I explained a couple of days ago that I could get a program to execute
correctly only after running it once under gdb.  Otherwise, I got
"Segmentation fault" errors without fail.

I have another piece of the puzzle: it seems that whatever is loading
the program from the NFS-mounted remote drive (kernel exec, I presume)
is failing, whereas loading the disk image (and running it, perhaps?)
under gdb seems to clear the path for further execution.

I got to this conclusion because the very same program fails once
again, after having run correctly a few hours ago.  My (uneducated)
guess is that gdb loads it correctly and somehow this creates a memory
image that eventually expires.  Would the UVM be involved in this?

++L