malloc patch for 2.2.4

[Wolfram Gloger]

Hello,

> Would you mind explaining how this improves thread safety?

Jakub explained the issue nicely.  I actually saw the NULL-setting
problem on Solaris and it doesn't happen any more with the patch
installed.

> And in fact, I can see it still being
> thread-unsafe if copying a function pointer is non-atomic.  Isn't that
> the case on, say, ia64?

If that would be the case, _many_ other things would break as well.

Regards,
Wolfram.