[Bug 21568] advapi32/crypt tests crash with +heap

http://bugs.winehq.org/show_bug.cgi?id=21568

--- Comment #1 from Juan Lang   2010-02-01 14:11:41 ---
It's hard to diagnose this, because your backtrace doesn't have debug symbols
for the test files.  Still, I'll take a guess that it's this one, in
test_incorrect_api_usage():

result = pCryptDestroyKey(hKey2);
    ok (result, "%d\n", GetLastError());

dwTemp = CRYPT_MODE_ECB;    
    result = pCryptSetKeyParam(hKey2, KP_MODE, (BYTE*)&dwTemp, sizeof(DWORD));
    ok (!result && GetLastError() == ERROR_INVALID_PARAMETER, "%d\n",
GetLastError());

That is, a destroyed key is than used.  The bad address is 0xfeeefeee, which
corresponds to the free filler (in dlls/ntdll/heap.c):
#define ARENA_FREE_FILLER      0xfeeefeee

Basically, I think this is a false positive.  An app was doing something bogus
in the first place, and we wrote a test that mimics its behavior.  I'm not sure
how to fix the tests to pass in this case.