On Sun, May 5, 2013 at 2:38 AM, Frank Terbeck <ft@bewatermyfriend.org>wrote:

> I've pushed this, because it's definitely an improvement. I tried
> looking for the right position to unblock SIGWINCH for the `exec' case
> in exec.c, but couldn't find it yet. Maybe someone has a more specific
> idea than I do?
>

 As far as I can tell, there are six calls to execve() in Src/exec.c.  If a
winch_unblock() were placed just before each of those six calls, the
unblock in zfork() from that last patch could probably be reverted.