[Bug 40384] New: dinput8:dinput regression caused by new compiler

https://bugs.winehq.org/show_bug.cgi?id=40384

Bug ID: 40384
           Summary: dinput8:dinput regression caused by new compiler
           Product: Wine
           Version: 1.9.6
          Hardware: x86
                OS: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: directx-dinput
          Assignee: [e-mail]
          Reporter: [e-mail]

The WineHQ upgrade changed the cross-compiler used to generate the reference
WineTest executables. Since the upgrade the dinput8:dinput test has these
failures on all 32 bit Windows platforms:

dinput.c:519: Test failed: EnumDevicesBySemantics did not call the callback
hr=80070057
dinput.c:520: Test failed: EnumDevicesBySemantics should enumerate the keyboard
dinput.c:521: Test failed: EnumDevicesBySemantics should enumerate the mouse
dinput.c:527: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:535: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:536: Test failed: There should be devices available before action
mapping available=0
dinput.c:543: Test failed: EnumDevicesBySemantics failed: hr=80070057
dinput.c:548: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:554: Test failed: EnumDevicesBySemantics failed: hr=80070057
dinput.c:559: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:565: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:571: Test failed: EnumDevicesBySemantics failed hr=80070057
dinput.c:577: Test failed: EnumDevicesBySemantics failed hr=80070057

See:
https://test.winehq.org/data/tests/dinput8:dinput.html
https://www.winehq.org/pipermail/wine-devel/2016-March/112371.html

I can reproduce the problem with this compiler:
$ i686-w64-mingw32-gcc --version
i686-w64-mingw32-gcc (GCC) 5.3.1 20160205

But if I recompile the tests with -O0 the failures go away which points to a
compiler bug.
We need to either:
* Switch to a non buggy compiler to build the reference WineTest executables.
  (which might involve reporting the bug in the first place)

* Add 'CFLAGS = -g -O0' to the test's Makefile.in as a workaround.
  (preferably once we can point to a relevant compiler bug)

* Find some other nicer workaround in the C file.