Nikodemus, thank you for your response.
My current problem is that with Hunchentoot server after some of these errors I got a lot of request handling threads hanging.
Is there a way in SBCL to debug post-mortem those threads to see where they hung?
(I can connect to the runing image and see the list of the threads)


On Sat, Nov 2, 2013 at 9:30 AM, Nikodemus Siivola <nikodemus@random-state.net> wrote:
On 1 November 2013 20:38, Andrei Stebakov <lispercat@gmail.com> wrote:
> I wonder how do I eliminate this bug:
>
> [ERROR]] Couldn't write to #<SB-SYS:FD-STREAM
>                                                   for "socket
> 127.0.0.1:3001, peer: 127.0.0.1:45338"
>                                                   {100941D0A3}>:
>                                 Broken pipe
>
>
> It's hard to reproduce it happens on my server ones in a while.
> Is it something fixable? I saw this bug mentioned a long time ago.
> I am using the latest SBCL for linux.

That's just a disconnected client. Business as usual when talking over
the network. See eg.

http://stackoverflow.com/questions/4584904/what-causes-the-broken-pipe-error

Just handle STREAM-ERRORs and you're fine. (A SIMPLE-STREAM-ERROR is a
subclass of STREAM-ERROR. If you need to identify the stream, it is
accessible as STREAM-ERROR-STREAM.)

Cheers,

 -- nikodemus

PS. sbcl-bugs is for reporting bugs in SBCL. For help on using SBCL,
please use sbcl-help.