LUA_ENV might not be the "_ENV" string

[Jure Bagić]

The new way of declaring global variables and maintaining backwards
compatibility is very elegant. Looking a bit at the implementation I stumbled
upon a small detail that might need fixing in lparser.c:502 the
'buildglobal' function:
> if (var->k == VGLOBAL)
> luaK_semerror(ls, "_ENV is global when accessing variable '%s'",
> getstr(varname));

The name of the 1st upvalue of the chunk (the environment) is dictated by the
definition of LUA_ENV, so maybe instead do:
> if (var->k == VGLOBAL)
> luaK_semerror(ls, "%s is global when accessing variable '%s'",
> LUA_ENV, getstr(varname));

This also works:
> if (var->k == VGLOBAL)
> luaK_semerror(ls, "%s is global when accessing variable '%s'",
> getstr(ls->envn), getstr(varname));

-- Jure

[Roberto Ierusalimschy]

Thanks for the feedback. I would prefer a third way:

luaK_semerror(ls, LUA_ENV " is global when accessing variable '%s'",

-- Roberto

[Jure Bagić]

On 25/12/04 01:20PM, Roberto Ierusalimschy wrote:
>
> Thanks for the feedback. I would prefer a third way:
>
> luaK_semerror(ls, LUA_ENV " is global when accessing variable '%s'",
>
> -- Roberto

Ok, you win.
--
Jure

[Roberto Ierusalimschy]

> > Thanks for the feedback. I would prefer a third way:
> >
> > luaK_semerror(ls, LUA_ENV " is global when accessing variable '%s'",
> >
> > -- Roberto
>
> Ok, you win.

Actually, I am not sure I prefer this :-) To know this code is safe we
need to know that LUA_ENV is a valid variable name and, therefore, cannot
contain a '%'. In the end, it is more complex than it looks.

-- Roberto

[Sainan]

I'm sure there's a macro metaprogamming exercise here somewhere. :^)

-- Sainan