On Sun, 20 Aug 2000, James Cameron wrote:
> I have not measured the rate of process memory usage yet against either
> time or connections made. I suspect it is unrelated to the number of
> simultaneous connections.
I checked the memory usage of netrekd with "ps -v", let a shell loop run
telnets to port 2591 for a while, and then checked again. Exactly the same
numbers, no memory leak. This is with glibc-2.1.3 on intel Linux. Does doing
this same thing cause a memory usage increase on your sparc/linux system?
> > Possibly. I've looked over the newstartd code, and it's pretty nasty. I
> > already spotted one bug in a select() call, but I doubt that is the problem.
>
> It was far worse until I hacked at it last year.
> Let me know what that bug is.
I'll just put it in CVS after I've fixed a few other things.
> {port = 0, sock = 0,
> nargs = 0, prog = "special", '\000' <repeats 504 times>,
> progname = '\000' <repeats 511 times>, arg = {'\000' <repeats 63 times>,
> '\000' <repeats 63 times>, '\000' <repeats 63 times>,
> '\000' <repeats 63 times>}, internal = 0, accepts = 0, denials = 0,
> forks = 0, addr = 0},
This one doesn't seem right to me. The port is set to 0, the port in the
default .ports file is 2596. Did you change it to 0? sock is also 0, which
shouldn't happen. It should be -1 if there is no listening socket, or >0 if
there is. progname should also be "statistics", and internal should be 1.