[PATCH v5 0/2] MR9639: Fix the integration of PowerPoint add-in in MS Office 2016
Encountered an issue with PowerPoint add-in integration. Reviewing the logs revealed the problem was due to the lack of CLSID lookup via the CurVer section and incorrect handling of the PARAMFLAG_NONE flag. A test was conducted in Windows to verify the handling of PARAMFLAG_NONE, PARAMFLAG_FIN, and PARAMFLAG_FOUT. Test class: ``` interface ITest : IDispatch { [id(1)] HRESULT TestMethod(BSTR* value); [id(2)] HRESULT TestMethodOut([out] BSTR* retVal); [id(3)] HRESULT TestMethodIn([in] BSTR* inVal); }; ``` Results in Windows: ``` --------TestMethod-------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethod] Hello from VT_BSTR Invoke returned: 0x00000000 -------TestMethodIn------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethodIn] Hello from VT_BSTR Invoke returned: 0x00000000 -------TestMethodOut------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethodOut] (null) Invoke returned: 0x00000000 ``` Results in Wine: ``` --------TestMethod-------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethod] (null) Invoke returned: 0x00000000 -------TestMethodIn------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethodIn] Hello from VT_BSTR Invoke returned: 0x00000000 -------TestMethodOut------- Variant type: 0x0008 (VT_BSTR) [CTest::TestMethodOut] (null) ``` Therefore, parameters with the PARAMFLAG_NONE flag are implied to be input parameters in Windows and are handled accordingly as PARAMFLAG_FIN. Added tests for the implemented changes. In the last two tests (their code is provided below), the CLSID is nullified in the guid_from_string function, which is why this test is left with a todo_wine tag. ``` clsid = CLSID_StdFont; hr = CLSIDFromString(L"MyApp.DocumentTest.3", &clsid); todo_wine ok(IsEqualCLSID(&clsid, &CLSID_StdFont), "Unexpected clsid %s.\n", wine_dbgstr_guid(&clsid)); clsid = CLSID_StdFont; hr = CLSIDFromString(L"MyApp.DocumentTest.5", &clsid); todo_wine ok(IsEqualCLSID(&clsid, &CLSID_StdFont), "Unexpected clsid %s.\n", wine_dbgstr_guid(&clsid)); ``` -- v5: combase: Implement CLSID lookup with CurVer fallback oleaut32: Fix parameter value loss in ITypeInfo_Invoke for byref parameters https://gitlab.winehq.org/wine/wine/-/merge_requests/9639
From: Ivan Lyugaev <valy@etersoft.ru> --- dlls/oleaut32/tests/test_reg.idl | 6 ++++ dlls/oleaut32/tests/typelib.c | 50 ++++++++++++++++++++++++++++++++ dlls/oleaut32/typelib.c | 2 +- 3 files changed, 57 insertions(+), 1 deletion(-) diff --git a/dlls/oleaut32/tests/test_reg.idl b/dlls/oleaut32/tests/test_reg.idl index 2ab7638f2fc..4b6c11d81bf 100644 --- a/dlls/oleaut32/tests/test_reg.idl +++ b/dlls/oleaut32/tests/test_reg.idl @@ -171,6 +171,12 @@ library register_test void testget8([out, retval] int* value); [propput, id(11)] void testput8([in] int value); + [id(12)] + HRESULT testbstrnone(BSTR* value, [out, retval] BSTR* retvalue); + [id(13)] + HRESULT testbstrin([in] BSTR* value, [out, retval] BSTR* retvalue); + [id(14)] + HRESULT testbstrout([out] BSTR* value, [out, retval] BSTR* retvalue); } /* uuid is same as for test_struct2 in test_tlb.idl, fields are different */ diff --git a/dlls/oleaut32/tests/typelib.c b/dlls/oleaut32/tests/typelib.c index 153964e17d5..b54e6be0806 100644 --- a/dlls/oleaut32/tests/typelib.c +++ b/dlls/oleaut32/tests/typelib.c @@ -316,6 +316,33 @@ static void WINAPI invoketest_get_testput8(IInvokeTest *iface, int value) { } +static HRESULT WINAPI invoketest_test_bstr_none(IInvokeTest *iface, BSTR* value, BSTR* retvalue) +{ + if (value != NULL && *value != NULL) + { + *retvalue = SysAllocString(*value); + } + return S_OK; +} + +static HRESULT WINAPI invoketest_test_bstr_in(IInvokeTest *iface, BSTR* value, BSTR* retvalue) +{ + if (value != NULL && *value != NULL) + { + *retvalue = SysAllocString(*value); + } + return S_OK; +} + +static HRESULT WINAPI invoketest_test_bstr_out(IInvokeTest *iface, BSTR* value, BSTR* retvalue) +{ + if (value != NULL && *value != NULL) + { + *retvalue = SysAllocString(*value); + } + return S_OK; +} + static const IInvokeTestVtbl invoketestvtbl = { invoketest_QueryInterface, invoketest_AddRef, @@ -337,6 +364,9 @@ static const IInvokeTestVtbl invoketestvtbl = { invoketest_get_testget7, invoketest_get_testget8, invoketest_get_testput8, + invoketest_test_bstr_none, + invoketest_test_bstr_in, + invoketest_test_bstr_out }; static IInvokeTest invoketest = { &invoketestvtbl }; @@ -868,6 +898,26 @@ static void test_invoke_func(ITypeInfo *typeinfo) dp.cArgs = 2; hres = ITypeInfo_Invoke(typeinfo, &invoketest, 3, DISPATCH_METHOD, &dp, &res, NULL, &i); ok(hres == DISP_E_BADPARAMCOUNT, "got 0x%08lx\n", hres); + + V_VT(args) = VT_BSTR; + V_BSTR(args) = SysAllocString(L"TestString"); + dp.cArgs = 1; + dp.cNamedArgs = 0; + + hres = ITypeInfo_Invoke(typeinfo, &invoketest, 12, DISPATCH_METHOD, &dp, &res, NULL, &i); + ok(hres == S_OK, "got 0x%08lx\n", hres); + ok(V_VT(&res) == VT_BSTR, "got %d\n", V_VT(&res)); + ok(!lstrcmpW(V_BSTR(&res), L"TestString"), "got %ls\n", !lstrcmpW(V_BSTR(&res), L"") ? L"(null)" : V_BSTR(&res)); + + hres = ITypeInfo_Invoke(typeinfo, &invoketest, 13, DISPATCH_METHOD, &dp, &res, NULL, &i); + ok(hres == S_OK, "got 0x%08lx\n", hres); + ok(V_VT(&res) == VT_BSTR, "got %d\n", V_VT(&res)); + ok(!lstrcmpW(V_BSTR(&res), L"TestString"), "got %ls\n", V_BSTR(&res)); + + hres = ITypeInfo_Invoke(typeinfo, &invoketest, 14, DISPATCH_METHOD, &dp, &res, NULL, &i); + ok(hres == S_OK, "got 0x%08lx\n", hres); + ok(V_VT(&res) == VT_BSTR, "got %d\n", V_VT(&res)); + ok(V_BSTR(&res) == NULL || !lstrcmpW(V_BSTR(&res), L""), "got %ls\n", V_BSTR(&res)); } static WCHAR *create_test_typelib(int res_no) diff --git a/dlls/oleaut32/typelib.c b/dlls/oleaut32/typelib.c index c4f00ca5bea..14a7baec20e 100644 --- a/dlls/oleaut32/typelib.c +++ b/dlls/oleaut32/typelib.c @@ -7326,7 +7326,7 @@ static HRESULT WINAPI ITypeInfo_fnInvoke( } else if ((rgvt[i] & VT_BYREF) && !V_ISBYREF(src_arg)) { - if (wParamFlags & PARAMFLAG_FIN) + if ((wParamFlags & PARAMFLAG_FIN) || !(wParamFlags & (PARAMFLAG_FIN|PARAMFLAG_FOUT))) hres = VariantChangeType(&missing_arg[i], src_arg, 0, rgvt[i] & ~VT_BYREF); else V_VT(&missing_arg[i]) = rgvt[i] & ~VT_BYREF; -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/9639
From: Ivan Lyugaev <valy@etersoft.ru> Some COM applications register version-independent ProgIDs that point to version-specific classes through the CurVer registry key. This implements the correct logic for handling such cases --- dlls/combase/combase.c | 64 ++++++++++++++++++++++++++++++++++++------ 1 file changed, 56 insertions(+), 8 deletions(-) diff --git a/dlls/combase/combase.c b/dlls/combase/combase.c index 34a9a42dc2a..d1284be885c 100644 --- a/dlls/combase/combase.c +++ b/dlls/combase/combase.c @@ -1454,22 +1454,70 @@ static HRESULT clsid_from_string_reg(LPCOLESTR progid, CLSID *clsid) lstrcpyW(buf, progid); lstrcatW(buf, L"\\CLSID"); - if (open_classes_key(HKEY_CLASSES_ROOT, buf, MAXIMUM_ALLOWED, &xhkey)) + if (!open_classes_key(HKEY_CLASSES_ROOT, buf, MAXIMUM_ALLOWED, &xhkey)) { - free(buf); + if (!RegQueryValueW(xhkey, NULL, buf2, &buf2len)) + { + RegCloseKey(xhkey); + free(buf); + return guid_from_string(buf2, clsid) ? S_OK : CO_E_CLASSSTRING; + } WARN("couldn't open key for ProgID %s\n", debugstr_w(progid)); - return CO_E_CLASSSTRING; + RegCloseKey(xhkey); + xhkey = NULL; } free(buf); - if (RegQueryValueW(xhkey, NULL, buf2, &buf2len)) + buf = malloc((lstrlenW(progid) + 8) * sizeof(WCHAR)); + if (!buf) return E_OUTOFMEMORY; + lstrcpyW(buf, progid); + lstrcatW(buf, L"\\CurVer"); + + if (!open_classes_key(HKEY_CLASSES_ROOT, buf, MAXIMUM_ALLOWED, &xhkey)) { - RegCloseKey(xhkey); - WARN("couldn't query clsid value for ProgID %s\n", debugstr_w(progid)); - return CO_E_CLASSSTRING; + LONG curverlen = 0; + if (!RegQueryValueW(xhkey, NULL, NULL, &curverlen)) + { + WCHAR *curver = malloc(curverlen * sizeof(WCHAR)); + if (!curver) + { + free(buf); + RegCloseKey(xhkey); + return E_OUTOFMEMORY; + } + if (!RegQueryValueW(xhkey, NULL, curver, &curverlen)) + { + RegCloseKey(xhkey); + free(buf); + + buf = malloc((lstrlenW(curver) + 8) * sizeof(WCHAR)); + if (!buf) + { + free(curver); + return E_OUTOFMEMORY; + } + + lstrcpyW(buf, curver); + lstrcatW(buf, L"\\CLSID"); + if(!open_classes_key(HKEY_CLASSES_ROOT, buf, MAXIMUM_ALLOWED, &xhkey)) + { + if (RegQueryValueW(xhkey, NULL, buf2, &buf2len)) + { + RegCloseKey(xhkey); + WARN("couldn't query clsid value for ProgID %s\n", debugstr_w(progid)); + return CO_E_CLASSSTRING; + } + free(buf); + free(curver); + RegCloseKey(xhkey); + return guid_from_string(buf2, clsid) ? S_OK : CO_E_CLASSSTRING; + } + } + } } + if (buf) free(buf); RegCloseKey(xhkey); - return guid_from_string(buf2, clsid) ? S_OK : CO_E_CLASSSTRING; + return CO_E_CLASSSTRING; } /****************************************************************************** -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/9639
On Wed Jan 28 08:24:20 2026 +0000, Nikolay Sivov wrote:
Both commits need work in my opinion. I already suggested what else to check for the typelib change. I tested this check variant !(wParamFlags & (FIN|FOUT)) in PowerPnt, and the add-in opened without any issues. Since the PARAMFLAG_FRETVAL and PARAMFLAG_FLCID flags are handled elsewhere, your proposed !(wParamFlags & (FIN|FOUT)) variant shouldn't cause any confusion.
During my earlier testing, I noticed that when flags are absent (PARAMFLAG_NONE) in Windows, the parameter is treated as PARAMFLAG_FIN. That's why I don't understand why you don't like the variant with the explicit check (wParamFlags == PARAMFLAG_NONE). 🤔 -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9639#note_128379
On Thu Jan 29 15:34:30 2026 +0000, Ivan Lyugaev wrote:
I tested this check variant !(wParamFlags & (FIN|FOUT)) in PowerPnt, and the add-in opened without any issues. Since the PARAMFLAG_FRETVAL and PARAMFLAG_FLCID flags are handled elsewhere, your proposed !(wParamFlags & (FIN|FOUT)) variant shouldn't cause any confusion. During my earlier testing, I noticed that when flags are absent (PARAMFLAG_NONE) in Windows, the parameter is treated as PARAMFLAG_FIN. That's why I don't understand why you don't like the variant with the explicit check (wParamFlags == PARAMFLAG_NONE). 🤔 Hi, Nikolay!
Could you please advise what you would suggest changing next to make this solution completely finished? Based on the tests I've added, it's clear that this solution works correctly in Windows, and it also solves my issue with the add-in. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9639#note_129904
On Wed Feb 18 16:24:32 2026 +0000, Ivan Lyugaev wrote:
Hi, Nikolay! Could you please advise what you would suggest changing next to make this solution completely finished? Based on the tests I've added, it's clear that this solution works correctly in Windows, and it also solves my issue with the add-in. Hi, Nikolay!
Are you planning a further review of my decision? -- https://gitlab.winehq.org/wine/wine/-/merge_requests/9639#note_132291
participants (2)
-
Ivan Lyugaev -
Ivan Lyugaev (＠valy)