Crashes in trunk

Hi all,
  I'm giving trunk a spin, there's a crash that hit me twice in a few
hours (running within gdb, single process, aufs)

Program received signal SIGPIPE, Broken pipe.
0x0012e416 in __kernel_vsyscall ()

stack trace:
#1 0x001454ab in write () at ../sysdeps/unix/syscall-template.S:82
#2 0x0825d488 in Comm::HandleWrite (fd=72, data=0x8451da8) at Write.cc:91
#3 0x08259ab5 in Comm::DoSelect (msec=<value optimized out>)
    at ModEpoll.cc:314
#4 0x0820eb1d in CommSelectEngine::checkEvents (this=0xbffff2b8, timeout=137)
        at comm.cc:2126
#5 0x08118b47 in EventLoop::checkEngine (this=0xbffff274, engine=0xbffff2b8,
            primary=true) at EventLoop.cc:51
#6 0x08118de8 in EventLoop::runOnce (this=0xbffff274) at EventLoop.cc:125
#7 0x08118ee0 in EventLoop::run (this=0xbffff274) at EventLoop.cc:95
#8 0x08175570 in SquidMain (argc=2, argv=0xbffff3b4) at main.cc:1506
#9 0x0817634c in SquidMainSafe (argc=2, argv=0xbffff3b4) at main.cc:1239
#10 main (argc=2, argv=0xbffff3b4) at main.cc:1231

where fd_table[72] is
.type = 3
.remote_port = 3861
.sock_family = 2
.desc = "Reading next request\000"
.flags = { open = 1, close_request = 0, write_daemon = 0, socket_eof =
0, nolinger = 0, nonblocking = 1, ipc = 0, called_connect = 0, nodelay
= 0, close_on_exec = 1, read_pending = 0, write_pending = 0,
transparent = 0}
.bytes_read = 10017
.bytes_written = 60149
.pconn = {uses = 0, pool = 0x0}
.epoll_state = 29
.read_handler = 0
.write_handler = 0

Hints?

-- 
    /kinkie