Segment Violation during light load

Hello,

a few minutes ago I upgraded my squid-1.1.b7 to b8 on my linux box.
(ELF libc.so.5.2.18, Kernel 2.0.22, squid-1.1.b8 with the following options
in src/Makefile set: -DLOG_FULL_HEADERS=1
                     -DCACHEMGR_HOSTNAME="getfullhostname()"
                     -DCACHE_HTTP_PORT=8080)

Since then I got massive problems with the file descriptors, which just die
away
on a really light loaded machine. Even switching back to b7 didn't help, as the
error occurs now with this version, too.

Here is the debug output with debugging set to ALL,9. I hope I pasted only the
relevant parts...

------------------------ cache.log, ALL.9 ---------------------------------

User-Agent: GetURL v1.3.4
Accept: */*

---
96/10/22 01:10:33| parseHttpRequest: Complete request received
96/10/22 01:10:33| mime_get_header: looking for 'User-Agent'
96/10/22 01:10:33| mime_get_header: checking 'User-Agent: GetURL v1.3.4
Accept: */*
'
96/10/22 01:10:33| mime_get_header: returning 'GetURL v1.3.4'
...
96/10/22 01:10:33| aclMatchAclList: returning 1
96/10/22 01:10:33| aclCheck: match found, returning 1
96/10/22 01:10:33| clientAccessCheckDone: 'http://www.nln.com:80/' answer=1
96/10/22 01:10:33| clientRedirectDone: 'http://www.nln.com/' result=NULL
96/10/22 01:10:33| mime_get_header: looking for 'If-Modified-Since'
96/10/22 01:10:33| mime_get_header: looking for 'Pragma'
96/10/22 01:10:33| mime_get_header: looking for 'Authorization'
96/10/22 01:10:33| icpProcessRequest: GET <URL:http://www.nln.com/>
96/10/22 01:10:33| icpProcessRequest: REQ_NOCACHE = NOT SET
96/10/22 01:10:33| icpProcessRequest: REQ_CACHABLE = SET
96/10/22 01:10:33| icpProcessRequest: REQ_HIERARCHICAL = SET
96/10/22 01:10:33| storeGeneratePublicKey: type=1 http://www.nln.com/
96/10/22 01:10:33| storeGet: looking up http://www.nln.com/
96/10/22 01:10:33| storeLockObject: key 'http://www.nln.com/' count=1
96/10/22 01:10:33| storeLockObject: OK: swap_status is SWAP_OK
96/10/22 01:10:33| new_MemObject: returning 0x82a4400
96/10/22 01:10:33| storeSwapInStart: initialized swap file '/usr/local/squid/cache/0E/36/0000036E' for <URL:http://www.nln.com/>
96/10/22 01:10:33| new_MemObjectData: calling memInit()
96/10/22 01:10:33| icpProcessRequest: TCP_HIT for 'http://www.nln.com/'
96/10/22 01:10:33| icpSendMoreData: <URL:http://www.nln.com/> sz 18291: len 0: off 0.
96/10/22 01:10:33| storeCopy: avail_to_write=0, store_offset=0
96/10/22 01:10:33| comm_write: FD 20: sz 0: tout 30: hndl 0x805b730: data 0x8103600.
96/10/22 01:10:33| comm_select: 2 sockets ready at 845939433
96/10/22 01:10:33| comm_select: FD 20 ready for writing
96/10/22 01:10:33| commHandleWrite: FD 20: state=0x8130dc0, off 0, sz 0.
96/10/22 01:10:33| icpHandleStoreComplete: FD 20: sz 0: err 0: off 0: len 18291: tsmp 845509140: lref 845939433.
96/10/22 01:10:33| storeRegister: FD 20 'http://www.nln.com/'
96/10/22 01:10:33| storeRegister: grew pending list to 1 for slot 0.
96/10/22 01:10:33| comm_select: FD 23 ready for reading
96/10/22 01:10:33| storeSwapInHandle: 'http://www.nln.com/'
96/10/22 01:10:33| storeSwapInHandle: e->swap_offset   = 0
96/10/22 01:10:33| storeSwapInHandle: len              = 8192
96/10/22 01:10:33| storeSwapInHandle: e->e_current_len = 0
96/10/22 01:10:33| storeSwapInHandle: e->object_len    = 18291
96/10/22 01:10:33| storeAppend: appending 8192 bytes for 'http://www.nln.com/'
96/10/22 01:10:33| memAppend: len 8192
96/10/22 01:10:33| icpHandleStore: FD 20: off 0: <URL:http://www.nln.com/>
96/10/22 01:10:33| icpSendMoreData: <URL:http://www.nln.com/> sz 18291: len 8192: off 0.
96/10/22 01:10:33| storeCopy: avail_to_write=4096, store_offset=0
96/10/22 01:10:33| memCopy: offset 0: size 4096
FATAL: Received Segment Violation...dying.
96/10/22 01:10:33| storeWriteCleanLog: Starting...
96/10/22 01:10:34|      4096 lines written so far.
96/10/22 01:10:34| comm_set_fd_lifetime: FD 19 lft -1
96/10/22 01:10:34|   Finished.  Wrote 6151 lines.
96/10/22 01:10:34|   Took 1 seconds (6151.0 lines/sec).
96/10/22 01:10:34| comm_set_fd_lifetime: FD 24 lft -1
CPU Usage: user 10 sys 17                                      
Memory Usage: rss 0 KB
Page faults with physical i/o: 302
96/10/22 01:10:34| sig_child: Ate pid 0
------------------------ end of cache.log ------------------------------------
No matter what browser I use (ns, "geturl ~/.bookmarks.html") a few seconds 
later squid dies and restarts again.
There is possibly a relationship between my L2 cache upgrade and squid. But
memtest says my memory is ok.
Any hint appreciated.
BYe Daniel
---
EMail: Daniel.Kuckuck@Muenchen.ORG
       Munich, Germany