2014-10-20 17:00 GMT+08:00 Piotr Caban <piotr.caban@gmail.com>:

On 10/19/14 17:44, Shuai Meng wrote:

+�� newstr = SysAllocStringLen(NULL, 1023);

You need to allocate correct length string. First argument should be used here. You're also leaking the string in error-handling paths.

��You mean I should replace 1023 with the final number of first argument? Seems good.��

+
+�� 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;

Please add a test with second argument larger then 256. It would be also interesting to see a test with second argument being BSTR with first character out of ASCII range.

��In fact it is hard to check what String returns when character is larger than 256 even below 256. For example, String(1,0) returns an empty string on xp, but Eval(String(1,0) = "") will return false, we all know 0 represents '\0' in ASCII which is not a print character. This is the same situation when the second argument is negative number.��

Please also check what happens if V_BSTR(arg+1)==NULL.

I have added such test:��

Call testStringError(2, Null, 94)

+�� hres = to_short(arg, &len);
+�� if(FAILED(hres))
+�� return hres;
+�� if(len <�� 0)
+�� return MAKE_VBSERROR(VBSE_ILLEGAL_FUNC_CALL);
+�� else if(len == 0)
+�� newstr = '\0';
+�� else if(len > 1023)
+�� len = 1023;

I guess that first argument is probably an integer.

��Do you mean we don't need ��to_short(arg, &len)?

The length of produced string is probably also not limited to 1023 characters.

I have answered this question in the reply to Nikolay~

Thanks,
Piotr

Thanks for commenting Piotr.