mirror of
https://github.com/cryb-to/cryb-to.git
synced 2024-11-21 21:25:50 +00:00
362 lines
9.4 KiB
C
362 lines
9.4 KiB
C
/*-
|
||
* Copyright (c) 2014-2016 Dag-Erling Smørgrav
|
||
* All rights reserved.
|
||
*
|
||
* Redistribution and use in source and binary forms, with or without
|
||
* modification, are permitted provided that the following conditions
|
||
* are met:
|
||
* 1. Redistributions of source code must retain the above copyright
|
||
* notice, this list of conditions and the following disclaimer.
|
||
* 2. Redistributions in binary form must reproduce the above copyright
|
||
* notice, this list of conditions and the following disclaimer in the
|
||
* documentation and/or other materials provided with the distribution.
|
||
* 3. The name of the author may not be used to endorse or promote
|
||
* products derived from this software without specific prior written
|
||
* permission.
|
||
*
|
||
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
|
||
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
||
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
||
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
||
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
||
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
||
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
||
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
||
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
||
* SUCH DAMAGE.
|
||
*/
|
||
|
||
#include "cryb/impl.h"
|
||
|
||
#include <errno.h>
|
||
#include <stdint.h>
|
||
#include <string.h>
|
||
#include <unistd.h>
|
||
|
||
#include <cryb/rfc3986.h>
|
||
|
||
#include <cryb/test.h>
|
||
|
||
struct t_case {
|
||
const char *desc;
|
||
int (*func)(const char *, size_t, char *, size_t *);
|
||
const char *in; /* input string */
|
||
size_t ilen; /* input length */
|
||
const char *out; /* expected output string or NULL */
|
||
size_t blen; /* initial value for olen or 0 */
|
||
size_t olen; /* expected value for olen */
|
||
int ret; /* expected return value */
|
||
int err; /* expected errno if ret != 0 */
|
||
};
|
||
|
||
/* basic encoding / decoding */
|
||
#define T_ENCODE(i, o) \
|
||
{ .func = percent_encode, \
|
||
.in = i, .ilen = sizeof i, \
|
||
.out = o, .blen = sizeof o, .olen = sizeof o, \
|
||
.ret = 0, .err = 0 }
|
||
#define T_DECODE(i, o) \
|
||
{ .func = percent_decode, \
|
||
.in = i, .ilen = sizeof i, \
|
||
.out = o, .blen = sizeof o, .olen = sizeof o, \
|
||
.ret = 0, .err = 0 }
|
||
|
||
/* roundtrip encoding tests */
|
||
#define T_ENCDEC(p, e) \
|
||
T_ENCODE(p, e), T_DECODE(e, p)
|
||
|
||
/* decoding failure */
|
||
#define T_DECODE_FAIL(e, i, o) \
|
||
{ .func = percent_decode, \
|
||
.in = i, .ilen = sizeof i - 1, \
|
||
.out = o, .blen = SIZE_MAX, .olen = sizeof o, \
|
||
.ret = -1, .err = e }
|
||
|
||
/* input string shorter than input length */
|
||
#define T_SHORT_INPUT_ENC(i, o) \
|
||
{ .desc = "encode (short input)", .func = percent_encode, \
|
||
.in = i, .ilen = SIZE_MAX, \
|
||
.out = o, .blen = SIZE_MAX, .olen = sizeof o }
|
||
#define T_SHORT_INPUT_DEC(i, o) \
|
||
{ .desc = "decode (short input)", .func = percent_decode, \
|
||
.in = i, .ilen = SIZE_MAX, \
|
||
.out = o, .blen = SIZE_MAX, .olen = sizeof o }
|
||
|
||
/* input string longer than input length */
|
||
#define T_LONG_INPUT_ENC(i, il, o, ol) \
|
||
{ .desc = "encode (long input)", .func = percent_encode, \
|
||
.in = i, .ilen = il, .out = o, .blen = ol, .olen = ol }
|
||
#define T_LONG_INPUT_DEC(i, il, o, ol) \
|
||
{ .desc = "decode (long input)", .func = percent_decode, \
|
||
.in = i, .ilen = il, .out = o, .blen = ol, .olen = ol }
|
||
|
||
/* output string longer than output length */
|
||
#define T_LONG_OUTPUT_ENC(i, o, l) \
|
||
{ .desc = "encode (long output)", .func = percent_encode, \
|
||
.in = i, .ilen = SIZE_MAX, \
|
||
.out = o, .blen = l, .olen = sizeof o, \
|
||
.ret = -1, .err = ENOSPC }
|
||
#define T_LONG_OUTPUT_DEC(i, o, l) \
|
||
{ .desc = "decode (long output)", .func = percent_decode, \
|
||
.in = i, .ilen = SIZE_MAX, \
|
||
.out = o, .blen = l, .olen = sizeof o, \
|
||
.ret = -1, .err = ENOSPC }
|
||
|
||
static const char unreserved[] =
|
||
"ABCDEFGHIJKLMNOPQRSTUVWXYZ"
|
||
"abcdefghijklmnopqrstuvwxyz"
|
||
"0123456789"
|
||
"-._~";
|
||
|
||
static const char reserved[] =
|
||
":/?#[]@" /* gen-delims */
|
||
"!$&'()*+,;="; /* sub-delims */
|
||
static const char reserved_upper[] =
|
||
"%3A%2F%3F%23%5B%5D%40"
|
||
"%21%24%26%27%28%29%2A%2B%2C%3B%3D";
|
||
static const char reserved_lower[] =
|
||
"%3a%2f%3f%23%5b%5d%40"
|
||
"%21%24%26%27%28%29%2a%2b%2c%3b%3d";
|
||
|
||
static const char mixed[] =
|
||
"Magic = xyzzy";
|
||
static const char mixed_upper[] =
|
||
"Magic%20%3D%20xyzzy";
|
||
static const char mixed_lower[] =
|
||
"Magic%20%3d%20xyzzy";
|
||
|
||
static struct t_case t_cases[] = {
|
||
/* empty string */
|
||
T_ENCDEC("", ""),
|
||
|
||
/* unreserved characters */
|
||
T_ENCDEC(unreserved, unreserved),
|
||
|
||
/* reserved characters */
|
||
T_ENCDEC(reserved, reserved_upper),
|
||
T_DECODE(reserved_lower, reserved),
|
||
|
||
/* individual reserved characters */
|
||
T_ENCDEC("!", "%21"),
|
||
T_ENCDEC("#", "%23"),
|
||
T_ENCDEC("$", "%24"),
|
||
T_ENCDEC("%", "%25"),
|
||
T_ENCDEC("&", "%26"),
|
||
T_ENCDEC("'", "%27"),
|
||
T_ENCDEC("(", "%28"),
|
||
T_ENCDEC(")", "%29"),
|
||
T_ENCDEC("*", "%2A"),
|
||
T_ENCDEC("+", "%2B"),
|
||
T_ENCDEC(",", "%2C"),
|
||
T_ENCDEC("/", "%2F"),
|
||
T_ENCDEC(";", "%3B"),
|
||
T_ENCDEC(":", "%3A"),
|
||
T_ENCDEC("=", "%3D"),
|
||
T_ENCDEC("?", "%3F"),
|
||
T_ENCDEC("@", "%40"),
|
||
T_ENCDEC("[", "%5B"),
|
||
T_ENCDEC("]", "%5D"),
|
||
T_DECODE("%2a", "*"),
|
||
T_DECODE("%2b", "+"),
|
||
T_DECODE("%2c", ","),
|
||
T_DECODE("%2f", "/"),
|
||
T_DECODE("%3a", ":"),
|
||
T_DECODE("%3b", ";"),
|
||
T_DECODE("%3d", "="),
|
||
T_DECODE("%3f", "?"),
|
||
T_DECODE("%5b", "["),
|
||
T_DECODE("%5d", "]"),
|
||
|
||
/* too few characters after % */
|
||
T_DECODE_FAIL(EINVAL, "x%", "x"),
|
||
T_DECODE_FAIL(EINVAL, "x%0", "x"),
|
||
|
||
/* non-hex characters after % */
|
||
T_DECODE_FAIL(EINVAL, "x%0z", "x"),
|
||
T_DECODE_FAIL(EINVAL, "x%zz", "x"),
|
||
T_DECODE_FAIL(EINVAL, "x%z0", "x"),
|
||
|
||
/* input shorter than claimed */
|
||
T_SHORT_INPUT_ENC(reserved, reserved_upper),
|
||
T_SHORT_INPUT_DEC(reserved_lower, reserved),
|
||
|
||
/* input longer than claimed */
|
||
T_LONG_INPUT_ENC(mixed, 6, mixed_upper, 9),
|
||
T_LONG_INPUT_DEC(mixed_lower, 8, mixed, 7),
|
||
|
||
/* encoding into short buffer */
|
||
T_LONG_OUTPUT_ENC(mixed, mixed_upper, 6),
|
||
T_LONG_OUTPUT_ENC(mixed, mixed_upper, 7),
|
||
T_LONG_OUTPUT_ENC(mixed, mixed_upper, 8),
|
||
T_LONG_OUTPUT_ENC(mixed, mixed_upper, 9),
|
||
|
||
/* decoding into short buffer */
|
||
T_LONG_OUTPUT_DEC(unreserved, unreserved, 5),
|
||
T_LONG_OUTPUT_DEC(reserved_lower, reserved, 5),
|
||
|
||
/* various no-output scenarios (decoding) */
|
||
{
|
||
.desc = "decode (no output)",
|
||
.func = percent_decode,
|
||
.in = mixed_lower,
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = SIZE_MAX,
|
||
.olen = sizeof mixed,
|
||
/* success */
|
||
},
|
||
{
|
||
.desc = "decode (no output, invalid sequence)",
|
||
.func = percent_decode,
|
||
.in = "x%",
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = SIZE_MAX,
|
||
.olen = 2,
|
||
.ret = -1,
|
||
.err = EINVAL,
|
||
},
|
||
{
|
||
.desc = "decode (no space)",
|
||
.func = percent_decode,
|
||
.in = mixed_lower,
|
||
.ilen = SIZE_MAX,
|
||
.out = "",
|
||
.blen = 0,
|
||
.olen = sizeof mixed,
|
||
.ret = -1,
|
||
.err = ENOSPC,
|
||
},
|
||
{
|
||
.desc = "decode (no space, invalid sequence)",
|
||
.func = percent_decode,
|
||
.in = "x%",
|
||
.ilen = SIZE_MAX,
|
||
.out = "",
|
||
.blen = 0,
|
||
.olen = 2,
|
||
/* invalid sequence trumps no space */
|
||
.ret = -1,
|
||
.err = EINVAL,
|
||
},
|
||
{
|
||
.desc = "decode (no output, no space)",
|
||
.func = percent_decode,
|
||
.in = mixed_lower,
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = 0,
|
||
.olen = sizeof mixed,
|
||
/* success */
|
||
},
|
||
{
|
||
.desc = "decode (no output, no space, invalid sequence)",
|
||
.func = percent_decode,
|
||
.in = "x%",
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = 0,
|
||
.olen = 2,
|
||
/* invalid sequence trumps no space */
|
||
.ret = -1,
|
||
.err = EINVAL,
|
||
},
|
||
|
||
/* various no-output scenarios (encoding) */
|
||
{
|
||
.desc = "encode (no output)",
|
||
.func = percent_encode,
|
||
.in = mixed,
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = SIZE_MAX,
|
||
.olen = sizeof mixed_upper,
|
||
/* success */
|
||
},
|
||
{
|
||
.desc = "encode (no space)",
|
||
.func = percent_encode,
|
||
.in = mixed,
|
||
.ilen = SIZE_MAX,
|
||
.out = "",
|
||
.blen = 0,
|
||
.olen = sizeof mixed_upper,
|
||
.ret = -1,
|
||
.err = ENOSPC,
|
||
},
|
||
{
|
||
.desc = "encode (no output, no space)",
|
||
.func = percent_encode,
|
||
.in = mixed,
|
||
.ilen = SIZE_MAX,
|
||
.out = NULL,
|
||
.blen = 0,
|
||
.olen = sizeof mixed_upper,
|
||
/* success */
|
||
},
|
||
};
|
||
|
||
/*
|
||
* Encoding test function
|
||
*/
|
||
static int
|
||
t_rfc3986(char **desc CRYB_UNUSED, void *arg)
|
||
{
|
||
struct t_case *t = arg;
|
||
char buf[256];
|
||
size_t len;
|
||
int ret;
|
||
|
||
memset(buf, 0, sizeof buf);
|
||
errno = 0;
|
||
len = t->blen > sizeof buf ? sizeof buf : t->blen;
|
||
ret = t_compare_i(t->ret,
|
||
t->func(t->in, t->ilen, t->out ? buf : NULL, &len));
|
||
ret &= t_compare_sz(t->olen, len);
|
||
if (t->ret != 0 || errno != 0)
|
||
ret &= t_compare_i(t->err, errno);
|
||
if (t->out)
|
||
ret &= t_compare_strn(t->out, buf, t->blen - 1);
|
||
return (ret);
|
||
}
|
||
|
||
|
||
/***************************************************************************
|
||
* Boilerplate
|
||
*/
|
||
|
||
static int
|
||
t_prepare(int argc, char *argv[])
|
||
{
|
||
struct t_case *t;
|
||
int i, n;
|
||
|
||
(void)argc;
|
||
(void)argv;
|
||
n = sizeof t_cases / sizeof t_cases[0];
|
||
for (i = 0; i < n; ++i) {
|
||
t = &t_cases[i];
|
||
if (t->desc) {
|
||
t_add_test(t_rfc3986, t, "%s", t->desc);
|
||
} else if (t->ret == 0) {
|
||
t_add_test(t_rfc3986, t,
|
||
"%s \"%s\" -> \"%s\"",
|
||
t->func == percent_encode ? "encode" : "decode",
|
||
t->in, t->out);
|
||
} else {
|
||
t_add_test(t_rfc3986, t,
|
||
"%s \"%s\" (failure)",
|
||
t->func == percent_encode ? "encode" : "decode",
|
||
t->in);
|
||
}
|
||
}
|
||
return (0);
|
||
}
|
||
|
||
int
|
||
main(int argc, char *argv[])
|
||
{
|
||
|
||
t_main(t_prepare, NULL, argc, argv);
|
||
}
|