Thanks all for commenting, I changed the implementation and put the codes here: https://github.com/Shuai-Meng/wine/commit/bb212763ffd639a238b5a47f2c6a27047a... It passed on both winxp and wine. But this patch still needs your comments. 2014-07-18 17:30 GMT+08:00 Jacek Caban <jacek(a)codeweavers.com>:
On 07/18/14 04:11, Shuai Meng wrote:
2014-07-17 11:11 GMT+08:00 Dmitry Timoshkov <dmitry(a)baikal.ru>:
Shuai Meng <mengshuaicalendr(a)gmail.com> wrote:
+ switch(V_VT(arg)) { + case VT_UI1: + case VT_I2: + case VT_I4: + case VT_I8: + case VT_R4: + case VT_R8: + case VT_BOOL: + case VT_EMPTY: + case VT_CY: + V_BOOL(res) = VARIANT_TRUE; + break;
This list is far from being complete. It seems that it should also contain at least VT_UI2, VT_UI4, VT_UI8, VT_I1, VT_INT, VT_UINT. VT_VECTOR, VT_ARRAY and VT_BYREF modifers probably also should be taken into account.
-- Dmitry.
Hi, I think over it again. Let's face the fact: what are all the possible results of switch(V_VT(arg)) ? Will VT_UI2, VT_UI4, VT_UI8, VT_I1, VT_INT, VT_UINT and so on exist? My answer is no. arg comes from users' input, which means it only belongs to the subtypes of variant, see here: http://msdn.microsoft.com/en-us/library/9e7a57cf(v=vs.84).aspx
No, your answer is wrong. VBScript code may be called from non-VBScript (and the other way around) and then any type of VARIANT may be passed around. Also, please add a test with an object having default value, which is of numeric type. I have a feeling that the right thing to do here is to call to_double and see if it fails or not, but this needs tests to confirm.
Jacek