Known bugs in JASS

[Strilanc]

Why isn't the TriggerDestroy bug in here? People apparently had a really hard time thanks to that one. I hit into it myself, a little bit.

JASS:

//Place in trigger called TriggerDestroy Bug and type "-bug"

///Exploits the bug
function CauseBugHelper takes nothing returns nothing
    call BJDebugMsg("Running Test")
    call DestroyTrigger(GetTriggeringTrigger())
    call TriggerSleepAction(0)
    call TriggerSleepAction(0) //both waits are needed
    if CreateGroup() == CreateGroup() then
        call BJDebugMsg("|cFFFF0000Bug Caught|r")
    else
        call BJDebugMsg("|cFF00FF00No Bug Caught|r")
    endif
endfunction

///Creates a trigger to bug, then calls it
///Note that calling the created trigger normally (timers or TriggerExecute) doesn't bug out
function CauseBug takes nothing returns nothing
    local trigger t = CreateTrigger()
    call TriggerAddAction(t, function CauseBugHelper)
    call TriggerRegisterAnyUnitEventBJ(t, EVENT_PLAYER_UNIT_DEATH) //indirect call by death
    call UnitApplyTimedLife(CreateUnit(Player(0), 'hfoo', 0, 0, 0), 'BTLF', 0.01)
    set t = null
    call BJDebugMsg("Setup Complete")
endfunction

function InitTrig_TriggerDestroy_Bug takes nothing returns nothing
    local trigger t = CreateTrigger()
    call TriggerAddAction(t, function CauseBug)
    call TriggerRegisterPlayerChatEvent(t, Player(0), "-bug", true)
endfunction

*edit* knocked off a bunch of lines and removed the ExecuteFunc call, since it's not needed.

[Pyrogasm]

Maybe you should explain this bug. Is it the handle-stack corruption one?

Edit: I have a bug of my own to add:

The function I2Code (a subset of the H2I-return bug stuff) does not work properly and may return erroneous things; instead, you need to add a helper function that converts the code to a real code to avoid bugs:

JASS:

function H2I takes handle h returns integer //Just fine when used with "code"s
    return h
    return 0
endfunction

function I2Code_Helper takes code c returns code
    return c
endfunction

function I2Code_Helper2 takes integer i returns code
    return i
    return null
endfunction

function I2Code takes integer i returns code
    return I2Code_Helper(I2Code_Helper2(i))
endfunction

Or at least I think that's right...

[Strilanc]

Originally Posted by Pyrogasm

Maybe you should explain this bug. Is it the handle-stack corruption one?

Yes that's the one.

[chobibo]

Using TriggerSleepAction inside timer function callbacks halts the functions execution.

[gekko]

Also found a weird one:

UnitRemoveAbility seems to unset GetSpellAbilityUnit().

I tried this:

JASS:

call UnitRemoveAbility(GetSpellAbilityUnit(),'A001')
call UnitAddAbility(GetSpellAbilityUnit(),'A002')

'A002' was never added. I tried several things, but then, after saving GetSpellAbilityUnit() in a local variable all worked fine! I haven't tested further, what GetSpellAbilityUnit() points to after a call of UnitRemoveAbility, but it was not the same unit as before.

[Vexorian]

Are you sure there weren't any waits involved?

[gekko]

Originally Posted by Vexorian

Are you sure there weren't any waits involved?

yes, these two lines are from my code. And it was unusable like that. Maybe the bug only appears in certain situations, but I was not able to add the second ability before I saved GetSpellAbilityUnit() in a local var and used this one on the second call.

Edit: The attached map shows the bug: A000 should be removed and A001 should be added, but A001 aint there after clicking A000!

[Troll-Brain]

UnitId2String always returns null when you load a saved game, same for UnitId2String and OrderId2String, and AbilityId2String doesn't work at all.

Originally Posted by common.j

// Not currently working correctly...
constant native AbilityId takes string abilityIdString returns integer
constant native AbilityId2String takes integer abilityId returns string

Added a demo map.

Note that i display the strings only for point orders and train orders, but it is the same with all other immediate orders and target orders, feel free to test it yourself.

[PurplePoot]

A little SubString bug that gave me hell.

SubStrings have some sort of bounds safety, as is well known. For example, SubString("test",0,9999)=="test" is true.

However, it seems their lower bound is a little stranger. Going "below" their start index can have a number of consequences:

JASS:

call SubString(<any string>,-4,0) == "Nmo"//Note that there is a zero-width space between N and m.
//And derivatives thereof, for example -3,0 == "mo" and -3,-1 == "m"
call SubString(<any string>,<-5 or lower>,<any number>) == null//Rather than just giving you the start of the string to the number you specified

EDIT: Apparently the N is not actually always an N—it's script specific (it seems to have to do with the number of globals or functions or something, but it's hard to tell).

[PurplePoot]

Yay for bugs.

GroupTargetOrderById does not work for 851971 (smart) on allied units. (Solution: issue 851990 (patrol) instead)

[Deaod]

Heres another one:

Assigning a new value to an image variable currently holding the invalid image (handle ID of -1) crashes the game.

[Chaos-Mapper]

Update, the first page pls with Return Bug or note that it's now in actuell warcraft version unable to use!

Thanks!

[weaaddar]

Create lightening native can return a null handle. Since this null handle can be manipulated, you have to go and attach like a boolean to determine if its created. Really annoying.

The fogstate exploit allows you to typecast from integers -> handles.

[Pyrogasm]

Isn't that what CreateLightningEx was supposed to solve, weaaddar?