Comment 1:

It seems that we need to ignore WSAECONNRESET from WSAAccept. And maybe also
WSAECONNREFUSED. Why does windows even return them?..

Comment 2:

Do you propose to ignore them in net/http or in net? We shouldn't be "ignoring" OS
errors (if possible). It should be up to the user what to do with them.
Alex

Comment 3 by trivital:

Alex, the problem is that you need to write different code for different OSes.
Plus, I cant handle this error even if I wanted to. http.ListenAndServe just plain
crashed.

Comment 4:

I would say that this is not an OS error, it's more of an informational notification
exposed through standard error reporting channel. Think of EINTR.

Comment 5 by trivital:

net/http snippet 
func (srv *Server) Serve(l net.Listener) error {
    defer l.Close()
    var tempDelay time.Duration // how long to sleep on accept failure
    for {
        rw, e := l.Accept() //this is where it happens ...
        if e != nil {
            if ne, ok := e.(net.Error); ok && ne.Temporary() {
                if tempDelay == 0 {
                    tempDelay = 5 * time.Millisecond
                } else {
                    tempDelay *= 2
                }
                if max := 1 * time.Second; tempDelay > max {
                    tempDelay = max
                }
                log.Printf("http: Accept error: %v; retrying in %v", e, tempDelay)
                time.Sleep(tempDelay)
                continue
            }
            return e
        }
        tempDelay = 0
        c, err := srv.newConn(rw)
        if err != nil {
            continue
        }
        go c.serve()
    }
}

Comment 6 by trivital:

i think we should continue on e.(net.Error)==WSAECONNRESET instead of return e

Comment 7:

WSAECONNRESET is actually about new connection, not about listening socket. I suspect,
it just the way AcceptEx is implemented that it returns that error. I suppose, we should
ignore it.
Alex

Labels changed: added release-go1.3, repo-main, os-windows.

Owner changed to @alexbrainman.

Status changed to Accepted.

Comment 8 by trivital:

exactly Alex, agree!

Comment 9:

Labels changed: added release-go1.2.1, removed release-go1.3.

Comment 10:

This is related to issue #6163, which is about Accept not ignoring all temporary errors
on Linux.

Comment 11:

Please review https://golang.org/cl/49490043/
Alex

Status changed to Started.

Comment 12:

This issue was closed by revision c7ef348.

Status changed to Fixed.

Comment 13 by rivercheng:

It seems this issue still exists on go1.2 amd 64.

Comment 14:

@rivercheng,
You won't see this fix as part of go1.2 - it has been fixed after go1.2 has been
released.
This is issue is marked as "Release-Go1.2.1", so, I expect, this change should be part
of go-1.2.1.
Alex

Comment 15 by rivercheng:

Thanks, Alex.
I forgot to check the date and thought this was a bug fixed long ago. :)

Comment 16:

This issue was closed by revision 4e312b5239f0.

CL https://golang.org/cl/9946 mentions this issue.