I still have a problem if the normal operation of my server (which
tests are simulating) requires more file descriptors than I am
given. From my perspective OSX is making it harder for me to get
what I want by giving child processes a smaller file handle limit
than the soft limit I have defined in the parent shell (e.g., by
setting default to MAX_OPEN). I'm not clear what purpose this serves.
If the soft limit is reachable (i.e. at or below hard-limit and
maxfilesperprocess and maxfiles), then the soft limit should be
inherited by the shell's child process.
If the soft limit is unreachable, i.e. above hard-limit or
maxfilesperprocess or maxfiles, then the soft limit will be capped at
the max reachable value.
I have never observed anything different from Mac OS X, using either
the shell's ulimit or using other means (e.g. child processes spawned
after Suite/P code changes hard or soft limits).
If you have a fail-case, please provide specific steps to reproduce
it, including any prior changes made to /etc/sysctl.conf.
Just because a hard-limit returns "unlimited" does not mean it really
is unlimited.
Also, the 'ulimit' builtin changes both hard and soft limits unless
told otherwise.
Finally, hard limits for a process can only be decreased, not raised,
unless the process is running as root. See 'man setrlimit' and find
the word "irreversibly". Remember this when using 'ulimit' without -
S.
-- GG
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Java-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/java-dev/email@hidden
This email sent to email@hidden
