Hacking the parser (Was: Re: Removing limitations of OP_SELF)

["Soni L." <fakedme@...>]

On 17/12/14 01:26 PM, Soni L. wrote:
> On 17/12/14 03:26 AM, Dirk Laurie wrote:
> > 2014-12-17 4:54 GMT+02:00 Soni L. <fakedme@gmail.com>:
> >
> > > First of all I want to be able to _G:table.insert("something").
> > > Secondly, I want to be able to _G:["print"]().
> > What exactly are these supposed to be syntax sugar for?
> It's opcode sugar, actually.
>
> _G:["print"](...) is already supported by the opcode, just use a 
> non-constant index. It basically translates to _G["print"](_G, ...) 
> (but _G only gets evaluated once).
>
> _G:table.insert("something") is a bit trickier to implement. (it uses 
> more opcodes) But think about it like this: in the same way that 
> _G:print() translates to _G.print(_G), _G:table.insert("something") 
> translates to _G.table.insert(_G, "something"). (add `do local _ = _G 
> ... end` boilerplate as needed)
>
> And ofc, _G:["table"]["insert"]("something"), I guess you can guess 
> what this does.
Here's what I tried: https://github.com/SoniEx2/sexlua/commit/b279aebc50b5

When I try _G:_G.print("test") I get

    [soniex2@soniex-pc src]$ ./luac -l -l -o /dev/null -
    _G:_G.print("test")

    main <stdin:0,0> (5 instructions at 0x10800c0)
    0+ params, 3 slots, 1 upvalue, 0 locals, 3 constants, 0 functions
        1    [1]    GETTABUP     0 0 -1    ; _ENV "_G"
        2    [1]    SELF         0 0 -1    ; "_G"
        3    [1]    LOADK        2 -3    ; "test"
        4    [1]    CALL         1 -131 1
        5    [1]    RETURN       0 1
    constants (3) for 0x10800c0:
        1    "_G"
        2    "print"
        3    "test"
    locals (0) for 0x10800c0:
    upvalues (1) for 0x10800c0:
        0    _ENV    1    0

It's missing a GETTABLE (for the 2nd _G, with key "print" aka -2) and 
that CALL should be... CALL 0 2 1? Completely broken.

And when I try _G:print("test") I get

    [soniex2@soniex-pc src]$ ./luac -l -l -o /dev/null -
    _G:print("test")

        2    [1]    SELF         0 0 -2    ; "print"
        3    [1]    LOADK        2 -3    ; "test"
        4    [1]    CALL         0 3 1
        5    [1]    RETURN       0 1
    constants (3) for 0x1a0d0c0:
        1    "_G"
        2    "print"
        3    "test"
    locals (0) for 0x1a0d0c0:
    upvalues (1) for 0x1a0d0c0:
        0    _ENV    1    0

Which has wrong CALL. (0 2 1 in unmodified/vanilla Lua. I'm not sure why 
this is happening.)

--
Disclaimer: these emails are public and can be accessed from <TODO: get a non-DHCP IP and put it here>. If you do not agree with this, DO NOT REPLY.