PJass updates

[LeP]

For leandro and all the crazy hackers out there i added the flags //# +nosyntaxerror and //# +nosemanticerror which are quite the sledgehammers. They do exactly as they are named: ignore syntax/semantic errors in the function so only use these if you know what you are doing.

I still would consider this a test release as this is based off of my refactor branch. I also added a //# +filter flag which enables the old behavior of the Filter check for functions that do not return booleans. I don't know if people are still on these old patches but i felt like adding it again.

 

[DracoL1ch]

hm could you make it not a flag, but part of the script? some tags to wrap the code we don't like to get notified about.
it's not a big deal, tho

 

[AGD]

Help, How do I update my PJass?

 

[LeP]

Help, How do I update my PJass?

You download the zip. Extract pjass.exe and replace your old pjass.exe with the new one.
Your old pjass.exe should reside in your jassnewgenpack-folder/jasshelper/

 

[AGD]

You download the zip. Extract pjass.exe and replace your old pjass.exe with the new one.
Your old pjass.exe should reside in your jassnewgenpack-folder/jasshelper/

Okay

How about the other zip file, the other one of the two above?

 

[LeP]

Okay

How about the other zip file, the other one of the two above?

You could download it and see it yourself. The other zip is named -src.zip.
I release the source code of pjass aswell as the exe.

 

[AGD]

I've already replaced my pjass.exe. I'm curious if there's still something I have to do with the source codes?

 

[DracoL1ch]

You could download it and see it yourself. The other zip is named -src.zip.
I release the source code of pjass aswell as the exe.

please, can we have something like //#noparse option, which will disable synt check for line it stands at? or //#noparsestart //#noparseend for blocks, idk. isn't big deal, but irritating

 

[LeP]

please, can we have something like //#noparse option, which will disable synt check for line it stands at? or //#noparsestart //#noparseend for blocks, idk. isn't big deal, but irritating

i dont see the use with //# +nosyntaxerror already in place.

I've already replaced my pjass.exe. I'm curious if there's still something I have to do with the source codes?

no.

 

[DracoL1ch]

ahhh I though it's disable syntax check competely. my bad, thanks

 

[AGD]

thanks for the help

 

[DracoL1ch]

one more thing
if function receives a real value
call MyFunc(.) won't cause any alert, tho this is invalid code. it should be either MyFunc(0.) or MyFunc(.0), but not just dot.

 

[LeP]

one more thing
if function receives a real value
call MyFunc(.) won't cause any alert, tho this is invalid code. it should be either MyFunc(0.) or MyFunc(.0), but not just dot.

Oh yes i was always somewhat aware of that but decided to ignore it.
Here's a new version with that fixed.
Also some new features:

$ cat tests/should-fail/not-enough-parameters.j
native Foo takes integer x, real bla, string wat, handle asdf, integer y returns nothing

function test takes nothing returns nothing
    call Foo(3, 3.1, "bla")
endfunction

$ ./pjass.exe tests/should-fail/not-enough-parameters.j
tests/should-fail/not-enough-parameters.j:5: Not enough arguments passed to function Foo. Still missing: asdf, y
tests/should-fail/not-enough-parameters.j failed with 1 error
Parse failed: 1 error total

$ cat tests/should-fail/wrong-types-in-call.j
native Foo takes integer x, real bla returns nothing

function test takes nothing returns nothing
    call Foo(3.4, "foo" )
endfunction

$ ./pjass.exe tests/should-fail/wrong-types-in-call.j
tests/should-fail/wrong-types-in-call.j:5: Cannot convert real to integer in parameter x in call to Foo
tests/should-fail/wrong-types-in-call.j:5: Cannot convert string to real in parameter bla in call to Foo
tests/should-fail/wrong-types-in-call.j failed with 2 errors
Parse failed: 2 errors total

Also this was cross-compiled from an ubuntu-vm... so it works for me atleast

 

[Trigger.edge]

Hi man! I do not want to be impertinent, but I wanted to know if you were making any changes in the pjass to work with Memory Hack? (I'm really impassive)

 

[LeP]

Hi man! I do not want to be impertinent, but I wanted to know if you were making any changes in the pjass to work with Memory Hack? (I'm really impassive)

But pjass is already able to handle with the memory hack (if not, hit me up) if you insert the right annotations. Im sure if you search this thread you can find which annotations im talking about. You have to modify the memory-hack j-file. Not the code, you just have to add a few comments. Once you've tried to do it yourself and still have problems feel free to ask.

e: look at my code here for an example.

 

[DracoL1ch]

Hi man! I do not want to be impertinent, but I wanted to know if you were making any changes in the pjass to work with Memory Hack? (I'm really impassive)

ye sry bout that, I dont have to pass pjass flawlessly, since Im using raw jass

 

[Trigger.edge]

But pjass is already able to handle with the memory hack (if not, hit me up) if you insert the right annotations. Im sure if you search this thread you can find which annotations im talking about. You have to modify the memory-hack j-file. Not the code, you just have to add a few comments.

Thanks man! I just had to use //# +nosemanticerror (I thought it was just a simple comment xd)

EDIT: <33

 

[DracoL1ch]

can you add listing of all ingame natives (including common/ai/blizzard) into check list to see whenever there's declared non-existant native in the script? just forgot to disable side-stuff and took a while to understand what went wrong

 

[LeP]

can you add listing of all ingame natives (including common/ai/blizzard) into check list to see whenever there's declared non-existant native in the script? just forgot to disable side-stuff and took a while to understand what went wrong

Hmm, i don't think this is any good for multiple reasons.
1) You very rarely declare a new native. Most of the time it's one of three known ones.
2) If you declare a new native you should be aware of it.
3) I would to have keep track of natives over patches. I know i could special case it to some patch as it would be avaible only via some flag anyways but it seems like a very specialized use-case to me.

 

[DracoL1ch]

Hmm, i don't think this is any good for multiple reasons.
1) You very rarely declare a new native. Most of the time it's one of three known ones.
2) If you declare a new native you should be aware of it.
3) I would to have keep track of natives over patches. I know i could special case it to some patch as it would be avaible only via some flag anyways but it seems like a very specialized use-case to me.

can we assume there would be no patches since 27? it barely awaited by anyone
ye, normally im aware of, but sometimes.. really sometimes

 

[LeP]

I did this. Still open for better wording.
Also now with all sources hopefully. I feel already said something likes this before...

 

[puffdragon]

Pjass plugin for notepad++ would be superb, even tho its awesome we get this updates anyway

 

[DracoL1ch]

can you plz add a check for GetObjectName used in globals block? this func causes instant fatal on selecting map in the list or starting game on it. basically bad idea to ask for object name, yeah, but why the fuck fatal, plz.

 

[LeP]

can you plz add a check for GetObjectName used in globals block? this func causes instant fatal on selecting map in the list or starting game on it. basically bad idea to ask for object name, yeah, but why the fuck fatal, plz.

yeah sounds like a good idea; i have several other such functions in my jassdb.

 

[IcemanBo]

Sounds good, but there are also some functions that should not be used for example onInits, like multiboard creations.
The point is you can't entirely protect the user from "wrong" jass usage, when it's about semantics.
Generaly maybe: syntax allowed -> technicaly no error from pjass.
But sure, in case you want to catch some cases, it will have only benefits, was just saying.

 

[DracoL1ch]

Can you please add support for the following bug:

globals
 boolean b
endglobals

function A takes nothing returns nothing
    local integer b
...blah blah
endfunction

function B takes nothing returns nothing
    set b=true
if b then
endif
endfunction

thats typecasting indeed, but sometimes - unwillingly. Im working with de-protected maps and 1-2 letters fuck up local declarations in custom functions sometimes. if the parser can detect if global variable somehow being redeclared as a different type, it would be great, as typecasting isn't big thing for 99.9% mapmakers anyway

 

[LeP]

can you plz add a check for GetObjectName used in globals block? this func causes instant fatal on selecting map in the list or starting game on it. basically bad idea to ask for object name, yeah, but why the fuck fatal, plz.

I've added a bunch of functions which crash in a globals block. Can be disabled via +noruntimeerror flag.
I also added functions which return null in a globals block. I'm open to discussion about this one though. Might be better to just write a general jass linter.

Can you please add support for the following bug:

globals
 boolean b
endglobals

function A takes nothing returns nothing
    local integer b
...blah blah
endfunction

function B takes nothing returns nothing
    set b=true
if b then
endif
endfunction

thats typecasting indeed, but sometimes - unwillingly. Im working with de-protected maps and 1-2 letters fuck up local declarations in custom functions sometimes. if the parser can detect if global variable somehow being redeclared as a different type, it would be great, as typecasting isn't big thing for 99.9% mapmakers anyway

I don't understand. What's the error here? Can't you just use +shadow flag?

 

[DracoL1ch]

that's invalid code which won't run but your tool won't notice that.
since declared as integer at local block 'b' can't get assigned boolean value, only integer. engine counts it as error

 

[LeP]

that's invalid code which won't run but your tool won't notice that.
since declared as integer at local block 'b' can't get assigned boolean value, only integer. engine counts it as error

Can you give me a map which wont run but which pjass accepts?

 

[DracoL1ch]

newgen will auto-add l__ prefix, so thats the fix purely for raw-coders (me)
here's the map, unless you re-compile it to add prefix, it won't run but code looks fine for pjas

 

[LeP]

Can you please add support for the following bug:

globals
 boolean b
endglobals

function A takes nothing returns nothing
    local integer b
...blah blah
endfunction

function B takes nothing returns nothing
    set b=true
if b then
endif
endfunction

thats typecasting indeed, but sometimes - unwillingly. Im working with de-protected maps and 1-2 letters fuck up local declarations in custom functions sometimes. if the parser can detect if global variable somehow being redeclared as a different type, it would be great, as typecasting isn't big thing for 99.9% mapmakers anyway

Have you used +shadow? That could've helped to find the error.
Nevertheless i have corrected the handling of shadowed variables. Here are the details:
If you shadow a variable with an incompatible type, that is a type which doesn't have the same primitive ancestor, every usage of that variable later in the script file(sic) will have the wrong type and wc3 will bail. You can still use the original global variable as long as all occurences of it in the source file are above the bad shadow.

This is of course how the current returnbug works.

 

[DracoL1ch]

thanks
idk, I didn't apply any flags before now

 

[LeP]

thanks
idk, I didn't apply any flags before now

But it even was you who requested error on shadowing...

 

[DracoL1ch]

But it even was you who requested error on shadowing...

my memory is rly short, especially for forums. tons of info everyday, sorry. I didn't need shadowing on my dota, but last month I got hooked to another project which is acutally clusterfuck with a single-letter globals like "x" or so. After all this time I would never recall about the shadow flags

 

[Flux]

Where can I see the changelog?

Also, is it possible to make pjass inlines one-liner ExecuteFunc? Example:

library L

    struct MyStruct
        private static real x
        private static method oneLiner takes nothing returns nothing
            set thistype.x = 42
        endmethod
    endstruct

endlibrary

function Main takes nothing returns nothing
    call ExecuteFunc("s__MyStruct__oneLiner")
endfunction

will turn into something like

function Main takes nothing returns nothing
    set s__MyStruct__s = 42
endfunction

 

[LeP]

Where can I see the changelog?

I don't have a changelog per se but all the changes are documented in this thread. And most of the time in the post where i attached a new version.

Also, is it possible to make pjass inlines one-liner ExecuteFunc? Example:

library L

    struct MyStruct
        private static real x
        private static method oneLiner takes nothing returns nothing
            set thistype.x = 42
        endmethod
    endstruct

endlibrary

function Main takes nothing returns nothing
    call ExecuteFunc("s__MyStruct__oneLiner")
endfunction

will turn into something like

function Main takes nothing returns nothing
    set s__MyStruct__s = 42
endfunction

No. pjass is only to check pure jass. This should be done in another tool like jasshelper/wurst/<insert any jass optimizer here>.
Also your proposed optmisation actaully changes the semantic of the code so no tool should rewrite it that way.

 

[Trigger.edge]

Where can I see the changelog?

Also, is it possible to make pjass inlines one-liner ExecuteFunc? Example:

library L

    struct MyStruct
        private static real x
        private static method oneLiner takes nothing returns nothing
            set thistype.x = 42
        endmethod
    endstruct

endlibrary

function Main takes nothing returns nothing
    call ExecuteFunc("s__MyStruct__oneLiner")
endfunction

will turn into something like

function Main takes nothing returns nothing
    set s__MyStruct__s = 42
endfunction

That would be very useful.

 

[TriggerHappy]

That would be very useful.

I don't see why. ExecuteFunc runs the code in a separate "thread", preventing thread crashes from stopping your entire initialization.

These optimizations are easy to make yourself, anyway.

 

[Flux]

It's mostly because of single lined onInit functions. The onInits are compiled into ExecuteFunc and most of them (if not, all) are placed at the very bottom of the script. The purpose of executing on a new thread seems pointless and I doubt these one-liner onInit would cause an OPLimit.

"easy to make yourself anyway"
How? Without doing it manually.

 

[TriggerHappy]

How? Without doing it manually.

Clearly I meant manually, and by having one init function?

 

[Flux]

I think that's not possible. Ok here's the thing, I'm using this which is a module. It has an onInit function that looks something like this

private static method onInit takes nothing returns nothing
            set recycler[0] = 1
endmethod

Everytime it is implemented, it generates it's own 1-lined onInit function that has to go through ExecuteFunc. No workaround can be done via script in my opinion. Anyway, this should be done by JassHelper not pjass but no one manages JassHelper anymore (I think?) while pjass is still somewhat active so I just tried my luck here.

Sorry for the offtopic.

 

[Lord of theDing]

Could you add something similar as //# +nosemanticerror but a bit weaker.
Ideally I would like something to annotate a function that takes a code parameter, to disable the parameter checks. I think an example describes it best:

function TakesParameter takes integer i returns nothing
endfunction

function Evil takes code c returns nothing
endfunction

//# +nosemanticerror
function Foo takes nothing returns nothing
    call Evil(function TakesParameter)
endfunction

To run this code I have to annotate Foo with //# +nosemanticerror. What I would like to do is:

//# +AllowArbitraryCodeParameter
function Evil takes code c returns nothing
endfunction

function Foo takes nothing returns nothing
    call Evil(function TakesParameter)
endfunction

If that is too much work to implement, something like this would also help:

function Foo takes nothing returns nothing
    //# +nosemanticerror
    call Evil(function TakesParameter)

   // rest of function has error checking like normal
endfunction

I'm working with evil(TM) bytecode hackery and it's a bit annoying to have no error checking on like half of all my functions.

P.S. Really big thank you for pjass, it's a really awesome tool that makes coding in jass bearable.

 

[LeP]

Could you add something similar as //# +nosemanticerror but a bit weaker.
Ideally I would like something to annotate a function that takes a code parameter, to disable the parameter checks. I think an example describes it best:

function TakesParameter takes integer i returns nothing
endfunction

function Evil takes code c returns nothing
endfunction

//# +nosemanticerror
function Foo takes nothing returns nothing
    call Evil(function TakesParameter)
endfunction

To run this code I have to annotate Foo with //# +nosemanticerror. What I would like to do is:

//# +AllowArbitraryCodeParameter
function Evil takes code c returns nothing
endfunction

function Foo takes nothing returns nothing
    call Evil(function TakesParameter)
endfunction

If that is too much work to implement, something like this would also help:

function Foo takes nothing returns nothing
    //# +nosemanticerror
    call Evil(function TakesParameter)

   // rest of function has error checking like normal
endfunction

I'm working with evil(TM) bytecode hackery and it's a bit annoying to have no error checking on like half of all my functions.

P.S. Really big thank you for pjass, it's a really awesome tool that makes coding in jass bearable.

Hm yes i can see that //# +no<x>error is quite the sledgehammer.
Adding a new flag would be very very easy but im unsure if it's the right way. If you don't watch out you get a million flags in the end. I like the approach of annotating functions instead of lines, which also constrains me a bit. I guess just adding a flag to not check code arguments is the pragmatic way here; and then see if we ever get a flag mess...

 

[Lord of theDing]

If you don't watch out you get a million flags in the end.

I see, you also use gcc

One way to avoid the flag hell may be not to add a flag for every possible error, but one flag with a parameter.
Like pjass +disable-error function-parameter +disable-error return-bug.

 

[LeP]

Strange, it does catch local variables which aren't initilized but used, but doesn't check globals for all the same issue? I just got this issue lately.

globals
 unit someGlobal
endglobals
function test takes nothing returns nothing
 call BJDebugMsg(GetUnitName(someGlobal))
endfunction

Because that is, atleast with pjass' architecture, a way harder problem.

 

[KILLCIDE]

Opened by request.

 

[LeP]

Just bringing the binaries posted here uptodate with my git.
Changes from 0500a60:

• Better errormessage for missing local identifier
• Better errormessage for passing arguments to a function literal: return function foo(3)
• Don't report an array variable with missing index as unintialized

 

[btdonald]

How to install this? I wanna try Memory hacks but I can't use it

 

[LeP]

How to install this? I wanna try Memory hacks but I can't use it

here

 

[zachhandley]

@LeP is PJASS updated to 1.29?