[Bug 28741] New: Odd coordinate in hit test causes uninitialized value references in MONTHCAL_CalculateDayOfWeek()?

http://bugs.winehq.org/show_bug.cgi?id=28741

Bug #: 28741
           Summary: Odd coordinate in hit test causes uninitialized value
                    references in MONTHCAL_CalculateDayOfWeek()?
           Product: Wine
           Version: 1.3.30
          Platform: x86
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: comctl32
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
    Classification: Unclassified

Running "make monthcal.ok" in comctl32/tests with Valgrind reports:

Conditional jump or move depends on uninitialised value(s)
   at __moddi3 (in dlls/ntdll/ntdll.dll.so)
   by RtlTimeToTimeFields (time.c:126)
   by FileTimeToSystemTime (time.c:858)
   by MONTHCAL_CalculateDayOfWeek (monthcal.c:469)
   by MONTHCAL_GetMonth (monthcal.c:486)
   by MONTHCAL_HitTest (monthcal.c:498)
   by MONTHCAL_WindowProc (monthcal.c:2861)
   by ??? (in /oldhome/dank/wine-git/dlls/user32/user32.dll.so)
   by call_window_proc (winproc.c:242)
   by WINPROC_CallProcAtoW (winproc.c:601)
   by CallWindowProcA (winproc.c:960)
   by monthcal_subclass_proc (monthcal.c:566)
   by ??? (in dlls/user32/user32.dll.so)
   by call_window_proc (winproc.c:242)
   by WINPROC_call_window (winproc.c:908)
   by call_window_proc (message.c:2211)
   by send_message (message.c:3084)
   by SendMessageA (message.c:3286)
   by test_hittest (monthcal.c:1042)
   by func_monthcal (monthcal.c:1957)
   by run_test (test.h:556)
   by main (test.h:624)
 Uninitialised value was created by a stack allocation
   at MONTHCAL_CalculateDayOfWeek (monthcal.c:462)

Something seems to be creating an out of range date; then when
MONTHCAL_CalculateDayOfWeek() converts to (and then back from)
a different time representation, the conversion fails, leading to
use of garbage.

My guess is that test_hittest() is picking some area to hittest that
is driving the code bonkers.  We get lucky, and the test passes anyway.
I don't know if a real hittest could give the wrong answer in some
situations.