handling borken pipes

So, I've got my server up and running nicely, but when the clients computer
crashes without finishing the connection, leaving me with a broken pipe, it
shuts down my server.
What's the best way to handle this? I'm tempted just to throw in an ignore
signal handle (signal(SIGPIPE, SIG_IGN) of the top of my head) and pay no
attention to broken pipes, but I don't know enough to be sure this is always
correct. For instance, I don't get why the signal is fatal by default, as it
seems relatively minor.


.

Relevant Pages

  • Re: "broken pipe" while reading/writing stream-based sockets
    ... > 1) Why does this server sometimes crash with "Broken Pipe"? ... connection while the server was still sending data. ... happens because your code does not deal with partial reads/writes. ...
    (comp.lang.ada)
  • Cluster Broken pipe & node Reboot
    ... I am running two node RHEL3U8 cluster of below cluster version on ... was no load on the server at the time system get rebooted as well as ... Broken pipe ...
    (RedHat)
  • EV_AGENT message error
    ... returns poor feedbacks. ... The server is a node of veritas cluster with a EMC CX400 storage. ... RWSocket::send: Broken pipe ...
    (SunManagers)
  • Re: handling borken pipes
    ... computer crashes without finishing the connection, leaving me with a ... broken pipe, it shuts down my server. ...
    (comp.programming)
  • Re: why lighttpd complain "sendfile: Broken pipe 32 "
    ... This is probably because the client closed the connection before the server ... send all data - so broken pipe. ... And ideally the handler in lighttpd should be handling SIGPIPE in this ...
    (freebsd-stable)