Hi Alistair,
On 03/30/12 12:29, Alistair Leslie-Hughes wrote:
Hi,
Changelog: jscript: Add VT_I2 support
hres = IDispatchEx_InvokeEx(dispex, id, ctx->lcid, flags, dp, retv, &ei->ei, &ctx->jscaller->IServiceProvider_iface); + if(SUCCEEDED(hres) && retv) + { + if(V_VT(retv) == VT_I2) + VariantChangeType(retv, retv, 0, VT_I4); + }
Since the conversion here is trivial, why not do it directly?
--- a/dlls/jscript/jsutils.c +++ b/dlls/jscript/jsutils.c @@ -191,6 +191,10 @@ HRESULT to_primitive(script_ctx_t *ctx, VARIANT *v, jsexcept_t *ei, VARIANT *ret case VT_R8: *ret = *v; break; + case VT_I2: + V_VT(ret) = VT_I4; + V_I4(ret) = V_I2(v); + break;
I thought the findings we've discussed off-list was that internal values should never be of VT_I2 types, so this should never happen. If it's not the case, I'd like to see a test case.
+function TestVT_I2() +{ + varI2 = GetShort(); + num3 = 3 + + ok(getVT(varI2) === "VT_I4", "getVT(varI2) = " + getVT(varI2));
Please include getVT(GetShort()) test I've showed you. This eliminates a lot of flow of value.
+ if(!strcmp_wa(bstrName, "GetShort")) { + *pid = DISPID_GLOBAL_SHORT; + return S_OK; + }
Please follow JavaScript names convention (this should be getShort).
Cheers, Jacek