Fix for comparisons in netapi32/tests

[Vincent Béron]

Two comparisons were always true due to the types. Not sure about the 
proper fix though.

Changelog:
Changed comparisons so they're not always true.

Vincent

Index: wine/dlls/netapi32/tests/apibuf.c
===================================================================
RCS file: /home/wine/wine/dlls/netapi32/tests/apibuf.c,v
retrieving revision 1.1
diff -u -r1.1 apibuf.c
--- wine/dlls/netapi32/tests/apibuf.c	11 Sep 2002 02:35:18 -0000	1.1
+++ wine/dlls/netapi32/tests/apibuf.c	10 Dec 2002 17:25:20 -0000
@@ -48,14 +48,14 @@
     ok(NetApiBufferFree(p) == NERR_Success, "Freed");

ok(NetApiBufferSize(p, &dwSize) == NERR_Success, "Got size");
-    ok(dwSize >= 0, "The size");
+    ok(dwSize < 0x80000000, "The size");
     ok(NetApiBufferSize(NULL, &dwSize) == ERROR_INVALID_PARAMETER, "Error for NULL pointer");

/* 0-length buffer */
     ok(NetApiBufferAllocate(0, (LPVOID *)&p) == NERR_Success,
        "Reserved memory");
     ok(NetApiBufferSize(p, &dwSize) == NERR_Success, "Got size");
-    ok((dwSize >= 0) && (dwSize < 0xFFFFFFFF),"The size of the 0-length buffer");
+    ok(dwSize < 0xFFFFFFFF, "The size of the 0-length buffer");
     ok(NetApiBufferFree(p) == NERR_Success, "Freed");
 }

[Francois Gouget]

ok(NetApiBufferSize(p, &dwSize) == NERR_Success, "Got size");
-    ok(dwSize >= 0, "The size");
+    ok(dwSize < 0x80000000, "The size");
[...]
     ok(NetApiBufferAllocate(0, (LPVOID *)&p) == NERR_Success,
        "Reserved memory");
     ok(NetApiBufferSize(p, &dwSize) == NERR_Success, "Got size");
-    ok((dwSize >= 0) && (dwSize < 0xFFFFFFFF),"The size of the 0-length
buffer");
+    ok(dwSize < 0xFFFFFFFF, "The size of the 0-length buffer");

I think this test does not make sense. Why would 0xfffffffe be
considered valid and not 0xffffffff? The truth is we have no idea what
value to expect from a call to NetApiBufferAllocate(0,...) and thus
there is nothing to test.

The < 0x80000000 test just above seems just as arbitrary (just a better
approximation of 'if this were a signed int would it be positive?').

So I vote to just remove both tests, or, for the first case, to actually
find out what value we should expect (but I doubt this is documented).