[Bug 47875] New: kernel32:file: FindFirstFile("foo\nul\bar") succeeds on AMD graphics!

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=47875

Bug ID: 47875
           Summary: kernel32:file: FindFirstFile("foo\nul\bar") succeeds
                    on AMD graphics!
           Product: Wine
           Version: 4.17
          Hardware: x86
                OS: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: kernel32
          Assignee: [e-mail]
          Reporter: [e-mail]

On some machines kernel32:file systematically gets the following two failures:

file.c:2709: Test failed: FindFirstFile on C:\foo\bar\nul should fail
file.c:2710: Test failed: Bad Error number -559038801

-559038801 is 0xdeadbeaf which is the value LastError is set to before the
test:
     SetLastError( 0xdeadbeaf );
     strcpy(buffer2, buffer);
     strcat(buffer2, "foo\bar\nul");
     handle = FindFirstFileA(buffer2, &data);
     ok ( handle == INVALID_HANDLE_VALUE, "FindFirstFile on %s should fail\n",
buffer2 );
     ok ( err == ERROR_PATH_NOT_FOUND, "Bad Error number %d\n", err );

This suggests that FindFirstFile() succeeds in finding "foo\bar\nul".

Furthermore this only happens on two Windows 8.1 machines, cw-rx460 and
dr-RS740, both in the 32 and 64 bit tests.

They share the same build number (9600) but so does cw2-gtx560 which does not
have these failures. The other common point is that they both have an AMD
graphics card and, presumably, the corresponding driver. But why would a
graphics driver change the behavior of FindFirstFile()?