.cpp needs a
different way to seed the item's LIMIT_TIMER_BASED_ON_WEAR configuration.
The seal previously toggled the mount via a manual AddAffect+socket
flag entirely inside UseItemEx, bypassing Equip()/Unequip() and the
item's LIMIT_TIMER_BASED_ON_WEAR machinery entirely - so with
item_proto set to LIMIT_NONE, the mount never expired at all.
Now: activating starts the item's own wear-based countdown (via
StartTimerBasedOnWearExpireEvent, reusing socket0 as remaining-seconds
exactly like the existing wear-timer convention), deactivating pauses
it (StopTimerBasedOnWearExpireEvent), and it only ticks down while the
seal is active - matching the intended "usage time" behavior. Moved
the seal's own active-flag from socket0 to socket2 to stop colliding
with that convention. item_proto now grants 2 days (172800s) per
activation. When the usage time fully runs out while riding, both the
mount and the seal item are removed (timer_based_on_wear_expire_event
now calls ClearMountAttributeAndAffect() for ride items before
destroying them, mirroring the existing Unequip() cleanup path).
Andreas Lierschaft