start fuzzing

CHANGELOG.md

@@ -7,6 +7,9 @@ fix:
 * allow empty template interpolations
 * allow empty templates
 
+quality:
+* add fuzzing
+
 ## 0.2.0 - 2021-06-26
 
 feature:

README.md

@@ -26,8 +26,9 @@ tree-sitter parse --quiet --stat example/real_world_stuff/*/*
 Total parses: 1892; successful parses: 1892; failed parses: 0; success percentage: 100.00%
 ```
 
-The aim is to build unit testcases from selected failure classes and slowly get to 100%. 
+## Fuzzing
 
+The directory `fuzz/crashers` contains a set of crashes that were found with fuzzing. To fuzz the parser i used the instrumentation of [tree-sitter](https://github.com/tree-sitter/tree-sitter/tree/master/test/fuzz) 
 
 ## Todo
 
@@ -37,3 +38,8 @@ The aim is to build unit testcases from selected failure classes and slowly get
     * [x] add quoted template interpolations
     * [ ] add quoted template directives
   * [x] add heredoc templates
+* [ ] fuzzing
+  * [x] start with fuzzing the parser
+  * [ ] upload fuzzing instrumentation
+  * [ ] document fuzzing process
+  * [ ] add parsing of crashers to CI process

src/scanner.cc

@@ -1,10 +1,10 @@
 #include <tree_sitter/parser.h>
 
+#include <climits>
 #include <vector>
 #include <string>
 #include <wctype.h>
 #include <assert.h>
-#include <stdio.h>
 
 namespace {
 
@@ -40,6 +40,10 @@ public:
   unsigned serialize(char* buf) {
     unsigned size = 0;
 
+    if (context_stack.size() > CHAR_MAX) {
+      return 0;
+    }
+
     buf[size++] = context_stack.size();
     for (vector<Context>::iterator it = context_stack.begin(); it != context_stack.end(); ++it) {
       if (size + 2 + it->heredoc_identifier.size() >= TREE_SITTER_SERIALIZATION_BUFFER_SIZE) {
@@ -54,12 +58,13 @@ public:
   }
 
   void deserialize(const char* buf, unsigned n) {
-    unsigned size = 0;
+    context_stack.clear();
+
     if (n == 0) {
       return;
     }
-    context_stack.clear();
 
+    unsigned size = 0;
     uint8_t context_stack_size = buf[size++];
     for (unsigned j = 0; j < context_stack_size; j++) {
       Context ctx;