Two fixup routines in fork_child seem misplaced (Corinna, Robert)

[Christopher Faylor]

There are two fixup routines in fork_child which seem to be too late in
fork process to be reliable: fixup_shms_after_fork and fixup_mmaps_after_fork.

Both occur after the last synchronization with the parent but may need to
rely on the parent's state, if I am reading things correctly.

I think this accounts for the strange behavior that was mentioned in the
cygwin mailing list with mmap.

There is obviously no shm usage these days.  I didn't look closely enough to see
if the regions that are being mapped are coming from the parent.  Maybe they're
just coming from the server, now that I think about it.  That would make more
sense.

cgf