Bug#866122: slapd-mtread crash on ppc64{,el} in stretch/sid

[Breno Leitao <brenohl@br.ibm.com>]

Some new discover I did today:

1) On function do_random(), the 'values' pointer is being corrupted after the
rand() syscall - In the failure case. If I remove the rand() function, I do
not see corruption.

2) If I get the original 'values' pointer, save it and check it later, it is
corrupted also. This guarantee that the there is no one else changing the
pointer. So, I suspect that this value is being corrupted during the context
switch or vsx unavailable tm (which is being called also).

This is the type of the pointer corruption I see:

>From 0x7fff80000970 to 0x7fff88000970.

Since the array pointer changes, the 'r' value will point to r + the
corruption displacement.