|author||anonym <firstname.lastname@example.org>||2016-05-13 14:43:01 +0200|
|committer||anonym <email@example.com>||2016-05-13 15:10:45 +0200|
Run tcpdump with --immediate-mode.
With this option, "packets are delivered to tcpdump as soon as they arrive, rather than being buffered for efficiency". It's not available in Jessie's tcpdump, but it is in jessie-backports. Background: Normally tcpdump uses a 1000 ms = 1 s timeout (set via `pcap_set_timeout`; for more details, see the "read timeout" section in the "pcap" man page from the libpcap*-dev package) which explaines that the `sleep 1` worked most of the time before on my lightly loaded system, but much rarely on jenkins highly loaded system. At the moment this should hopefully fix the issues we have with the "Anti test: Detecting UDP leaks of DNS lookups with the firewall leak detector" scenario in the test/9521-chutney branch.
2 files changed, 2 insertions, 10 deletions
diff --git a/features/support/helpers/sniffing_helper.rb b/features/support/helpers/sniffing_helper.rb
index 1f48e71..38b1382 100644
@@ -27,6 +27,7 @@ class Sniffer
@@ -37,15 +38,6 @@ class Sniffer
- # Sometimes tcpdump/pcap is not fast enough to write a captured
- # packet to file vs when we kill the process. For instance, if we
- # sniff then network, run some step that generates traffic that we
- # are interested in, and then immediately run this method, we
- # might kill tcpdump before it has handled said traffic, so
- # there's a race. Since we already run tcpdump with `-U` there
- # seems to be little else we can do than a static sleep, short of
- # switching to another sniffing technology that is faster.
- sleep 1
diff --git a/wiki/src/contribute/release_process/test/setup.mdwn b/wiki/src/contribute/release_process/test/setup.mdwn
index e28bf36..583fb62 100644
@@ -52,7 +52,7 @@ The following packages are necessary on Debian Jessie:
- tcpdump \
+ tcpdump/jessie-backports \