Re: [PATCH] vbscript: Implemented String

[Nikolay Sivov]

On 10/20/2014 19:58, Shuai Meng wrote:
> Thank you very much for commenting on this patch.
>
> 2014-10-20 0:06 GMT+08:00 Nikolay Sivov <[e-mail] 
> >:
>     +    str = SysAllocStringLen(NULL, 1023);
>     +    newstr = SysAllocStringLen(NULL, 1023);
>
> Where this length comes from?
> Well, I tested String on windows xp, and found that 1023 was the
> limit, when given a number bigger than that, the output kept the
> length of 1023.

Can you add this too as a test? In some compact way if possible.
>     +    switch(V_VT(arg + 1)) {
>     +    case VT_NULL:
>     +        return MAKE_VBSERROR(VBSE_ILLEGAL_NULL_USE);
>     +    case VT_BSTR:
>     +        str = V_BSTR(arg + 1);
>     +        break;
>     +    case  VT_ARRAY|VT_BYREF|VT_VARIANT:
>     +        return DISP_E_TYPEMISMATCH;
>     +    default:
>     +        hres = to_short(arg + 1, &tmp);
>     +        if(FAILED(hres))
>     +            return hres;
>     +        str[0] = (char)tmp;
>     +        break;
>     +    }
>
> You only need first character, right? Then why do you need a full
> BSTR pointer in VT_BSTR case? And assigning it
> to 'str' you leak a previously allocated buffer. 
>
> So how do I get the first character of (arg + 1)?  How about this: 
> str[0] = * V_BSTR(arg + 1)
> In fact I don't quite understand how SysAllocStringLen work, but I see 
> it is used in the former function, so I think maybe it is necessary.

What I don't get is why do you allocate 'str' at all, it should be just 
'WCHAR ch;' variable.
> Why cast to (char)tmp?
>
> I think the type of str[0] is WCHAR, and tmp is an integer, shouldn't 
> we make a cast?

WCHAR is 'unsigned short'. And it feels like it needs more tests for 
other variant types.
>     +    else if(len == 0)
>     +        newstr = '\0';
>
> Same way you're losing pointer to allocated buffer.