1 files changed, 37 insertions, 0 deletions
diff --git a/third_party/libwebrtc/test/fuzzers/corpora/README b/third_party/libwebrtc/test/fuzzers/corpora/README
new file mode 100644
index 0000000000..cc87025ff6
--- /dev/null
+++ b/third_party/libwebrtc/test/fuzzers/corpora/README
@@ -0,0 +1,37 @@
+This is a collection of corpora for various WebRTC fuzzers. To use
+them, the gn targets define seed_corpus=$corpus_dir, which causes the
+ClusterFuzz upload bot to bundle $corpus_dir and upload it.
+
+The format is simple: one file per test case. Specific notes are
+included below.
+
+### SDP ###
+This corpus was initially assembled manually from the following
+sources:
+
+  - curl --silent https://www.ietf.org/rfc/rfc4317.txt | grep '^[ a-z]*=[^=]*$' | sed 's/^[[:space:]]*//' | awk -v RS='(^|\n)v=' '/./ {print "v="$0 > NR".sdp"}'
+  - all the SDPs used in the parser unit tests
+  - some manually gathered SDPs from Firefox and Opera
+
+The SDP tokens come from:
+
+ -  grep "^static const " webrtc/api/webrtcsdp.cc | cut -d'=' -f2 | cut -d ';' -f1 | tr -d '"' | tr -d "'" | tr -d ' ' | sort -u | grep -v '^(\n|\r|\r\n)$|^$' | sed -e 's/^/"/' -e 's/$/"/' | tail -n +2
+
+### STUN ###
+This corpus was initially assembled from the STUN unit tests, together
+with a crash that it found relatively quickly.
+
+### RT(C)P ###
+This corpus was initially assembled from the unittests. RTCP was
+minimised first.
+
+There is also rt(c?)p-corpus-with-extra-byte, in which each sample is
+prefixed by the byte 0xff. Some of the rtp fuzzers need to decide
+which header extensions to enable, and the first byte of the fuzz data
+is used for this.
+
+### PseudoTCP ###
+Very small corpus minimised from the unit tests.
+
+### SCTP ###
+This corpus was extracted from a few manually recorder wireshark dumps.