REGTESTS: http-messaging: test the option http-no-delay

The option was listed as out of reach because it only changes when the
data are pushed out and because no vtest assertion expresses a duration.
It turns out haproxy measures that duration itself: %Tt reports the
whole transaction, so it is enough to build an exchange whose length is
exactly the delay being looked for and to let the syslog expect count
the digits.

The server sends a chunked response, then its first chunk alone, and
only terminates the message once a barrier tells it the client has
received that chunk. Without the option, the incomplete message is held
by the system for its cork timeout, which is 200 ms on Linux, and the
transaction lasts that long; with the option it lasts zero millisecond.
Five runs here gave 200 to 222 ms against 0, so a three digit value
against a one or two digit one leaves a comfortable margin.

The delay comes from MSG_MORE, which is compiled out as a no-op where
it does not exist, leaving nothing to observe, hence the excluded
targets.

diff --git a/reg-tests/http-messaging/http_no_delay.vtc b/reg-tests/http-messaging/http_no_delay.vtc
new file mode 100644 (file)
index 0000000..2044c0f
--- /dev/null
+++ b/reg-tests/http-messaging/http_no_delay.vtc
@@ -0,0 +1,100 @@
+#EXCLUDE_TARGETS=freebsd,osx,generic
+varnishtest "Test the option http-no-delay"
+
+# "option http-no-delay" had no coverage. It disables the optimizations
+# which tell the system that more data is coming for the message being
+# forwarded, so that a partially transferred message is not pushed out
+# immediately. Nothing reports the option itself, only the moment at
+# which the data leave haproxy changes, so the test measures that delay.
+#
+# The server sends a chunked response, then its first chunk alone, and
+# only terminates the message once the client has received that chunk,
+# which the barrier below synchronises. The whole transaction therefore
+# lasts as long as that first chunk takes to reach the client, and %Tt
+# reports it. Without the option the system holds the incomplete message
+# for its cork timeout, 200 ms on Linux, which is far above the few
+# milliseconds a loopback exchange takes.
+#
+# The delay is produced by MSG_MORE, which does not exist everywhere and
+# is compiled out as a no-op in that case, leaving nothing to observe,
+# hence the excluded targets above.
+
+feature ignore_unknown_macro
+
+barrier b1 cond 2
+barrier b2 cond 2
+
+server s1 {
+       rxreq
+       txresp -nolen -hdr "transfer-encoding: chunked"
+       chunkedlen 5
+       # the client releases it once it has received the chunk
+       barrier b1 sync
+       chunkedlen 0
+} -start
+
+server s2 {
+       rxreq
+       txresp -nolen -hdr "transfer-encoding: chunked"
+       chunkedlen 5
+       barrier b2 sync
+       chunkedlen 0
+} -start
+
+syslog Slg -level info {
+       # three digits: the incomplete message waited for the cork timeout
+       recv
+       expect ~ ": delayed tt=[0-9]{3,} "
+       # one or two digits: it was pushed out immediately
+       recv
+       expect ~ ": nodelay tt=[0-9]{1,2} "
+} -start
+
+haproxy h1 -conf {
+    global
+    .if feature(THREAD)
+        thread-groups 1
+    .endif
+        chroot /
+       log "${Slg_addr}:${Slg_port}" local0 info
+
+    defaults
+       mode http
+       log global
+       timeout connect "${HAPROXY_TEST_TIMEOUT-5s}"
+       timeout client  "${HAPROXY_TEST_TIMEOUT-5s}"
+       timeout server  "${HAPROXY_TEST_TIMEOUT-5s}"
+
+    listen fe1
+       bind "fd@${fe1}"
+       log-format "delayed tt=%Tt td=%Td"
+       server s1 ${s1_addr}:${s1_port}
+
+    listen fe2
+       bind "fd@${fe2}"
+       option http-no-delay
+       log-format "nodelay tt=%Tt td=%Td"
+       server s2 ${s2_addr}:${s2_port}
+} -start
+
+client c1 -connect ${h1_fe1_sock} {
+       txreq
+       rxresphdrs
+       rxchunk
+       expect resp.chunklen == 5
+       barrier b1 sync
+       rxchunk
+       expect resp.bodylen == 5
+} -run
+
+client c2 -connect ${h1_fe2_sock} {
+       txreq
+       rxresphdrs
+       rxchunk
+       expect resp.chunklen == 5
+       barrier b2 sync
+       rxchunk
+       expect resp.bodylen == 5
+} -run
+
+syslog Slg -wait