[Bug 20224] New: dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

Summary: dlls/d3dx9_36/tests/math.C: Test failed: Expected
                    matrix
           Product: Wine
           Version: 1.1.30
          Platform: PC-x86-64
               URL: http://test.winehq.org/data/f3b9fb554dd116c88142209873
                    8a1f0223087cd1/wine_ae-904-x64-nogecko/d3dx9_36:math.h
                    tml
        OS/Version: Linux
            Status: NEW
          Keywords: download, source, testcase, win64
          Severity: trivial
          Priority: P2
         Component: directx-d3d
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Normally, I wouldn't file bugs related to testcases failing on 64-bit, but Jörg
pointed out something interesting:

http://test.winehq.org/data/f3b9fb554dd116c881422098738a1f0223087cd1/wine_ae...
http://test.winehq.org/data/f3b9fb554dd116c881422098738a1f0223087cd1/wine_jc...

(both are winetest results from 1.1.30, my win64 tests, and his 32-bit OS X
tests)

Both fail in the same way:
dlls/d3dx9_36/tests/math.c
math.c:512: Test failed: Expected matrix=
(-0.214800,1.311600,0.475200,0.000000
0.950400,-0.883600,0.924400,0.000000
1.021200,0.193600,-1.358800,0.000000
18.298500,-29.624001,15.683499,1.000000
)
Got matrix=
(-0.214800,1.311600,0.475200,0.000000
0.950402,-0.883606,0.924399,0.000000
1.021202,0.193600,-1.358803,0.000000
18.298527,-29.624161,15.683442,1.000000)
math.c:756: Test marked todo: Expected Vector= (0.206945, 0.103473, 0.310418,
0.000000)
, Got Vector= (0.093768, 0.187536, 0.375073, 0.000000)
math: 364 tests executed (1 marked as todo, 1 failure), 0 skipped.

Since it also affects a 32-bit wine, I figured I should file a bug. Perhaps
driver related, and so only affects OS X/64-bit?

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

Jörg Höhle  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
                   |                            |t

--- Comment #1 from Jörg Höhle   2009-10-01 03:05:34 ---
> OS X/64-bit

? AFAIK, my Leopard machine is 32 bit. Snow Leopard went to 64bits, whatever
that means.  I believe 64bit does not make sense in any way on a Mac Mini that
Apple is said to have crippled to not allow more than 4GB RAM (mine has 1GB).
I'm using gcc-4.2 from the Leopard install DVD, not gcc-4.0.  gcc-4.2 is said
to have changed defaults from 32 to 64 bit in Snow Leopard -- not Leopard.
So until now, I've no indication that my machine should do anything with
64bits, but how could I tell?

Doesn't this look like a gross floating point rounding or propagation error?

I've bought the Snow Leopard upgrade, but a) I was too busy to install it so
far and b) before upgrading, I found it valuable to analyse all known bugs
before encountering new ones.

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

--- Comment #2 from Austin English   2009-10-01 03:10:40 ---
(In reply to comment #1)
> > OS X/64-bit

Sorry, meant 32-bit OS X and 64-bit Wine on Linux. 64-bit wine doesn't yet
build on OS X (I think).

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

joaopa  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #3 from joaopa   2009-10-14 19:06:53 ---
It looks like a floating point rounding error. One could release the condition
for equality, but is it useful for only one machine not working?

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

--- Comment #4 from Austin English   2009-10-14 19:19:12 ---
(In reply to comment #3)
> It looks like a floating point rounding error. One could release the condition
> for equality, but is it useful for only one machine not working?

It's not just one machine...64-bit wine has the failure (on my machine and
AJ's:http://test.winehq.org/data/e8ce3e600a40a739f75d6716aae555c31b0b4d0d/wine_aj...),
as well as OS X.

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

--- Comment #5 from Alexandre Julliard   2009-10-15 02:55:00 ---
It may not necessarily need to be relaxed, but the accepted difference should
be relative to the magnitude of the compared values. The code is hardcoding a
0.0001 difference and that is broken.

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #6 from Austin English   2009-12-15 11:31:57 ---
Appears fixed:
http://test.winehq.org/data/33a193439d03a06555e8517d371e27b12e64c00d/wine_ae...

by
http://source.winehq.org/git/wine.git/?a=commitdiff;h=3ff4d07c401d856fdf5926...

Thanks David!

[Bug 20224] dlls/d3dx9_36/tests/math.C: Test failed: Expected matrix

http://bugs.winehq.org/show_bug.cgi?id=20224

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #7 from Alexandre Julliard   2009-12-18 13:07:43 ---
Closing bugs fixed in 1.1.35.