From 976dfc7195bd72cd799964eef59eeb261538781e Mon Sep 17 00:00:00 2001
From: Andrew Chambers <ac@acha.ninja>
Date: Sat, 23 Nov 2019 16:55:23 +1300
Subject: [PATCH] Minor fixes for parser

Check length before dereferencing buffer in tokenchar.
Check keywords are valid utf-8.
Fix minor typos.
---
 src/core/parse.c  | 15 +++++++++------
 test/suite6.janet | 10 ++++++++++
 2 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/src/core/parse.c b/src/core/parse.c
index eb4ad24a..92486500 100644
--- a/src/core/parse.c
+++ b/src/core/parse.c
@@ -38,7 +38,7 @@ static int is_whitespace(uint8_t c) {
 
 /* Code generated by tools/symcharsgen.c.
  * The table contains 256 bits, where each bit is 1
- * if the corresponding ascci code is a symbol char, and 0
+ * if the corresponding ascii code is a symbol char, and 0
  * if not. The upper characters are also considered symbol
  * chars and are then checked for utf-8 compliance. */
 static const uint32_t symchars[8] = {
@@ -233,7 +233,7 @@ static int escapeh(JanetParser *p, JanetParseState *state, uint8_t c) {
         p->error = "invalid hex digit in hex escape";
         return 1;
     }
-    state->argn = (state->argn << 4) + digit;;
+    state->argn = (state->argn << 4) + digit;
     state->counter--;
     if (!state->counter) {
         push_buf(p, (state->argn & 0xFF));
@@ -329,6 +329,12 @@ static int tokenchar(JanetParser *p, JanetParseState *state, uint8_t c) {
     int start_dig = p->buf[0] >= '0' && p->buf[0] <= '9';
     int start_num = start_dig || p->buf[0] == '-' || p->buf[0] == '+' || p->buf[0] == '.';
     if (p->buf[0] == ':') {
+        /* Don't do full utf-8 check unless we have seen non ascii characters. */
+        int valid = (!state->argn) || valid_utf8(p->buf + 1, blen - 1);
+        if (!valid) {
+            p->error = "invalid utf-8 in keyword";
+            return 0;
+        }
         ret = janet_keywordv(p->buf + 1, blen - 1);
     } else if (start_num && !janet_scan_number(p->buf, blen, &numval)) {
         ret = janet_wrap_number(numval);
@@ -338,7 +344,7 @@ static int tokenchar(JanetParser *p, JanetParseState *state, uint8_t c) {
         ret = janet_wrap_false();
     } else if (!check_str_const("true", p->buf, blen)) {
         ret = janet_wrap_true();
-    } else if (p->buf) {
+    } else {
         if (start_dig) {
             p->error = "symbol literal cannot start with a digit";
             return 0;
@@ -351,9 +357,6 @@ static int tokenchar(JanetParser *p, JanetParseState *state, uint8_t c) {
             }
             ret = janet_symbolv(p->buf, blen);
         }
-    } else {
-        p->error = "empty symbol invalid";
-        return 0;
     }
     p->bufcount = 0;
     popstate(p, ret);
diff --git a/test/suite6.janet b/test/suite6.janet
index e53ad0c8..b3e836e0 100644
--- a/test/suite6.janet
+++ b/test/suite6.janet
@@ -118,6 +118,16 @@
 (assert (deep= (parser/status p) (parser/status p2)) "parser 2")
 (assert (deep= (parser/state p) (parser/state p2)) "parser 3")
 
+# Parser errors
+(defn parse-error [input]
+  (def p (parser/new))
+  (parser/consume p input)
+  (parser/error p))
+
+# Invalid utf-8 sequences
+(assert (not= nil (parse-error @"\xc3\x28")) "reject invalid utf-8 symbol")
+(assert (not= nil (parse-error @":\xc3\x28")) "reject invalid utf-8 keyword")
+
 # String check-set
 (assert (string/check-set "abc" "a") "string/check-set 1")
 (assert (not (string/check-set "abc" "z")) "string/check-set 2")