[Bug 20684] New: Memory leak in CryptGenKey/CryptDestroyKey?

http://bugs.winehq.org/show_bug.cgi?id=20684

Summary: Memory leak in CryptGenKey/CryptDestroyKey?
           Product: Wine
           Version: 1.1.32
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Keywords: download, source, testcase
          Severity: normal
          Priority: P2
         Component: rsaenh
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=24711)
 --> (http://bugs.winehq.org/attachment.cgi?id=24711)
minimal test showing problem

Chromium's unit_tests.exe's SignatureCreatorTest.BasicTest
seems to show a leak in CryptGenKey():

972 bytes in 1 blocks are definitely lost in loss record 836 of 921
   at RtlAllocateHeap (heap.c:1423)
   by ???
   by ???
   by ???
   by CryptGenKey (crypt.c:1434)
   by base::RSAPrivateKey::Create (rsa_private_key_win.cc:35)
   by SignatureCreatorTest_BasicTest_Test::TestBody
(signature_creator_unittest.cc:15)

I've extracted a minimal test case into the attached patch.  
To repeat, apply the patch and then run the rsaenh test under valgrind.
It complains

972 bytes in 1 blocks are definitely lost in loss record 463 of 518
   at RtlAllocateHeap (heap.c:1423)
   by new_object (handle.c:362)
   by new_key (rsaenh.c:834)
   by RSAENH_CPGenKey (rsaenh.c:3064)
   by CryptGenKey (crypt.c:1434)
   by func_rsaenh (rsaenh.c:2546)
   by run_test (test.h:535)
   by main (test.h:585)

It seems the reference count of the generated key starts off one
too high, or something?