When buff suppression expires, only clients get ReapplyBuff() called to
restore visual and state effects (illusions, procs, silence, etc.).
Pets and NPCs had no equivalent handling, causing permanent loss of
illusions, weapon procs, and other active effects after suppression.
Add non-client effect restoration in the suppression expiry path of
BuffFadeBySlot() that handles illusions, silence, amnesia, and weapon
procs for pets, NPCs, and bots.
Fixes#32
When BuffFadeBySlot() is called on a suppressed slot (e.g., when
suppression timer expires), the effect removal loop accesses
spells[SPELL_SUPPRESSED] (index 0xFFFD) which is out-of-bounds.
This is unnecessary since effects were already removed when the buff
was first suppressed. Skip the loop entirely for SPELL_SUPPRESSED
slots to prevent OOB memory access and let execution proceed directly
to the restoration branch.
Fixes#29
BuffFadeNonPersistDeath(), BuffFadeBeneficial(), BuffFadeDetrimental(),
and BuffFadeByEffect() all use IsValidSpell() which returns false for
SPELL_SUPPRESSED (0xFFFD), causing suppressed buffs to survive death
and mass-dispels. Worse, if BuffFadeBySlot() were called on a
suppressed slot, it would restore the buff instead of clearing it.
Add explicit handling for SPELL_SUPPRESSED slots in each function:
check the suppressedid against the fade criteria and clear the slot
directly (set to SPELL_UNKNOWN) rather than calling BuffFadeBySlot().
Fixes#31
BuffProcess() calls DoBuffTic() on all valid-or-suppressed spells, but
DoBuffTic() accesses spells[buff.spellid] which for SPELL_SUPPRESSED
(0xFFFD) could read out-of-bounds memory. Add an explicit check to skip
DoBuffTic() for suppressed slots since they should only have their timer
decremented, not their effects ticked.
Fixes#30
When a player zones while having suppressed buffs (via the spell
suppression system), the suppression state was not saved to the
database. On zone-in, suppressed buffs were restored as normal active
buffs, causing non-persistent illusions to be incorrectly removed.
Changes:
- Add 'suppressed' column to character_buffs table (tinyint, default 0)
- SaveBuffs(): Save suppressed flag (1 when buff is in SPELL_SUPPRESSED state)
- LoadBuffs(): Restore buffs in suppressed state when suppressed=1,
setting spellid=SPELL_SUPPRESSED and populating suppressedid/
suppressedticsremaining from the saved spell_id and ticsremaining
- LoadBuffs(): Skip suppressed buffs in the illusion/charm removal loop
so they are not incorrectly stripped on zone-in
Fixes#33
Resolve CMake style conflict in common/CMakeLists.txt by adopting
the lowercase CMake command style used consistently throughout
the upstream codebase.
FindFirstFreeSlotThatFitsItem was returning 0 instead of INVALID_INDEX
when no free slot was found. This caused items to be placed in slot 0
(charm equipment slot) when inventory was full, overwriting equipped items.
Changes:
- Fix FindFirstFreeSlotThatFitsItem to return INVALID_INDEX when no slot found
- Add defensive check in PutItemInInventoryWithStacking to protect equipment
slots 0-22 from being targeted for item placement
* Start rewrite, add vcpkg
* Simple vcpkg manifest, will almost certainly need tweaking
* Remove cmake ext we wont be using anymore
* Update vcpkg to no longer be from 2022, update cmake lists (wip)
* Add finds to the toplevel cmakelists
* WIP, luabind and perlbind build. Common only partially builds.
* Fix common build.
* shared_memory compiles
* client files compile
* Tests and more cmake version updates
* World, had to swap out zlib-ng for now because it wasn't playing nicely along side the zlib install. May revisit.
* UCS compiles now too!
* queryserv and eqlaunch
* loginserver works
* Zone works but is messy, tomorrow futher cleanup!
* Cleanup main file
* remove old zlibng, remove perlwrap, remove hc
* More cleanup
* vcpkg baseline set for CI
* Remove pkg-config, it's the suggested way to use luajit with vcpkg but it causes issues with CI and might be a pain point for windows users
* Actually add file
* Set perlbind include dir
* Perl link got lost
* PERL_SET_INTERP causes an issue on newer versions of perl on windows because a symbol is not properly exported in their API, change the lines so it's basically what it used to be
* Remove static unix linking, we dont do automated released anymore and this was tightly coupled to that. Can explore this again if we decide to change that.
* Remove unused submodules, set cmake policy for boost
* Fix some cereal includes
* Improve some boilerplate, I'd still like to do better about getting linker stuff set.
* Going through and cleaning up the build.
* Fix world, separate out data_buckets.
* add fixes for other servers
* fix zone
* Fix client files, loginserver and tests
* Newer versions of libmariadb default to tls forced on, return to the default of not forcing that.
auto_login were breaking on linux builds
loginserver wasn't setting proper openssl compile flag
* Move set out of a giant cpp file include.
* Convert show
* convert find
* Add uuid to unix builds
* Remove some cpp includes.
* Restructure to remove more things.
* change db update manifest to header
change build yml
* Move world CLI include cpps to cmake.
* Move zone cli out of source and into cmake
* Sidecar stuff wont directly include cpp files now too.
* Fix uuid-dev missing on linux runner
* Reorg common cmake file
* Some cleanup
* Fix libsodium support (oops). Fix perl support (more oops)
* Change doc
---------
Co-authored-by: KimLS <KimLS@peqtgc.com>
* Initial plan
* Initial plan
* Enhance SummonItemIntoInventory to support stacking
Co-authored-by: Valorith <76063792+Valorith@users.noreply.github.com>
* Update .gitignore
* Revert "Update .gitignore"
This reverts commit 16159398d8a69c53a719a1d54d068bbe0fa5284c.
* Disable PCH for patch sources compiled with -O0
Disables precompiled headers for specific patch source files that are compiled with -O0 on UNIX. This avoids Clang errors caused by __OPTIMIZE__ macro state mismatches between the PCH and translation units.
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
