1 files changed, 343 insertions, 0 deletions

diff --git a/tests/t_urlencoded.c b/tests/t_urlencoded.c
new file mode 100644
index 0000000..6e0ee31
--- /dev/null
+++ b/tests/t_urlencoded.c
@@ -0,0 +1,343 @@
+/* This software was written by Dirk Engling <erdgeist@erdgeist.org>
+   It is considered beerware. Prost. Skol. Cheers or whatever.
+
+   $id$ */
+
+/* System */
+#include <string.h>
+
+/* Opentracker */
+#include "ot_urlencoded.h"
+#include "test.h"
+
+/* The scanner returns the number of bytes written to the destination, or
+   -1 for a parse error and -2 for a hard terminator encountered before any
+   output was produced. It advances the source pointer past the separator it
+   stopped on, except for hard terminators and the '&' ending a parameter,
+   which it leaves for the next call to see. */
+
+static void case_path(void) {
+  char    query[] = "/announce?info_hash=A";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("path stops at the question mark");
+  n = urlencoded_parse(&p, dest, SCAN_PATH);
+  CHECK_INT(n, 9);
+  dest[n] = 0;
+  CHECK_STR(dest, "/announce");
+  CHECK_INT(*p, 'i');
+}
+
+static void case_param_and_value(void) {
+  char    query[] = "info_hash=abc&port=80";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("parameter stops at '=', value stops at '&'");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_PARAM);
+  CHECK_INT(n, 9);
+  dest[n] = 0;
+  CHECK_STR(dest, "info_hash");
+
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 3);
+  dest[n] = 0;
+  CHECK_STR(dest, "abc");
+
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_PARAM);
+  CHECK_INT(n, 4);
+  dest[n] = 0;
+  CHECK_STR(dest, "port");
+
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 2);
+  dest[n] = 0;
+  CHECK_STR(dest, "80");
+}
+
+static void case_percent_decoding(void) {
+  char    query[] = "%41%42%43&x=1";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("percent escapes decode");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 3);
+  CHECK_MEM(dest, "ABC", 3);
+}
+
+static void case_binary_info_hash(void) {
+  /* The twenty byte info_hash is what the announce path really decodes */
+  char          query[] = "%00%01%02%ff%fe%7f%80ABCDEFGHIJKLM&x=1";
+  unsigned char want[]  = {0x00, 0x01, 0x02, 0xff, 0xfe, 0x7f, 0x80, 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M'};
+  char          dest[64];
+  char         *p = query;
+  ssize_t       n;
+
+  test_case("twenty byte binary value survives decoding");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 20);
+  CHECK_MEM(dest, want, 20);
+}
+
+static void case_in_place(void) {
+  char    query[] = "%41%42&x=1";
+  char   *p       = query;
+  ssize_t n;
+
+  test_case("decoding in place over the source is safe");
+  n = urlencoded_parse(&p, query, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 2);
+  CHECK_MEM(query, "AB", 2);
+}
+
+static void case_bad_escape(void) {
+  char    query[] = "%zz";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("a non hexadecimal escape is a parse error");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, -1);
+}
+
+static void case_truncated_escape(void) {
+  char    query[] = "%4";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("an escape cut short by the terminator is a parse error");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, -1);
+}
+
+static void case_terminator(void) {
+  char    empty[] = "";
+  char    space[] = " rest";
+  char    dest[64];
+  char   *p = empty;
+  ssize_t n;
+
+  test_case("a hard terminator before any output reports end of input");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_PARAM);
+  CHECK_INT(n, -2);
+
+  p = space;
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_PARAM);
+  CHECK_INT(n, -2);
+}
+
+static void case_terminator_after_output(void) {
+  char    query[] = "abc";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("a hard terminator after output is left for the next call");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 3);
+  CHECK_INT(*p, 0);
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, -2);
+}
+
+/* The character matrix makes '?' terminate a path only, and '=' terminate a
+   path or a parameter only. Both are ordinary characters inside a value, so a
+   value is free to contain a nested query string. */
+
+static void case_separators_are_contextual(void) {
+  char    question[] = "a?b&x=1";
+  char    equals[]   = "a=b&x=1";
+  char    dest[64];
+  char   *p = question;
+  ssize_t n;
+
+  test_case("a question mark is literal inside a value");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 3);
+  dest[n] = 0;
+  CHECK_STR(dest, "a?b");
+
+  test_case("an equals sign is literal inside a value");
+  p = equals;
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, 3);
+  dest[n] = 0;
+  CHECK_STR(dest, "a=b");
+}
+
+static void case_ampersand_in_path(void) {
+  char    query[] = "/path&x";
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("an ampersand in a path is a parse error");
+  n = urlencoded_parse(&p, dest, SCAN_PATH);
+  CHECK_INT(n, -1);
+}
+
+static void case_raw_high_byte(void) {
+  char    query[] = {'a', 'b', (char)0x80, 0};
+  char    dest[64];
+  char   *p = query;
+  ssize_t n;
+
+  test_case("an undecoded byte above ASCII is a parse error");
+  n = urlencoded_parse(&p, dest, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(n, -1);
+}
+
+static void case_skipvalue(void) {
+  char  query[] = "abc&next=1";
+  char  bare[]  = "abc";
+  char *p       = query;
+
+  test_case("skipvalue steps over the value and its separator");
+  urlencoded_skip_value(&p);
+  CHECK_STR(p, "next=1");
+
+  test_case("skipvalue leaves a hard terminator in place");
+  p = bare;
+  urlencoded_skip_value(&p);
+  CHECK_INT(*p, 0);
+}
+
+static void case_find_keywords(void) {
+  static const ot_keywords keywords[] = {{"port", 1}, {"left", 2}, {"compact", 3}, {NULL, -3}};
+  char                     match[]    = "left=0";
+  char                     nomatch[]  = "bogus=1";
+  char                     prefix[]   = "por=1";
+  char                    *p          = match;
+
+  test_case("a known keyword returns its value");
+  CHECK_INT(urlencoded_match_keyword(keywords, &p, SCAN_SEARCHPATH_PARAM), 2);
+
+  test_case("an unknown keyword returns -3");
+  p = nomatch;
+  CHECK_INT(urlencoded_match_keyword(keywords, &p, SCAN_SEARCHPATH_PARAM), -3);
+
+  test_case("a prefix of a keyword does not match it");
+  p = prefix;
+  CHECK_INT(urlencoded_match_keyword(keywords, &p, SCAN_SEARCHPATH_PARAM), -3);
+}
+
+/* The encoding direction. Which bytes survive unescaped is fixed by what
+   existing parsers of the urlencoded full scrape expect, so these cases pin
+   the set rather than describe a preference. */
+
+static void case_fmt_passes_safe_bytes(void) {
+  const uint8_t src[] = "aZ09$-./=_~";
+  char          dest[64];
+  size_t        n;
+
+  test_case("alphanumerics and the safe punctuation pass through");
+  n       = urlencoded_format(dest, src, 11);
+  dest[n] = 0;
+  CHECK_INT(n, 11);
+  CHECK_STR(dest, "aZ09$-./=_~");
+}
+
+static void case_fmt_escapes_the_rest(void) {
+  const uint8_t src[] = {' ', '%', '&', '+', ':', '?'};
+  char          dest[64];
+  size_t        n;
+
+  test_case("everything else becomes a percent escape");
+  n       = urlencoded_format(dest, src, sizeof(src));
+  dest[n] = 0;
+  CHECK_INT(n, 18);
+  CHECK_STR(dest, "%20%25%26%2b%3a%3f");
+}
+
+static void case_fmt_hex_is_lower_case(void) {
+  const uint8_t src[] = {0x00, 0xab, 0xff};
+  char          dest[64];
+  size_t        n;
+
+  test_case("escapes use lower case hex digits");
+  n       = urlencoded_format(dest, src, sizeof(src));
+  dest[n] = 0;
+  CHECK_INT(n, 9);
+  CHECK_STR(dest, "%00%ab%ff");
+}
+
+static void case_fmt_zero_length(void) {
+  char dest[4];
+
+  test_case("a zero length input writes nothing");
+  memset(dest, 0x5a, sizeof(dest));
+  CHECK_INT(urlencoded_format(dest, (const uint8_t *)"abc", 0), 0);
+  CHECK_MEM(dest, "\x5a\x5a\x5a\x5a", 4);
+}
+
+static void case_fmt_worst_case_fits(void) {
+  uint8_t src[20];
+  char    dest[61];
+  size_t  i, n;
+
+  test_case("a hash of entirely unsafe bytes needs three times its length");
+  for (i = 0; i < sizeof(src); ++i)
+    src[i] = 0xff;
+  n       = urlencoded_format(dest, src, sizeof(src));
+  dest[n] = 0;
+  CHECK_INT(n, 60);
+  for (i = 0; i < 20; ++i)
+    if (memcmp(dest + 3 * i, "%ff", 3)) {
+      test_failf(__FILE__, __LINE__, "escape %zu is wrong", i);
+      return;
+    }
+  ++test_checks;
+}
+
+static void case_fmt_decoder_disagrees_on_dollar(void) {
+  /* Documenting a known asymmetry rather than a wish. Of the bytes the
+     encoder passes through unescaped, '$' is the only one the decoder then
+     refuses, and it refuses the whole string rather than stopping short.
+     Output of the urlencoded full scrape therefore does not round trip
+     through urlencoded_parse when a hash contains 0x24. Harmless today
+     because only third party tooling reads that output. */
+  const uint8_t src[]  = {'a', '$', 'b'};
+  char          dest[16];
+  char          scanned[16];
+  char         *p = dest;
+  size_t        n;
+  ssize_t       got;
+
+  test_case("the encoder passes '$' but the decoder rejects it");
+  n       = urlencoded_format(dest, src, sizeof(src));
+  dest[n] = 0;
+  CHECK_STR(dest, "a$b");
+
+  got = urlencoded_parse(&p, scanned, SCAN_SEARCHPATH_VALUE);
+  CHECK_INT(got, -1);
+}
+
+void test_urlencoded(void) {
+  case_path();
+  case_param_and_value();
+  case_percent_decoding();
+  case_binary_info_hash();
+  case_in_place();
+  case_bad_escape();
+  case_truncated_escape();
+  case_terminator();
+  case_terminator_after_output();
+  case_separators_are_contextual();
+  case_ampersand_in_path();
+  case_raw_high_byte();
+  case_skipvalue();
+  case_find_keywords();
+  case_fmt_passes_safe_bytes();
+  case_fmt_escapes_the_rest();
+  case_fmt_hex_is_lower_case();
+  case_fmt_zero_length();
+  case_fmt_worst_case_fits();
+  case_fmt_decoder_disagrees_on_dollar();
+}