Re: Large number bytecode weirdness on different platforms

[Chris <coderight@...>]

Hmm, I didn't remember that you can do that.   luac -l crashes with
the Windows binary from Lua Binaries.  My own compiled version works
though.

Linux:
main <test.lua:0,0> (7 instructions, 28 bytes at 0x8060d48)
0+ params, 2 slots, 0 upvalues, 0 locals, 3 constants, 0 functions
       1       [1]     GETGLOBAL       0 -1    ; num
       2       [1]     LT              0 0 -2  ; - 4294967296
       3       [1]     JMP             0       ; to 4
       4       [2]     GETGLOBAL       0 -1    ; num
       5       [2]     LT              0 0 -3  ; - 68719476736
       6       [2]     JMP             0       ; to 7
       7       [2]     RETURN          0 1

main <test.lua:0,0> (7 instructions, 28 bytes at 002F5B78)
0+ params, 2 slots, 0 upvalues, 0 locals, 2 constants, 0 functions
       1       [1]     GETGLOBAL       0 -1    ; num
       2       [1]     LT              0 0 -2  ; - 4294967295
       3       [1]     JMP             0       ; to 4
       4       [2]     GETGLOBAL       0 -1    ; num
       5       [2]     LT              0 0 -2  ; - 4294967295
       6       [2]     JMP             0       ; to 7
       7       [2]     RETURN          0 1

As you mentioned it looks like the Windows version is overflowing.
Ah, now I see it's a problem with the "0x" hex parsing because if I
use the normal integer value there is no problem.  Ugh, I hate
inconsistent behaviour ;)

Thanks!  I'll have to remember "-l" next time.  I'll look into the 0x
parsing and see what the issue is.

CR


On 1/29/07, Rici Lake <lua@ricilake.net> wrote:
> On 29-Jan-07, at 4:16 PM, Chris wrote:
>
> > I have this simple test.lua file:
> >
> >  if num < 0x100000000 then end
> >  if num < 0x1000000000 then end
> >
> > Using vanilla source Lua 5.1.1:
> >
> > On Ubuntu Linux:
> > luac -o test.linux test.lua
> >
> > On Windows:
> > luac.exe -o test.win test.lua
> >
> > $ ls -l
> > -rw-r--r-- 1 user user    143 2007-01-29 15:46 test.linux
> > -rw------- 1 user user    134 2007-01-29 15:48 test.win
> >
> > Notice the size difference.  The bytecode is very different when I
> > compare it.  What's happening?
> >
> > The weird thing is, if you remove one of the "if" statements they
> > compile the same.  You have to have both of them in there.  Also, both
> > numbers must be greater than 32-bit but not all combinations produce
> > different bytecode.
>
> You'd get a clearer idea if you used luac -l to print the bytecode in a
> human readable form. Or maybe you did, and just didn't put the results
> in this message.
>
> My guess, since I don't have windows handy, is that on Windows both of
> those numbers are actually 0, while on Ubuntu they're actually the
> correct value. That means that on Windows, there is only one constant
> (0) and hence the compiled byte code is shorter by one number (type
> byte + 8 bytes for the numbers == 9 bytes, exactly the difference
> you're showing.)