[PATCH 1/5] msi/tests: Add tests for external UI callback.

[Zebediah Figura]

Ah, I see. I'll fix these and the other patches.

On 06/26/2017 09:37 AM, Hans Leidekker wrote:
> On Mon, 2017-06-26 at 09:17 -0500, Z Figura wrote:
>> On 06/26/2017 04:07 AM, Hans Leidekker wrote:
>>> On Sun, 2017-06-25 at 21:13 -0500, Zebediah Figura wrote:
>>>> @@ -822,6 +823,8 @@ static UINT package_from_db(MSIHANDLE hdb, MSIHANDLE *handle)
>>>>        CHAR szPackage[12];
>>>>        MSIHANDLE hPackage;
>>>>    
>>>> +    CoInitialize(NULL);
>>>> +
>>>>        sprintf(szPackage, "#%u", hdb);
>>>>        res = MsiOpenPackageA(szPackage, &hPackage);
>>>>        if (res != ERROR_SUCCESS)
>>>> @@ -837,22 +840,39 @@ static UINT package_from_db(MSIHANDLE hdb, MSIHANDLE *handle)
>>>>            return res;
>>>>        }
>>>>    
>>>> +    CoUninitialize();
>>>> +
>>>
>>> Why do you need this?
>>
>> The message tests fail on Windows > 8 if COM is not initialized. This is
>> as per MSDN:
>>
>> "Note  Initialize COM on the same thread before calling the
>> MsiOpenPackage, MsiOpenPackageEx, or MsiOpenProduct function."
> 
> Then it would be better to wrap them only for the message tests.
>   
>>>> -static void create_test_file(const CHAR *name)
>>>> +static void create_file_data(LPCSTR name, LPCSTR data, DWORD size)
>>>>    {
>>>>        HANDLE file;
>>>>        DWORD written;
>>>>    
>>>>        file = CreateFileA(name, GENERIC_WRITE, 0, NULL, CREATE_ALWAYS, 0, NULL);
>>>>        ok(file != INVALID_HANDLE_VALUE, "Failure to open file %s\n", name);
>>>> -    WriteFile(file, name, strlen(name), &written, NULL);
>>>> +    if (file == INVALID_HANDLE_VALUE)
>>>> +        return;
>>>> +
>>>> +    WriteFile(file, data, strlen(data), &written, NULL);
>>>>        WriteFile(file, "\n", strlen("\n"), &written, NULL);
>>>> +
>>>> +    if (size)
>>>> +    {
>>>> +        SetFilePointer(file, size, NULL, FILE_BEGIN);
>>>> +        SetEndOfFile(file);
>>>> +    }
>>>> +
>>>>        CloseHandle(file);
>>>>    }
>>>
>>> You're not using this in this patch set.
>>>
>>>
>>
>> I use it below (line 261 of the patch) to set the codepage, which I do
>> in order to test the correct setting of field 3 of the
>> INSTALLMESSAGE_COMMONDATA messages.
> 
> You're calling it with size 0, so you're not using the code to explicitly
> set the file size.
> 
>