Here's what a truss shows when pen is taking up the whole CPU:
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
accept(5, 0xFFBEFAB8, 0xFFBEFAC8, 1) Err#24 EMFILE
poll(0xFFBEF0A0, 251, -1) = 227
...
repeated as fast as can be. What's that mean?
Stopping pen and restarting cleared it.
This is version 10.1 on Sun Sparc/Solaris 8 compiled with Sun Workshop 6
update 2. I had dropped down to version 0.10.1 because I thought the
problem started when I switched to version 0.12.0. The semi-related
email slowdown (I'm load balancing a webmail frontend) was due to a bad
disk in a RAID5 array.
This archive was generated by hypermail 2.1.2 : Mon Nov 17 2003 - 21:21:38 CET