(a few weeks ago Havard said this...)

> There are (were?) some problems concerning
> dangling pointers to tty structures in the process structure whenever 
> the session leader goes away.  

> So it appears to be fixed in NetBSD-current.

It's worked around, by not free'ing the malloc'd tty structures at
XXclose() time. Something seems to still be using them. This caused
panics during 'w'.

For what it's worth though, yesterday I had a 'w' go into an infinite
loop (lseek some address, read a few bytes, lseek same address, read
same number of bytes, repeat forever..) but we were not able to debug
it further. The kvm library and w is still doing something wrong.

But the problem is a more generic tty subsystem lossage. At least
cgd, Andrew, and I have noticed some really weird and bad things
happen though on pty's....
 <tdr.

------------------------------------------------------------------------------