PJass updates

[LeP]

@LeP is PJASS updated to 1.29?

As i mentioned i have problems with the newest patch.
I can't really test what valid jass is; for example if return null is still bugged or what kind of type-casting is allowed now. That being said, pjass doesn't really need to be adapted to the new patch (modulo the modulo operator) since it always gets passed the common.j and Blizzard.j from somewhere. So the new natives and such should just work, if your World Editor of choice passes the right files.

That being said, i have a 1.29 branch in my git but i can only modify it based on heresay at the moment. I will attach it but this is pjass for 1.29 ptr so it might no be 100% correct anymore.

And i just hope that blizzard will release quick patch.

 

[TriggerHappy]

As i mentioned i have problems with the newest patch.
I can't really test what valid jass is; for example if return null is still bugged or what kind of type-casting is allowed now. That being said, pjass doesn't really need to be adapted to the new patch (modulo the modulo operator) since it always gets passed the common.j and Blizzard.j from somewhere. So the new natives and such should just work, if your World Editor of choice passes the right files.

That being said, i have a 1.29 branch in my git but i can only modify it based on heresay at the moment. I will attach it but this is pjass for 1.29 ptr so it might no be 100% correct anymore.

And i just hope that blizzard will release quick patch.

Yeah I think it just needs support for the modulo operator. The return null bug has been fixed.

Typecasting no longer works but shadowing still does so you shouldn't need to update much.

 

[LeP]

Yeah I think it just needs support for the modulo operator. The return null bug has been fixed.

Typecasting no longer works but shadowing still does so you shouldn't need to update much.

The return null stuff for example is just a git revert away. The typing rules are a bit more delicate and i would like to test them myself. Otherwise the best way is to send code snippets which either should compile but dont or which shouldnt compile but do. That should atleast be sufficient for the short run.

 

[zachhandley]

@LeP thank you by the way for your work. You, Vexorian, and everyone else who has created the tools that allow us Map Developers to just learn ZINC/vJASS and just use it is all because of you guys.

 

[pyf]

As i mentioned i have problems with the newest patch. [...]

I therefore understand that this post was unfortunately of no help?

 

[LeP]

I therefore understand that this post was unfortunately of no help?

unfortunatley not. I ofcourse tried your suggestion and redownloaded directx. and 1.29.1 also didn't fix my problems. maybe wex works around it? i haven't tried it yet. i have reported my issues, now it's up to blizzard to fix them.

 

[pyf]

This looks like the description of your issue:
https://us.battle.net/forums/en/bnet/topic/20763256231

 

[LeP]

@Wareditor requested to report the line where a block was opened when the corresponding endblock is missing. Consider this beta though.

function f takes nothing returns nothing
    local integer x
    local integer y
    if true then
        set x = 4
        loop
        exitwhen true
      set y = 3
   set y = 8
endfunction

$ pjass test.j
test.j:11: Missing endloop for block opened in line 7.
test.j:11: Missing endif for block opened in line 5.
test.j failed with 2 errors
Parse failed: 2 errors total

 

[Wareditor]

Thank you for the update! Gonna test it soon
Can you make it so next update has a max recursion limit of 1000000 so I don't have to recompile it each time? If you don't want to do it just for one person, that's okay.

EDIT: my pjass crash when trying to compile my map. The map has not syntax error as my previous pjass didn't find any mistake and the map compiled fine. Funny enough when I try to compile this same map but this time with a missing endblock error it doesn't crash but report the error like it used to (without actually giving the line of the start of the block). @LeP

 

[LeP]

Thank you for the update! Gonna test it soon
Can you make it so next update has a max recursion limit of 1000000 so I don't have to recompile it each time? If you don't want to do it just for one person, that's okay.

EDIT: my pjass crash when trying to compile my map. The map has not syntax error as my previous pjass didn't find any mistake and the map compiled fine. Funny enough when I try to compile this same map but this time with a missing endblock error it doesn't crash but report the error like it used to (without actually giving the line of the start of the block). @LeP

eh. i guess i would need the map/war3map.j. or maybe the core dump. or come to irc

 

[LeP]

Fixed some really silly bug thanks to @Wareditor. Also enabled some optimisation in cross compiling (my gcc for cross-compiling is rather old... well what can you do)
Also if more people reach that stack limit error pls say so and i will increase it again.

 

[LeP]

Added modulo operator and merged the above release aswell.

 

[PurgeandFire]

wait. jass supports a modulo operator? o.o

or does pjass translate it somehow? it looks like it only does type-checking

 

[Illidan(Evil)X]

set i = 1000 % 8 is valid Jass code since patch 1.29.

 

[PurgeandFire]

set i = 1000 % 8 is valid Jass code since patch 1.29.

Awesome! I didn't know they added that.

 

[Prometheus3375]

Added modulo operator and merged the above release aswell.

Well, you forgot to update link in the 1st post.

And it's still passes next mistake.

globals
    boolean B true
endglobals

 

[LeP]

And it's still passes next mistake.

globals
    boolean B true
endglobals

Sorry can't replicate this one:

$ ./pjass -v
./pjass version git-ea093b2
$ cat tests/should-fail/vardecl-missing-equal.j
globals
    boolean B true
endglobals
$ ./pjass tests/should-fail/vardecl-missing-equal.j
tests/should-fail/vardecl-missing-equal.j:2: syntax error
tests/should-fail/vardecl-missing-equal.j failed with 1 error
Parse failed: 1 error total

 

[Prometheus3375]

Hmmm, I created new map and write that piece of code. No mistake was found when I clicked "Save map".

Added modulo operator and merged the above release aswell.

Used pjass.exe from that post.

 

[LeP]

Hmmm, I created new map and write that piece of code. No mistake was found when I clicked "Save map".

Used pjass.exe from that post.

That's on jasshelper. Jasshelper translates boolean B true to boolean B.

 

[Prometheus3375]

I noticed that the latest pjass has problems with encoding.

No such bug with this version. Tested it because I had it on my computer.
Changing encoding in PowerShell with chcp command does not help.

 

[LeP]

I noticed that the latest pjass has problems with encoding.

No such bug with this version. Tested it because I had it on my computer.
Changing encoding in PowerShell with chcp command does not help.

Having a script to reproduce it myself would help. Minimal example or whole mapscript is fine.
PM is also fine, if you dont want it to be public.

 

[LeP]

Thanks to @Prometheus3375 i have fixed an off-by-one error in the error message construction.

 

[DracoL1ch]

globals
trigger TriggerGoldPerSecond
endglobals

//somewhere else
call DisableTrigger(TriggerGoldPerSecond)

Would be great if Pjass could analyze that variable never being used withing "set" statement therefore using it will crash. Currently it only does it for local scopes

 

[LeP]

globals
trigger TriggerGoldPerSecond
endglobals

//somewhere else
call DisableTrigger(TriggerGoldPerSecond)

Would be great if Pjass could analyze that variable never being used withing "set" statement therefore using it will crash. Currently it only does it for local scopes

Sure that would be nice but it's also far from trivial. Even when you don't consider pjass' way of operation.
Also at one point one would be better served with a standalone jass linter or the like instead of stuffing everything into pjass.

 

[DracoL1ch]

hm, I thought it's the same as finding non-init'd locals, just bigger scope. W/e, just saying

 

[LeP]

hm, I thought it's the same as finding non-init'd locals, just bigger scope. W/e, just saying

Thing with locals is that you can see the whole control flow. I would assume with globals it would give false positives due to interprocedural nature of globals. I think the local check never gives false positives.
I could use the same code for globals and see what happens.

 

[LeP]

The uninitialized var check has been expanded. Now it also checks for uninitialized globals in the globals block by default.
Example:

globals
    integer a
    integer b = a + 3
endglobals

$ ./pjass tests/should-fail/uninit-globals-in-globals.j
tests/should-fail/uninit-globals-in-globals.j:3: Variable a is uninitalized
tests/should-fail/uninit-globals-in-globals.j failed with 1 error
Parse failed: 1 error total

You can also now use the same check there already is for locals on globals with the flag +checkglobalsinit
It is not default since it might be reporting false positives such as this

$ ./pjass +checkglobalsinit tests/common.j tests/Blizzard.j
Parse successful:     2418 lines: tests/common.j
tests/Blizzard.j:19891: Variable bj_stockPickedItemType might be uninitalized
tests/Blizzard.j failed with 1 error
Parse failed: 1 error total

 

[DracoL1ch]

thank you!

 

[LeP]

Just to bring this in sync with git.

• Various performance improvments
• Better error message for non-integer array indices
  

  globals
          integer array foo
          integer bla = foo[3.1]
  endglobals
  function xx takes nothing returns integer
          local integer array asdf
          return asdf["foo"]
  endfunction

  $ pjass test.j
  test.j:3: Cannot convert real to integer as index for array foo
  test.j:7: Cannot convert string to integer as index for array asdf
  test.j failed with 2 errors
  Parse failed: 2 errors total

 

[Prometheus3375]

Is this report okay? I'm using windows-1251 encoding and latest pjass.exe.

Spoiler: Screenshot

 

[LeP]

Is this report okay? I'm using windows-1251 encoding and latest pjass.exe.

Spoiler: Screenshot

Nope that's a jasshelper error message. Pjass would simply report "Syntax Error".

 

[Prometheus3375]

Why _SomeCharSequence is invalid identifier? Used for name of globals.
In other words, must not names in JASS2 start with _ or is it pjass mistake?

 

[LeP]

Why _SomeCharSequence is invalid identifier? Used for name of globals.
In other words, must not names in JASS2 start with _ or is it pjass mistake?

Indeed, names in jass have to start with a letter and have to end in either a letter or a number. inbetween numbers, letters and underscores are allowed.

 

[LeP]

Small update triggered by this. But some other stuff added aswell.
Here's the shortlog from the last version released here.

Jesse Rogers (1):
    Support Apple by defining _aligned_malloc
lep (8):
    Merge pull request #1 from theQuazz/feature/support-apple
    Specify minimum bison version in readme
    Report natives after function defintions
    Merge branch 'master' of ssh://192.168.1.5:52/home/git/pjass
    More newline stuff
    Added reserved keyword 'alias'
    Better error message for `type` as a name
    Fix release target

 

[Ralle]

Open!

 

[LeP]

Added a new option +checkstringhash which looks for calls to StringHash with a literal string as its argument. If pjass finds two calls with different strings which would hash to the same integer pjass will report an error. Do note though that you have to enable that option yourself; it's not on by default.
To do this i use the code i have reversed years ago. But afaik there have been changes to the StringHash native; it should still work for plain old ASCII strings but for multibyte characters my reversed code produces different hashes. If anyone wants to dump the new asm i will happily reverse engineer it (or maybe Blizzard sees their mistake and reverses the change).
While we're at it: some patches ago Blizzard made the backward incompatible change of giving alias a keyword. From what i know it never did anything so maybe remove it again? Or atleast document such things.

I've also taken this opportunity to bring the Github page up-to date.

 

[Aniki]

pjass does not report an error when the operand of 'function' is a native:

native foo takes nothing returns nothing
globals
    code c = function foo // no error
endglobals

Tested with "pjass version git-814075e".
I think this crashes the game at runtime.

 

[DracoL1ch]

well, as long as you can launch the map with that code, it's not really an issue

 

[LeP]

pjass does not report an error when the operand of 'function' is a native:

native foo takes nothing returns nothing
globals
    code c = function foo // no error
endglobals

Tested with "pjass version git-814075e".
I think this crashes the game at runtime.

Checking that is easy. If you make sure that this actually crashes ingame i will gladly add this as a check. For testing purpose you can use native TriggerSyncStart takes nothing returns nothing.

 

[Aniki]

Tested on 1.31 and it crashes when trying to execute the instruction that implements the 'function' operation:

    local code c = function ClearSelection

I suppose instead of giving an error (which it doesn't) the compiler could've generated an anonymous stub like:

function ClearSelection* takes nothing retuns nothing
    call ClearSelection()
endfunction

or something, but that would've been too clever for very little gain.

 

[LeP]

Tested on 1.31 and it crashes when trying to execute the instruction that implements the 'function' operation:

    local code c = function ClearSelection

I suppose instead of giving an error (which it doesn't) the compiler could've generated an anonymous stub like:

function ClearSelection* takes nothing retuns nothing
    call ClearSelection()
endfunction

or something, but that would've been too clever for very little gain.

Thanks for testing. Indeed the fix is very easy but it's not pjass' domain to do so.
Here is a fresh pjass version which reports this.

$ cat tests/should-fail/native-as-code.j
native TriggerSyncStart takes nothing returns nothing

function foo takes nothing returns nothing
    local code bla = function TriggerSyncStart
endfunction

$ ./pjass tests/should-fail/native-as-code.j
tests/should-fail/native-as-code.j:4: Cannot use native 'TriggerSyncStart' as code
tests/should-fail/native-as-code.j failed with 1 error
Parse failed: 1 error total

 

[Aniki]

More nit-picking:

Hex literals using the '0X' prefix are not recognized:

globals
    integer a1 = 0x0
    integer a2 = 0X0 // error
endglobals

Variable shadowing interacting with constant function:

globals
    integer x
endglobals
constant function foo takes nothing returns nothing
    local integer x
    set x = 5 // Assignment to global variable x in constant function
endfunction

 

[LeP]

More nit-picking:

Hex literals using the '0X' prefix are not recognized:

globals
    integer a1 = 0x0
    integer a2 = 0X0 // error
endglobals

Variable shadowing interacting with constant function:

globals
    integer x
endglobals
constant function foo takes nothing returns nothing
    local integer x
    set x = 5 // Assignment to global variable x in constant function
endfunction

Cheers m8 keep 'em coming.

 

[Prometheus3375]

Can you, please, add link to the readme or repo to the main post?

 

[Prometheus3375]

I believe, last version is bugged.
Consider this file.
Version 02.08.2020 gives a syntax error.
Version 04.08.2020 crashes.
@LeP, can you look at this?

 

[LeP]

I believe, last version is bugged.
Consider this file.
Version 02.08.2020 gives a syntax error.
Version 04.08.2020 crashes.
@LeP, can you look at this?

Should be fixed. Thanks for the report.

 

[Prometheus3375]

Too many errors (~25) due to missing call. Is it OK?

function Trig_Vampirizm_Actions takes nothing returns nothing
    local real h= ( GetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE) / 400 ) * GetUnitAbilityLevel(udg_VampirizmU, 'A018')
    SetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE, GetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE) - h)
    SetUnitState(udg_VampirizmU, UNIT_STATE_LIFE, GetUnitState(udg_VampirizmU, UNIT_STATE_LIFE) + h)
    call DestroyEffect(AddSpecialEffectTarget("Abilities\\Spells\\Human\\Heal\\HealTarget.mdl", udg_VampirizmU, "origin"))
endfunction

 

[Prometheus3375]

It also seems like that such an error makes pjass to freeze on Win7. Does pjass still support Win7? If not, which version is the latest for Win7?

 

[LeP]

Too many errors (~25) due to missing call. Is it OK?

function Trig_Vampirizm_Actions takes nothing returns nothing
    local real h= ( GetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE) / 400 ) * GetUnitAbilityLevel(udg_VampirizmU, 'A018')
    SetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE, GetUnitState(GetTriggerUnit(), UNIT_STATE_LIFE) - h)
    SetUnitState(udg_VampirizmU, UNIT_STATE_LIFE, GetUnitState(udg_VampirizmU, UNIT_STATE_LIFE) + h)
    call DestroyEffect(AddSpecialEffectTarget("Abilities\\Spells\\Human\\Heal\\HealTarget.mdl", udg_VampirizmU, "origin"))
endfunction

View attachment 376998

Error recovery is rather hard, it can only be handled case by case (in pjass). I can try if i can give a better experience for this but no promises.

It also seems like that such an error makes pjass to freeze on Win7. Does pjass still support Win7? If not, which version is the latest for Win7?

I also still have a win7 machine as my only windows machine. Feel free to PM me a freezing .j file and i can have a look.

 

[LeP]

Added special case for missing call which should both provide a better error message and also doesn't mess up the following checking.

$ cat tests/should-fail/no-call.j
function foo takes nothing returns nothing
    return
    foo()
    call foo()
endfunction
$ pjass tests/should-fail/no-call.j
tests/should-fail/no-call.j:3: Missing 'call'
tests/should-fail/no-call.j failed with 1 error
Parse failed: 1 error total