Comment 1:

Status changed to Accepted.

Comment 2:

Hi,
Does this happen so frequently?

Comment 3 by steview.woodcock:

no its quite difficult to reproduce which suggests some form of race.

Comment 4:

[The time for maybe has passed.]

Comment 5:

I can reproduce this easily after a few minutes running this script on an 8 core machine
--- FAIL: TestDualStackTCPListener-107 (0.00 seconds)
        unicast_posix_test.go:407: Second Listen("tcp4", "0.0.0.0:40030") returns listen tcp4 0.0.0.0:40030: address already in use, expected <nil>
ip-10-248-36-242(~/go/src) % cat stress.bash 
#!/bin/bash
set -e
eval $(go env)
export GOROOT
pkgs() {
        go list std | grep -v cmd
}
rand() {
        echo $[ 1 + $[ RANDOM % $[ $1 - 1 ] ] ]
}
pkgs | xargs -n 1 -P 4 -I % sh -c 'cd $GOROOT/src/pkg/$0 && go test -v -c' %
while true ; do 
        pkgs | xargs -n1 -P $(rand 8) -I % -t bash -xe -c \
                'cd $GOROOT/src/pkg/$0 && if [ -x $(basename $0).test ] ; then GOMAXPROCS=$1 $GOROOT/src/pkg/$0/$(basename $0).test || exit 255; fi ' % $(rand 128) 
done

Comment 6:

Hi Dave,
Can you try to repro w/ https://golang.org/cl/7880046/ ?
CL 7880046 will report the first TCP listener state when second one fails.

Comment 7:

Sorry, still occurs with 7880046 applied
--- FAIL: TestDualStackTCPListener-78 (0.00 seconds)
        unicast_posix_test.go:19: &syscall.TCPInfo{State:0xa, Ca_state:0x0, Retransmits:0x0, Probes:0x0, Backoff:0x0, Options:0x0, Pad_cgo_0:[2]uint8{0x0, 0x0}, Rto:0xf4240, Ato:0x0, Snd_mss:0x218, Rcv_mss:0x0, Unacked:0x0, Sacked:0x80, Lost:0x0, Retrans:0x0, Fackets:0x0, Last_data_sent:0x224ecc, Last_ack_sent:0x0, Last_data_recv:0x224ecc, Last_ack_recv:0x224ecc, Pmtu:0x0, Rcv_ssthresh:0x0, Rtt:0x0, Rttvar:0x3d090, Snd_ssthresh:0x7fffffff, Snd_cwnd:0xa, Advmss:0x0, Reordering:0x3, Rcv_rtt:0x0, Rcv_space:0x0, Total_retrans:0x0}
        unicast_posix_test.go:20: Second Listen("tcp4", ":53501") returns listen tcp4 <nil>:53501: address already in use, expected <nil>

Comment 8:

No, that's fine and what I expected. Thanks a lot!
  State:0xa
  Other fields are pretty clean and looks still larval
Hm, which version of Linux kernel are you running?
(Have to have a look at correct definitions for states)

Comment 9:

This was an m2.xlarge running in ec2 with whatever flavour of the
Ubuntu 12.10 kernel runs in that environment, probably 3.4 or 3.5.

Comment 10:

Aha, the first one still stays in TCP_LISTEN state.
Probably test cases might be wrong.

Comment 11:

Lowering priority, this doesn't happen that frequently.

Labels changed: added priority-later, removed priority-triage.

Comment 12:

Labels changed: added go1.2maybe.

Comment 13:

We continue to see this on the builders.
e.g. http://build.golang.org/log/f641c1499be25ad271b91dbc0656309375bcccee

Comment 14:

Labels changed: added testing.

Comment 15:

Labels changed: added go1.3, removed go1.2maybe.

Comment 16:

Labels changed: added release-go1.3.

Comment 17:

Labels changed: removed go1.3.

Comment 18:

Labels changed: added repo-main.

Comment 19:

This issue was updated by revision 8bc3278.

This test is flakey on linux servers and fails otherwise good builds. Mikio has some
proposals to fix the test, but they require additional plumbing.
In the meantime, disable this test in -short mode so it will run during the full net
test suite, but not during builder ci.
R=golang-codereviews, iant
CC=golang-codereviews
https://golang.org/cl/53410043

Comment 20:

Labels changed: added release-go1.4, removed release-go1.3.

Comment 21:

Labels changed: added release-none, removed release-go1.4.

Comment 22:

Issue #4176 has been merged into this issue.

Comment 23:

CL https://golang.org/cl/154950043 mentions this issue.

Comment 24:

This issue was updated by revision 83001ff.

LGTM=bradfitz
R=rlh, bradfitz
CC=golang-codereviews
https://golang.org/cl/154950043

CL https://golang.org/cl/9661 mentions this issue.