CAN_I_DAMAGEABLE not updating status bar #1259
Conversation
…efault container from map of vectors to a multimap, also trying to delete elements as little as possible, to avoid cpu usage, reallocations and tree rebalancing). Added #define MT_ENGINES: it doesn't make sense to keep unneeded mutex locks, at least until (if ever) we'll have a fully multithreaded engine. Some advancement in making ASAN work on Windows with Clang. Added .cmake-format
while using EF_FollowerList Sphereserver#1251
Reduced CSector member data types as needed, to save memory. Added a security check to CSFile::r_Write, at shutdown there might be a request to write to an invalid file descriptor (tested on Linux, tried to write on fd -1). Added macro SECTORSIZE_DEFAULT instead of using 64 as a magic number.
…dynamically allocating memory for it each time, pre-allocate and request cached CWorldSearch instances. Use CWorldSearch::GetInstance(). Added cdrc external library (reference counted smart pointers, std::shared_ptr is too slow) and new CSReferenceCount sphere library. Cdrc has been slightly modified in order to be used and compiled in a multi-file project. See custom folder. Added HAS_FLAGS_STRICT and HAS_FLAGS_ANY macros. Fixed macOS compilation error.
CWorldSearchHolder now uses CSReferenceCount instead of rc_ptr, which is more complex and slightly slower. Modified some ADDTOCALLSTACK calls for frequently called functions. Changed some rand val generations in CCharFight to use the fast algorithm. Inlined some small functions.
Fixed some Linux compilation warnings and errors. MSVC will compile by default with SSE2 CPU extensions enabled on x86_64. Updated CMakeDetectArch.cmake with 3rd party code.
…significantly reduced CPU usage and (slightly) reduced startup time. - Fixed: RAM usage always increasing during uptime. Caused by a couple of memory leaks, the greatest happening because the cached map blocks were not properly released. - Fixed: adopted a more conservative memory allocation policy for strings, resulting in reduced RAM usage. - Changed: I_MEMORY (ITEMID_MEMORY) items won't have anymore the ComponentProps Faction, ItemChar, Item, reducing their memory footprint.
…umber with a different base (seconds or tenths of seconds or milliseconds) in different contexts. Now they expect in scripts only values in seconds. Fixed: crash when moving a char in an area full of teleporters, or whichever items made the char move multiple times in the same tick. Fixed: buffer read past the end in CCacheableScriptFile first reading. Fixed: if a char had the COMM_CRYSTAL flag enabled, it called ::OnHear on every item in range, without checking if the item is actually capable of hearing. Further optimized distance checks. Separated the Github workflows into different files, in order to get a different badge for each build platform/type. Dismissed AppVeyor build. Added sphere_library/sfastmath.h.
Fixeid an invalid memory read when reading the last script file line.
Added more log messages for Linux x86 build and Coverity upload workflow.
By changing hits on item with can_i_damageable the status bar wouldnt get updated automatically. You had to reopen the status, change some value like more2 or target the item.
|
does that fix works? |
does for me. Everytime the hits gets changed from the custom script, statusbar updated. Before it didnt. If its correct fix im not sure. |
|
Confirmed bug and confirmed that Gladie's fix works. But you lack the fix as well for maxhits I dont know if it is how it should be fixed but works. Anyway, I found something weird. maybe is normal I dont know.
|
|
I wasnt sure if we need to have it update on maxhits. Maybe if there is like 50/100 hits and you decrease maxhits to 75 so the ratio of status would change too? |
same fix like for HITS, only missed the MAXHITS
You should split commits because there is a fix for follower slots. Well, I think maxhits could be set after the item is created and placed in the world so..., and custom scripts could need it too. For example, a repair system that reduces maxhits but sets hits to 100% I think u can force the ticks just giving a timer to the item. let me check. tested, Nopes u can't, Not sure if this is the way. I think so, but I confirm the fix works. |
I Made new branch for this fix but somehow it shows also the followerslots I dunno why 🤐 |
|
U need to do a branch from source-x instead from ur own fork. Then do the changes on that branch and pull. Remember u need to edit changelog as well. |
…ags. (Sphereserver#1237) Fixed duration of hallucination spell. Fixed wrong layer for potion delay. Fixed wrong sound id for changing monster sounds schema. Fixed Attacker_GetHighestThreat: return -1 would make the NPC threat lesser than the ATTACKER_THREAT_TOLDBYMASTER and make CChar::NPC_GetAttackMotivation not work correctly for attacking pets. Fixed NPC_Act_Follow use combat target even if the npc is not in combat.
- Fixed: ini CommandPrefix being overwritten by data from other ini settings. - Fixed: function call stack sometimes being subject to a race condition between two writing threads, corrupting it. - Fixed: lots of memory corruption errors leaving the server unstable when loading a broken save file and recovering by loading a backup save. - Fixed: CCacheableScriptFile sometimes not reading the last character. - Fixed: .clang_format. - Changed: RES_GET_INDEX macro to ResGetIndex function, allowing more integer type safety. Same for RES_GET_TYPE. - Changed: Renamed THREAD_* macros to MT_*. Split them in always active macros (MT_*) and macros active only when MT_ENGINE is set to 1 (MT_ENGINE_*). There are some situations where we still need a mutex even if not working with multithreaded script/networking engines (which are NOT complete). - Changed: Disable signal handlers under Linux if running with ASAN or a debugger. - Added: NOEXCEPT_DEBUG, it's simply the noexcept specifier, but it's ignored if running in a debug build (useful if we're using DEBUG_ASSERT inside a function which wouldn't otherwise throw an exception). - Fixed: some undefined behaviors, like signed to unsigned conversion and vice versa. - Added: additional checks for GCC and Clang ASAN builds.
And fixed non-windows compilation error.
|
Did you manage to remove the FOLLOWERSLOTS change from this PR? |
This reverts commit fa87433.
I think I managed to do it right now. |
cbnolok
force-pushed
the
dev
branch
3 times, most recently
from
0913018
to
7a9a90c
Compare
|
I'm sorry, could you fix the conflicts please? |
should be done |
|
Manually merged. |
Another attempt for the small fix. #1248 also together with this we should consider #1260
The status bar would not get updated when you lower hits on the item with the flag.
Only way it would get updated is: