[PATCH v2 1/8] kernel32/tests: Fix NtMapViewOfSection tests results for w1064v1809

[Huw Davies]

On Tue, Jun 18, 2019 at 12:34:07PM +0200, Rémi Bernon wrote:
> On Mon, 2019-06-17 at 10:39 +0100, Huw Davies wrote:
> > On Fri, Jun 14, 2019 at 03:11:10PM +0200, Rémi Bernon wrote:
> > > @@ -1328,25 +1329,30 @@ static void test_NtMapViewOfSection(void)
> > >      size = 0;
> > >      offset.QuadPart = 0;
> > >      status = pNtMapViewOfSection( mapping, hProcess, &ptr2, 12, 0,
> > > &offset, &size, 1, 0, PAGE_READWRITE );
> > 
> > Also, changing this and the other non-zero zero_bits below to '1'
> > would improve the tests.  i.e. it's something that one would expect
> > to
> > succeed if ptr2 was set to NULL.
> 
> After doing some testing it looks like it would change the test results
> as well. The comment here is wrong in the sense that it looks like
> Windows returns INVALID_PARAMETER whenever the fixed address parameter
> does not satisfy the provided zero_bits constraint. Whenever it does,
> then the tests fail with MAPPED_ALIGNMENT instead.
> 
> So with 12 or more zero bits, there was little chance for the provided
> address to ever satisfy the check, and INVALID_PARAMETER was returned.
> If we change it to 1 zero bit, then the tests will usually succeed.
> 
> This also explains the "broken" behaviour that was sometimes seen on
> w1064v1809.
> 
> This first patch was merely to add results seen on w1064v1809, and I
> didn't want to change test logic before moving them to ntdll. I can do
> the logic changes before moving the tests, but I don't think doing
> logic changes both before and after moving them is very clear.

The idea is to fix the tests before moving them.  So yes, making these
changes first makes sense.

Huw.