Make dlls-kernel-tests-file.c work on Windows 95 and XP.

[Jakob Eriksson]

Alexandre Julliard wrote:

>Jakob Eriksson <jakov at vmlinux.org> writes:
>
>  
>
>>+static int test_win9x_based()
>>+{
>>+    OSVERSIONINFOEX ver;
>>+
>>+    ver.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
>>+    ok( GetVersionEx ((OSVERSIONINFO *) &ver), " --- Can't get OS version.\n" );
>>+    if ((4 == ver.dwMajorVersion && 0 == ver.dwMinorVersion && 1381 == ver.dwBuildNumber)
>>+        || ver.dwMajorVersion >= 5)
>>+    {
>>+        return 0;
>>+    }
>>+
>>+    return 1;
>>    
>>
>
>You should not do version checks in tests, the checks should be based
>on the features we are testing. Otherwise the results are going to
>depend on the Windows version that Wine is configured for, which is
>the wrong thing to do since the behavior of the Wine code in that case
>does not depend on the version.
>
>  
>

I just thought hard about winetest.exe... how about this approach:

if (running_under_wine || !win9x_based)
{
   /* Do stuff that would crash Windows 9x. */

   /* Could also be stuff that Wine has not yet implemented */
   todo_wine { stuff(); }
}



Pros:
* Under Wine, run every test.
* Easy skip of tests under platforms we know crashes or hangs winetest.exe.

* If I am allowed to check for platform, it is easier to avoid crashing
  winetest.exe. (I.e. how do I check for capabilities on something that 
hangs
  winetest.exe when probed for capabilities?)


Cons:
Could be easy to be lazy - for instance there can be different
versions of DLLs under Windows 98, some more capable than others.
This scheme would invite to not test for capabilities.



I just want do deploy winetest.exe on as many Win32 platforms as possible
without having the app crashing all the time... Maybe that desire is leading
me to cut corners where I shouldn't?



regards,
Jakob