Added files to enable fuzz testing and fixed bug found with fuzzing (#150)

* Added files to enable fuzz testing
* Added test for crashes found with fuzzing
* fixed divide-by-zero error in profile/legacy_java_profile.go found with fuzzing

fuzz/README.md

@@ -0,0 +1,37 @@
+This is an explanation of how to do fuzzing of ParseData. This uses github.com/dvyukov/go-fuzz/ for fuzzing.
+
+# How to use
+First, get go-fuzz 
+```
+$ go get github.com/dvyukov/go-fuzz/go-fuzz
+$ go get github.com/dvyukov/go-fuzz/go-fuzz-build
+```
+
+Build the test program by calling the following command 
+(assuming you have files for pprof located in github.com/google/pprof within go's src folder)
+
+```
+$ go-fuzz-build github.com/google/pprof/fuzz
+```
+The above command will produce pprof-fuzz.zip 
+
+
+Now you can run the fuzzer by calling
+
+```
+$ go-fuzz -bin=./pprof-fuzz.zip -workdir=fuzz
+```
+
+This will save a corpus of files used by the fuzzer in ./fuzz/corpus, and
+all files that caused ParseData to crash in ./fuzz/crashers.
+
+For more details on the usage, see github.com/dvyukov/go-fuzz/
+
+# About the to corpus
+
+Right now, fuzz/corpus contains the corpus initially given to the fuzzer
+
+If using the above commands, fuzz/corpus will be used to generate the initial corpus during fuzz testing.
+
+One can add profiles into the corpus by placing these files in the corpus directory (fuzz/corpus)
+prior to calling go-fuzz-build.

fuzz/fuzz_test.go

@@ -0,0 +1,24 @@
+package pprof
+
+import (
+	"github.com/google/pprof/profile"
+	"io/ioutil"
+	"testing"
+)
+
+func TestParseData(t *testing.T) {
+	const path = "testdata/"
+	files, err := ioutil.ReadDir(path)
+	if err != nil {
+		t.Errorf("Problem reading directory %s : %v", path, err)
+	}
+	for _, f := range files {
+		file := path + f.Name()
+		inbytes, err := ioutil.ReadFile(file)
+		if err != nil {
+			t.Errorf("Problem reading file: %s : %v", file, err)
+			continue
+		}
+		profile.ParseData(inbytes)
+	}
+}

fuzz/main.go

@@ -0,0 +1,11 @@
+package pprof
+
+import (
+	"github.com/google/pprof/profile"
+)
+
+// Fuzz can be used with https://github.com/dvyukov/go-fuzz to do fuzz testing on ParseData
+func Fuzz(data []byte) int {
+	profile.ParseData(data)
+	return 0
+}

fuzz/testdata/7e3c92482f6f39fc502b822ded792c589849cca8

@@ -0,0 +1,2 @@
+--- heapz 1 ---
+0 0 @ 0

profile/legacy_java_profile.go

@@ -212,6 +212,9 @@ func parseJavaSamples(pType string, b []byte, p *Profile) ([]byte, map[uint64]*L
 			switch pType {
 			case "heap":
 				const javaHeapzSamplingRate = 524288 // 512K
+				if s.Value[0] == 0 {
+					return nil, nil, fmt.Errorf("parsing sample %s: second value must be non-zero", line)
+				}
 				s.NumLabel = map[string][]int64{"bytes": {s.Value[1] / s.Value[0]}}
 				s.Value[0], s.Value[1] = scaleHeapSample(s.Value[0], s.Value[1], javaHeapzSamplingRate)
 			case "contention":