[1/2] imagehlp/tests: Add test for BindImageEx (try 12)

[Bernhard Reiter]

Am 2014-07-24 um 18:33 schrieb Alexandre Julliard:
> Bernhard Reiter <ockham at raz.or.at> writes:
> 
>> @@ -31,6 +31,8 @@
>>  static HMODULE hImageHlp;
>>  
>>  static BOOL (WINAPI *pImageGetDigestStream)(HANDLE, DWORD, DIGEST_FUNCTION, DIGEST_HANDLE);
>> +static BOOL (WINAPI *pBindImageEx)(DWORD Flags, const char *ImageName, const char *DllPath,
>> +                                   const char *SymbolPath, void *StatusRoutine);
> 
> That's not the correct prototype.

Meaning I should change the const char*s to char*s (as MSDN says PSTR)?
I was looking at the BindImageEx stub in dlls/imagehlp/modify.c, and
that had PCSTRs.

(Or do you mean using PIMAGEHLP_STATUS_ROUTINE instead of void* for
StatusRoutine? Because that would require me to typecast
testing_status_routine to that when invoking pBindImageEx, wouldn't it?)

>> +static BOOL WINAPI testing_status_routine(IMAGEHLP_STATUS_REASON reason, char *ImageName,
>> +                                          char *DllName, ULONG_PTR Va, ULONG_PTR Parameter)
>> +{
>> +    char kernel32_path[MAX_PATH];
>> +
>> +    status_routine_called[reason]++;
> 
> This will have nasty results if you get an unexpected code.

Okay, I'll add checks if reason is in the expected range.

> 
>> +        default:
>> +            ok(0, "expected reason to be one of BindImportModule, BindImportProcedure, or "
>> +               "BindForwarderNOT, got %d\n", reason);
> 
> Please simplify your error messages. In particular, this one would need
> to be changed every time the test is expanded, that's not a good idea.

I'm not exactly creative how to change this... would

	ok(0, "testing_status_routine got wrong reason %d\n", reason);

be okay?

>> +static void test_bind_image_ex(void)
>> +{
>> +    BOOL ret;
>> +    HANDLE file;
>> +    char temp_file[MAX_PATH] = "nonexistant.dll";
> 
> There's no reason to use a variable for this.

But I need it later to retrieve the name of the file created by
create_temp_file. Plus, if I change the const char*s in pBindImageEx's
declaration to char*s, I will run into problems here if I pass a const
char* to pBindImageEx, won't I?

So is it okay to leave it as it is or am I overlooking something?

>> +    /* Under 64 bit images of Vista Ultimate, 2008 Server, and 7 Pro, StatusRoutine is called with
>> +     * reason BindForwarderNOT instead of BindImportProcedure. */
>> +    todo_wine ok((status_routine_called[BindImportProcedure] == 1) ||
>> +                 (status_routine_called[BindForwarderNOT] == 1),
>> +                 "Expected StatusRoutine to be called once with reason BindImportProcedure or "
>> +                 "BindForwarderNOT, but it was called %d and %d times, respectively\n",
>> +                 status_routine_called[BindImportProcedure], status_routine_called[BindForwarderNOT]);
> 
> You'd most likely want the test to depend on 32/64-bit instead of always
> accepting both.

I had #if defined(_WIN64) macros there earlier, but the behavior isn't
consistent: Windows 8 gives BindImportProcedure also with the 64 bit
image. (I tried it with test job 7953, if that can be retrieved somehow.
Could that BindForwarderNOT be due to a bug in WoW that was fixed in
Windows 8?) I asked on #winehackers if I should add a check for the
Windows version, but people suggested not to.
What's your stance on this?

Bernhard