[squid-users] squid 3.2.1 on solaris

Hello,

I'm migrating from squid 3.1.19 to 3.2.1 on a solaris 10 machine (Solaris 10
10/08 s10x_u6wos_07b X86).

It worked fine before. Under 3.2.1, after adjusting some config lines, squid is
dying each small number of minutes. I've increased the debug level and I have
the following at cache_log.

----------
2012/09/04 12:06:11.411 kid1| connReadWasError: FD local=194.214.158.222:8080
remote=192.168.128.116:2962 FD 176 flags=1: got flag -
1
2012/09/04 12:06:11.411 kid1| client_side.cc(764) swanSong:
local=194.214.158.222:8080 remote=192.168.128.116:2962 flags=1
2012/09/04 12:06:11.412 kid1| storeCreate: Selected dir 0 for -1_at_-1=0/2/1/1
FATAL: Received Segment Violation...dying.
2012/09/04 12:06:11.412 kid1| Closing HTTP port 0.0.0.0:8080
2012/09/04 12:06:11.412 kid1| Stop receiving ICP on 0.0.0.0:3130
2012/09/04 12:06:11.412 kid1| Stop sending ICP from 0.0.0.0:3130
2012/09/04 12:06:11.412 kid1| BUG: Orphan Comm::Connection: local=0.0.0.0:3130
remote=[::] FD 15 flags=9
2012/09/04 12:06:11.412 kid1| NOTE: 3 Orphans since last started.
2012/09/04 12:06:11.412 kid1| Not currently OK to rewrite swap log.
2012/09/04 12:06:11.412 kid1| storeDirWriteCleanLogs: Operation aborted.
CPU Usage: 22.028 seconds = 5.285 user + 16.743 sys
Maximum Resident Size: 0 KB
Page faults with physical i/o: 0
2012/09/04 12:06:14.542 kid1| main.cc(1413) SquidMain: Doing post-config
initialization

The cache filesystem is zfs, but it worked before, with the same cache_dir config :

cache_dir ufs /var/squid/cache 50000 64 512

This version is running on another machine, same hardware, same OS version, but
no cache.

Some hints ? Maybe I missed something ?

Thanks

--