[Bug 58293] New: Regression caused by switch to realloc which doesn't zero added memory

[WineHQ Bugzilla]

http://bugs.winehq.org/show_bug.cgi?id=58293

Bug ID: 58293
           Summary: Regression caused by switch to realloc which doesn't
                    zero added memory
           Product: Wine
           Version: 10.7
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: wintrust
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 78644
  --> http://bugs.winehq.org/attachment.cgi?id=78644
Log showing run with and without reverting 6db69d0 under wine 10.7

In wine 8 `LibOVR.dll` (the shim) would load `LibOVRRT32_1.dll` (the Oculus
API). In wine 9 and later this is no longer the case. Reverting 6db69d0
(wintrust: Use CRT allocation function) makes it work again.

Looking over this commit I expect the issue is the switch from
```
static void* WINTRUST_ReAlloc(void *ptr, DWORD cb) __WINE_ALLOC_SIZE(2);
static void* WINTRUST_ReAlloc(void *ptr, DWORD cb)
{
    return HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, ptr, cb);
}
```
which zeros added memory to realloc which does not. Presumably this is
resulting in some field being populated with garbage, and that makes
`LibOVR.dll` unhappy when it verifies the signatures on `LibOVRRT32_1.dll`.

I put more details and a test program in the following github repo

https://github.com/twhitehead/issue-wine-ovrf