[PATCH] D20301: [libfuzzer] Trying random unit prefixes during corpus load.

[Kostya Serebryany via llvm-commits]

kcc added inline comments.

================
Comment at: lib/Fuzzer/FuzzerFlags.def:87
@@ -86,2 +86,3 @@
     "reaching this limit of RSS memory usage.")
+FUZZER_FLAG_INT(truncate_units, 1, "Try truncated units when loading corpus.")
 
----------------
I'd prefer to have this flag of by default for now

================
Comment at: lib/Fuzzer/FuzzerLoop.cpp:362
@@ +361,3 @@
+  size_t MaxCorpusLen = 0;
+  for (const auto &U : Corpus) {
+    MaxCorpusLen = std::max(MaxCorpusLen, U.size());
----------------
no {}, same below

================
Comment at: lib/Fuzzer/FuzzerLoop.cpp:379
@@ +378,3 @@
+  }
+  std::sort(Sizes.begin(), Sizes.end());
+
----------------
do you expect duplicates here? 

================
Comment at: lib/Fuzzer/test/FuzzerUnittest.cpp:16
@@ -15,1 +15,3 @@
 
+extern "C" int EmptyLLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {
+  return 0;
----------------
make it static and not extern "C"


http://reviews.llvm.org/D20301