"keepalive_timeout 0" (really) disables keepalive

Although this behavior is mentioned on the documentation,
this was broken under EventMachine, Rev*, and Revactor.

Furthermore, we set the "Connection: close" header to allow the
client to optimize is handling of non-keepalive connections.

1 files changed, 43 insertions, 0 deletions

diff --git a/t/t0017-keepalive-timeout-zero.sh b/t/t0017-keepalive-timeout-zero.sh
new file mode 100755
index 0000000..4ba5dd0
--- /dev/null
+++ b/t/t0017-keepalive-timeout-zero.sh
@@ -0,0 +1,43 @@
+#!/bin/sh
+. ./test-lib.sh
+t_plan 6 "keepalive_timeout 0 tests for $model"
+
+t_begin "setup and start" && {
+        rainbows_setup $model 2 0
+        grep 'keepalive_timeout 0' $unicorn_config
+        rainbows -D env.ru -c $unicorn_config
+        rainbows_wait_start
+}
+
+t_begin 'check server responds with Connection: close' && {
+        curl -sSfi http://$listen/ | grep 'Connection: close'
+}
+
+t_begin "send keepalive response that does not expect close" && {
+        req='GET / HTTP/1.1\r\nHost: example.com\r\n\r\n'
+        t0=$(date +%s)
+        (
+                cat $fifo > $tmp &
+                printf "$req"
+                wait
+                date +%s > $ok
+        ) | socat - TCP:$listen > $fifo
+        now="$(cat $ok)"
+        elapsed=$(( $now - $t0 ))
+        t_info "elapsed=$elapsed (expecting <=3)"
+        test $elapsed -le 3
+}
+
+t_begin "'Connection: close' header set" && {
+        grep 'Connection: close' $tmp
+}
+
+t_begin "killing succeeds" && {
+        kill $rainbows_pid
+}
+
+t_begin "check stderr" && {
+        check_stderr
+}
+
+t_done