simple upvalue decompile incrrect

[kernus]

lua code:

local b = 2
function add()
c = 3
return b + c
end
print(add())

decompiled:

slot0 = 2
function add()
c = 3
return 3 + c
end
print(add())
return

with bugfix-locals branch, ljd crashed:
Traceback (most recent call last):
File "main.py", line 123, in
retval = main()
File "main.py", line 104, in main
ljd.ast.unwarper.unwarp(ast)
File "/Users/../ljd/ast/unwarper.py", line 38, in unwarp
_glue_flows(node)
File "/Users/../ljd/ast/unwarper.py", line 61, in _glue_flows
assert isinstance(blocks[-1].warp, nodes.EndWarp)
AttributeError: 'Return' object has no attribute 'warp'

[jjdredd]

yes, bugfix-locals is bugged, please use bugfix or master.

[kernus]

well, does this upvalue thing hard to fix? I get the bytecode parsed like below:

1b 4c 4a 01 02

23
00 flags: no ffi, no vararg, no child
00 argsize
02 framesize
01 upvalues number
01
00
06 instruction count
27 00 03 00 KSHORT 0 3
35 00 00 00 GSET 0 0 ; "c"
2b 00 00 00 UGET 0 0 ; b
34 01 00 00 GGET 1 0 ; "c"
1e 00 01 00 ADDVV 0 0 1
48 00 02 00 RET1 0 2

00 c0 upvalue refs
06 63 c

36
03 flags: no ffi, vararg, has child
00
03
00 kup
03 kgc
00 knum
09 instruction number
27 00 02 00 KSHORT 0 2
31 01 00 00 FNEW 1 0 ; number.lua:3
35 01 01 00 GSET 1 1 ; "add"
34 01 02 00 GGET 1 2 ; "print"
34 02 01 00 GGET 2 1 ; "add"
3e 02 01 00 CALL 2 0 1
3d 01 00 01 CALLM 1 1 0
30 00 00 80 UCLO 0 0009
47 00 01 00 RET0 0 1

0a 70 72 69 6e 74 print
08 61 64 64 add
00 child

00 eof

hard to run it in mind.

[jjdredd]

It's hard for me to fix, the easiest solution here is to use a less bugged branch like master or bugfix.
Why did you chose to use bugfix-locals anyway?

[kernus]

With master, I get:
slot0 = 2
function add()
c = 3
return 3 + c
end
print(add())
return

the upvalue parsing is wrong.

As for bugfix-locals, I just thought it's latest enough, lol.

Really hope this project is active, I have reading it for some weeks, it's interesting, however, I have little experience with luajit or ljd.

the upvalue ref in add function is 00 c0, how can it refs to b = 2 (i.e. KSHORT 0 2?)? googled it for whole day, just no useful ideas.

[jjdredd]

As for bugfix-locals, I just thought it's latest enough, lol.

It's kinda experimental.
As you can see I have made several lame attempts to fix the locals/slotworks, but I don't understand
the ljd algorithm very well and don't have much time to work with it.

Really hope this project is active

I'm sorry, it's not. the original author abandoned it and I don't think I can fix this. IDK if it's easier to rewrite the whole thing or try to fix it.

If you ever return to irc, be sure to ping me (Judge_Dredd).
EDIT: you can find some discussions withe the original author in the original repo's issues on github:
https://github.com/NightNord/ljd/issues

[kernus]

Another found, scripts below works, just remove local c

local b = 2

function add()
return b
end

print(add())

decompiled

slot0 = 2

function add()
return slot0
end

print(add())

Guess the slot0 within add function stands for the original c variable, not the outside b slot0, will check it tomorrow.