Files
libtorrent/test/test_torrent_info.cpp

1517 lines
47 KiB
C++

/*
Copyright (c) 2013-2022, Arvid Norberg
Copyright (c) 2016, 2018, Alden Torres
Copyright (c) 2017, Pavel Pimenov
Copyright (c) 2017-2019, Steven Siloti
Copyright (c) 2019, Andrei Kurushin
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* 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.
* Neither the name of the author nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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 "test.hpp"
#include "setup_transfer.hpp" // for load_file
#include "test_utils.hpp"
#include "libtorrent/file_storage.hpp"
#include "libtorrent/load_torrent.hpp"
#include "libtorrent/aux_/path.hpp"
#include "libtorrent/torrent_info.hpp"
#include "libtorrent/create_torrent.hpp"
#include "libtorrent/announce_entry.hpp"
#include "libtorrent/disk_interface.hpp" // for default_block_size
#include "libtorrent/aux_/escape_string.hpp" // for convert_path_to_posix
#include "libtorrent/piece_picker.hpp"
#include "libtorrent/hex.hpp" // to_hex
#include "libtorrent/write_resume_data.hpp" // write_torrent_file
#include <iostream>
using namespace lt;
#ifndef TORRENT_DISABLE_MUTABLE_TORRENTS
TORRENT_TEST(mutable_torrents)
{
file_storage fs;
fs.add_file("test/temporary.txt", 0x4000);
lt::create_torrent t(fs, 0x4000);
// calculate the hash for all pieces
for (auto const i : fs.piece_range())
t.set_hash(i, sha1_hash::max());
t.add_collection("collection1");
t.add_collection("collection2");
t.add_similar_torrent(sha1_hash("abababababababababab"));
t.add_similar_torrent(sha1_hash("babababababababababa"));
entry tor = t.generate();
std::vector<char> tmp;
bencode(std::back_inserter(tmp), tor);
torrent_info ti(tmp, from_span);
std::vector<sha1_hash> similar;
similar.push_back(sha1_hash("abababababababababab"));
similar.push_back(sha1_hash("babababababababababa"));
std::vector<std::string> collections;
collections.push_back("collection1");
collections.push_back("collection2");
TEST_CHECK(similar == ti.similar_torrents());
TEST_CHECK(collections == ti.collections());
}
#endif
namespace {
struct test_torrent_t
{
test_torrent_t(char const* f, std::function<void(torrent_info*)> t = {}) // NOLINT
: file(f), test(std::move(t)) {}
char const* file;
std::function<void(torrent_info*)> test;
};
using namespace lt;
#ifdef TORRENT_WINDOWS
#define SEPARATOR "\\"
#else
#define SEPARATOR "/"
#endif
static test_torrent_t const test_torrents[] =
{
{ "base.torrent"},
{ "empty_path.torrent" },
{ "parent_path.torrent" },
{ "hidden_parent_path.torrent" },
{ "single_multi_file.torrent" },
{ "slash_path.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{0}), "temp" SEPARATOR "bar");
}
},
{ "slash_path2.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{0}), "temp" SEPARATOR "abc....def" SEPARATOR "bar");
}
},
{ "slash_path3.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{0}), "temp....abc");
}
},
{ "backslash_path.torrent" },
{ "url_list.torrent" },
{ "url_list2.torrent" },
{ "url_list3.torrent" },
{ "httpseed.torrent" },
{ "empty_httpseed.torrent" },
{ "long_name.torrent" },
{ "whitespace_url.torrent", [](torrent_info const* ti) {
// make sure we trimmed the url
TEST_CHECK(ti->trackers().size() > 0);
if (ti->trackers().size() > 0)
TEST_CHECK(ti->trackers()[0].url == "udp://test.com/announce");
}
},
{ "duplicate_files.torrent", [](torrent_info const* ti) {
// make sure we disambiguated the files
TEST_EQUAL(ti->num_files(), 2);
TEST_CHECK(ti->files().file_path(file_index_t{0}) == combine_path(combine_path("temp", "foo"), "bar.txt"));
TEST_CHECK(ti->files().file_path(file_index_t{1}) == combine_path(combine_path("temp", "foo"), "bar.1.txt"));
}
},
{ "pad_file.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
TEST_EQUAL(bool(ti->files().file_flags(file_index_t{0}) & file_storage::flag_pad_file), false);
TEST_EQUAL(bool(ti->files().file_flags(file_index_t{1}) & file_storage::flag_pad_file), true);
}
},
{ "creation_date.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->creation_date(), 1234567);
}
},
{ "no_creation_date.torrent", [](torrent_info const* ti) {
TEST_CHECK(!ti->creation_date());
}
},
{ "url_seed.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 1);
TEST_EQUAL(ti->web_seeds()[0].url, "http://test.com/file");
#if TORRENT_ABI_VERSION == 1
TEST_EQUAL(ti->http_seeds().size(), 0);
TEST_EQUAL(ti->url_seeds().size(), 1);
TEST_EQUAL(ti->url_seeds()[0], "http://test.com/file");
#endif
}
},
{ "url_seed_multi.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 1);
TEST_EQUAL(ti->web_seeds()[0].url, "http://test.com/file/");
#if TORRENT_ABI_VERSION == 1
TEST_EQUAL(ti->http_seeds().size(), 0);
TEST_EQUAL(ti->url_seeds().size(), 1);
TEST_EQUAL(ti->url_seeds()[0], "http://test.com/file/");
#endif
}
},
{ "url_seed_multi_single_file.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 1);
TEST_EQUAL(ti->web_seeds()[0].url, "http://test.com/file/temp/foo/bar.txt");
}
},
{ "url_seed_multi_space.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 1);
TEST_EQUAL(ti->web_seeds()[0].url, "http://test.com/test%20file/foo%20bar/");
#if TORRENT_ABI_VERSION == 1
TEST_EQUAL(ti->http_seeds().size(), 0);
TEST_EQUAL(ti->url_seeds().size(), 1);
TEST_EQUAL(ti->url_seeds()[0], "http://test.com/test%20file/foo%20bar/");
#endif
}
},
{ "url_seed_multi_space_nolist.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 1);
TEST_EQUAL(ti->web_seeds()[0].url, "http://test.com/test%20file/foo%20bar/");
#if TORRENT_ABI_VERSION == 1
TEST_EQUAL(ti->http_seeds().size(), 0);
TEST_EQUAL(ti->url_seeds().size(), 1);
TEST_EQUAL(ti->url_seeds()[0], "http://test.com/test%20file/foo%20bar/");
#endif
}
},
{ "empty_path_multi.torrent" },
{ "duplicate_web_seeds.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->web_seeds().size(), 3);
}
},
{ "invalid_name2.torrent", [](torrent_info const* ti) {
// if, after all invalid characters are removed from the name, it ends up
// being empty, it's set to the info-hash. Some torrents also have an empty name
// in which case it's also set to the info-hash
TEST_EQUAL(ti->name(), "b61560c2918f463768cd122b6d2fdd47b77bdb35");
}
},
{ "invalid_name3.torrent", [](torrent_info const* ti) {
// windows does not allow trailing spaces in filenames
#ifdef TORRENT_WINDOWS
TEST_EQUAL(ti->name(), "foobar");
#else
TEST_EQUAL(ti->name(), "foobar ");
#endif
}
},
{ "symlink1.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
TEST_EQUAL(ti->files().symlink(file_index_t{1}), "temp" SEPARATOR "a" SEPARATOR "b" SEPARATOR "bar");
}
},
{ "symlink2.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 5);
TEST_EQUAL(ti->files().symlink(file_index_t{0}), "Some.framework" SEPARATOR "Versions" SEPARATOR "A" SEPARATOR "SDL2");
TEST_EQUAL(ti->files().symlink(file_index_t{4}), "Some.framework" SEPARATOR "Versions" SEPARATOR "A");
}
},
{ "unordered.torrent" },
{ "symlink_zero_size.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
TEST_EQUAL(ti->files().symlink(1_file), "temp" SEPARATOR "a" SEPARATOR "b" SEPARATOR "bar");
}
},
{ "pad_file_no_path.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
TEST_EQUAL(ti->files().file_path(file_index_t{1}), combine_path(".pad", "2124"));
}
},
{ "large.torrent" },
{ "absolute_filename.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
TEST_EQUAL(ti->files().file_path(file_index_t{0}), combine_path("temp", "abcde"));
TEST_EQUAL(ti->files().file_path(file_index_t{1}), combine_path("temp", "foobar"));
}
},
{ "invalid_filename.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 2);
}
},
{ "invalid_filename2.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 3);
}
},
{ "overlapping_symlinks.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->num_files() > 3);
TEST_EQUAL(ti->files().symlink(file_index_t{0}), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "Headers");
TEST_EQUAL(ti->files().symlink(file_index_t{1}), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "Resources");
TEST_EQUAL(ti->files().symlink(file_index_t{2}), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "SDL2");
}
},
{ "v2.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{ 0 }), "test64K"_sv);
TEST_EQUAL(ti->files().file_size(file_index_t{ 0 }), 65536);
TEST_EQUAL(aux::to_hex(ti->files().root(file_index_t{ 0 })), "60aae9c7b428f87e0713e88229e18f0adf12cd7b22a0dd8a92bb2485eb7af242"_sv);
TEST_EQUAL(ti->info_hashes().has_v1(), true);
TEST_EQUAL(ti->info_hashes().has_v2(), true);
TEST_EQUAL(aux::to_hex(ti->info_hashes().v2), "597b180c1a170a585dfc5e85d834d69013ceda174b8f357d5bb1a0ca509faf0a"_sv);
TEST_CHECK(ti->v2());
TEST_CHECK(ti->v1());
}
},
{ "v2_multipiece_file.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{ 0 }), "test1MB"_sv);
TEST_EQUAL(ti->files().file_size(file_index_t{ 0 }), 1048576);
TEST_EQUAL(aux::to_hex(ti->files().root(file_index_t{ 0 })), "515ea9181744b817744ded9d2e8e9dc6a8450c0b0c52e24b5077f302ffbd9008"_sv);
TEST_EQUAL(ti->info_hashes().has_v1(), true);
TEST_EQUAL(ti->info_hashes().has_v2(), true);
TEST_EQUAL(aux::to_hex(ti->info_hashes().v2), "108ac2c3718ce722e6896edc56c4afa98f1d711ecaace7aad74fca418ebd03de"_sv);
}
},
{ "v2_only.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{ 0 }), "test1MB"_sv);
TEST_EQUAL(ti->files().file_size(file_index_t{ 0 }), 1048576);
TEST_EQUAL(aux::to_hex(ti->files().root(file_index_t{ 0 })), "515ea9181744b817744ded9d2e8e9dc6a8450c0b0c52e24b5077f302ffbd9008"_sv);
TEST_EQUAL(ti->info_hashes().has_v1(), false);
TEST_EQUAL(ti->info_hashes().has_v2(), true);
TEST_EQUAL(aux::to_hex(ti->info_hashes().v2), "95e04d0c4bad94ab206efa884666fd89777dbe4f7bd9945af1829037a85c6192"_sv);
TEST_CHECK(ti->v2());
TEST_CHECK(!ti->v1());
}
},
{ "v2_invalid_filename.torrent", [](torrent_info const* ti) {
TEST_EQUAL(ti->num_files(), 1);
TEST_EQUAL(ti->files().file_path(file_index_t{0}), "_estMB"_sv);
}
},
{ "v2_multiple_files.torrent", [](torrent_info* ti) {
TEST_EQUAL(ti->v2_piece_hashes_verified(), true);
TEST_EQUAL(ti->num_files(), 5);
TEST_CHECK(ti->v2());
ti->free_piece_layers();
TEST_CHECK(ti->v2());
TEST_EQUAL(ti->v2_piece_hashes_verified(), false);
}
},
{ "v2_symlinks.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->num_files() > 3);
TEST_EQUAL(ti->files().symlink(0_file), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "Headers");
TEST_EQUAL(ti->files().symlink(1_file), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "Resources");
TEST_EQUAL(ti->files().symlink(2_file), "SDL2.framework" SEPARATOR "Versions" SEPARATOR "Current" SEPARATOR "SDL2");
}
},
{ "v2_hybrid.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "empty-files-1.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "empty-files-2.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "empty-files-3.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "empty-files-4.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "empty-files-5.torrent", [](torrent_info const* ti) {
TEST_CHECK(ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "v2_no_piece_layers.torrent", [](torrent_info const* ti) {
// it's OK to not have a piece layers field.
// It's just like adding a magnet link
TEST_CHECK(!ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "v2_incomplete_piece_layer.torrent", [](torrent_info const* ti) {
// it's OK for some files to not have a piece layer.
// It's just like adding a magnet link
TEST_CHECK(!ti->info_hashes().has_v1());
TEST_CHECK(ti->info_hashes().has_v2());
}
},
{ "similar.torrent", [](torrent_info const* ti) {
TEST_CHECK((ti->similar_torrents() == std::vector<lt::sha1_hash>{sha1_hash("aaaaaaaaaaaaaaaaaaaa")}));
}
},
{ "similar2.torrent", [](torrent_info const* ti) {
TEST_CHECK((ti->similar_torrents() == std::vector<lt::sha1_hash>{sha1_hash("aaaaaaaaaaaaaaaaaaaa")}));
}
},
{ "collection.torrent", [](torrent_info const* ti) {
TEST_CHECK((ti->collections() == std::vector<std::string>{"bar", "foo"}));
}
},
{ "collection2.torrent", [](torrent_info const* ti) {
TEST_CHECK((ti->collections() == std::vector<std::string>{"bar", "foo"}));
}
},
{ "dht_nodes.torrent", [](torrent_info const* ti) {
using np = std::pair<std::string, int>;
TEST_CHECK((ti->nodes() == std::vector<np>{np("127.0.0.1", 6881), np("192.168.1.1", 6881)}));
}
},
};
struct test_failing_torrent_t
{
char const* file;
error_code error; // the expected error
};
test_failing_torrent_t test_error_torrents[] =
{
{ "missing_piece_len.torrent", errors::torrent_missing_piece_length },
{ "invalid_piece_len.torrent", errors::torrent_missing_piece_length },
{ "negative_piece_len.torrent", errors::torrent_missing_piece_length },
{ "no_name.torrent", errors::torrent_missing_name },
{ "bad_name.torrent", errors::torrent_missing_name },
{ "invalid_name.torrent", errors::torrent_missing_name },
{ "invalid_info.torrent", errors::torrent_missing_info },
{ "string.torrent", errors::torrent_is_no_dict },
{ "negative_size.torrent", errors::torrent_invalid_length },
{ "negative_file_size.torrent", errors::torrent_invalid_length },
{ "invalid_path_list.torrent", errors::torrent_invalid_name},
{ "missing_path_list.torrent", errors::torrent_missing_name },
{ "invalid_pieces.torrent", errors::torrent_missing_pieces },
{ "unaligned_pieces.torrent", errors::torrent_invalid_hashes },
{ "invalid_file_size.torrent", errors::torrent_invalid_length },
{ "invalid_symlink.torrent", errors::torrent_invalid_name },
{ "many_pieces.torrent", errors::too_many_pieces_in_torrent },
{ "no_files.torrent", errors::no_files_in_torrent},
{ "zero.torrent", errors::torrent_invalid_length},
{ "zero2.torrent", errors::torrent_invalid_length},
{ "v2_mismatching_metadata.torrent", errors::torrent_inconsistent_files},
{ "v2_no_power2_piece.torrent", errors::torrent_missing_piece_length},
{ "v2_invalid_file.torrent", errors::torrent_file_parse_failed},
{ "v2_deep_recursion.torrent", bdecode_errors::depth_exceeded},
{ "v2_non_multiple_piece_layer.torrent", errors::torrent_invalid_piece_layer},
{ "v2_piece_layer_invalid_file_hash.torrent", errors::torrent_invalid_piece_layer},
{ "v2_invalid_piece_layer.torrent", errors::torrent_invalid_piece_layer},
{ "v2_invalid_piece_layer_size.torrent", errors::torrent_invalid_piece_layer},
{ "v2_bad_file_alignment.torrent", errors::torrent_inconsistent_files},
{ "v2_unordered_files.torrent", errors::invalid_bencoding},
{ "v2_overlong_integer.torrent", errors::invalid_bencoding},
{ "v2_missing_file_root_invalid_symlink.torrent", errors::torrent_missing_pieces_root},
{ "v2_large_file.torrent", errors::torrent_invalid_length},
{ "v2_large_offset.torrent", errors::too_many_pieces_in_torrent},
{ "v2_piece_size.torrent", errors::torrent_missing_piece_length},
{ "v2_invalid_pad_file.torrent", errors::torrent_invalid_pad_file},
{ "v2_zero_root.torrent", errors::torrent_missing_pieces_root},
{ "v2_zero_root_small.torrent", errors::torrent_missing_pieces_root},
};
} // anonymous namespace
// TODO: test remap_files
// TODO: torrent with 'p' (padfile) attribute
// TODO: torrent with 'h' (hidden) attribute
// TODO: torrent with 'x' (executable) attribute
// TODO: torrent with 'l' (symlink) attribute
// TODO: torrent with multiple trackers in multiple tiers, making sure we
// shuffle them (how do you test shuffling?, load it multiple times and make
// sure it's in different order at least once)
// TODO: torrents with a zero-length name
// TODO: torrent with a non-dictionary info-section
// TODO: torrents with DHT nodes
// TODO: torrent with url-list as a single string
// TODO: torrent with http seed as a single string
// TODO: torrent with a comment
// TODO: torrent with an SSL cert
// TODO: torrent with attributes (executable and hidden)
// TODO: torrent_info constructor that takes an invalid bencoded buffer
// TODO: verify_encoding with a string that triggers character replacement
TORRENT_TEST(add_tracker)
{
torrent_info ti(info_hash_t(sha1_hash(" ")));
TEST_EQUAL(ti.trackers().size(), 0);
ti.add_tracker("http://test.com/announce");
TEST_EQUAL(ti.trackers().size(), 1);
announce_entry ae = ti.trackers()[0];
TEST_EQUAL(ae.url, "http://test.com/announce");
ti.clear_trackers();
TEST_EQUAL(ti.trackers().size(), 0);
}
TORRENT_TEST(url_list_and_httpseeds)
{
entry info;
info["pieces"] = "aaaaaaaaaaaaaaaaaaaa";
info["name.utf-8"] = "test1";
info["name"] = "test__";
info["piece length"] = 16 * 1024;
info["length"] = 3245;
entry::list_type l;
l.push_back(entry("http://foo.com/bar1"));
l.push_back(entry("http://foo.com/bar1"));
l.push_back(entry("http://foo.com/bar2"));
entry const e(l);
entry torrent;
torrent["url-list"] = e;
torrent["httpseeds"] = e;
torrent["info"] = info;
std::vector<char> buf;
bencode(std::back_inserter(buf), torrent);
torrent_info ti(buf, from_span);
TEST_EQUAL(ti.web_seeds().size(), 4);
}
TORRENT_TEST(add_url_seed)
{
torrent_info ti(info_hash_t(sha1_hash(" ")));
TEST_EQUAL(ti.web_seeds().size(), 0);
ti.add_url_seed("http://test.com");
TEST_EQUAL(ti.web_seeds().size(), 1);
web_seed_entry we = ti.web_seeds()[0];
TEST_EQUAL(we.type, web_seed_entry::url_seed);
TEST_EQUAL(we.url, "http://test.com");
}
TORRENT_TEST(add_http_seed)
{
torrent_info ti(info_hash_t(sha1_hash(" ")));
TEST_EQUAL(ti.web_seeds().size(), 0);
ti.add_http_seed("http://test.com");
TEST_EQUAL(ti.web_seeds().size(), 1);
web_seed_entry we = ti.web_seeds()[0];
TEST_EQUAL(we.type, web_seed_entry::http_seed);
TEST_EQUAL(we.url, "http://test.com");
}
TORRENT_TEST(set_web_seeds)
{
torrent_info ti(info_hash_t(sha1_hash(" ")));
TEST_EQUAL(ti.web_seeds().size(), 0);
std::vector<web_seed_entry> seeds;
web_seed_entry e1("http://test1.com", web_seed_entry::url_seed);
seeds.push_back(e1);
web_seed_entry e2("http://test2com", web_seed_entry::http_seed);
seeds.push_back(e2);
ti.set_web_seeds(seeds);
TEST_EQUAL(ti.web_seeds().size(), 2);
TEST_CHECK(ti.web_seeds() == seeds);
}
TORRENT_TEST(sanitize_path_truncate)
{
using lt::aux::sanitize_append_path_element;
std::string path;
sanitize_append_path_element(path,
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_");
sanitize_append_path_element(path,
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcde.test");
TEST_EQUAL(path,
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_" SEPARATOR
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_.test");
}
TORRENT_TEST(sanitize_path_truncate_utf)
{
using lt::aux::sanitize_append_path_element;
std::string path;
// msvc doesn't like unicode string literals, so we encode it as UTF-8 explicitly
sanitize_append_path_element(path,
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi" "\xE2" "\x80" "\x94" "abcde.jpg");
TEST_EQUAL(path,
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi_abcdefghi_"
"abcdefghi_abcdefghi_abcdefghi_abcdefghi" "\xE2" "\x80" "\x94" ".jpg");
}
TORRENT_TEST(sanitize_path_trailing_dots)
{
std::string path;
using lt::aux::sanitize_append_path_element;
sanitize_append_path_element(path, "a");
sanitize_append_path_element(path, "abc...");
sanitize_append_path_element(path, "c");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "a" SEPARATOR "abc" SEPARATOR "c");
#else
TEST_EQUAL(path, "a" SEPARATOR "abc..." SEPARATOR "c");
#endif
path.clear();
sanitize_append_path_element(path, "abc...");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "abc");
#else
TEST_EQUAL(path, "abc...");
#endif
path.clear();
sanitize_append_path_element(path, "abc.");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "abc");
#else
TEST_EQUAL(path, "abc.");
#endif
path.clear();
sanitize_append_path_element(path, "a. . .");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "a");
#else
TEST_EQUAL(path, "a. . .");
#endif
}
TORRENT_TEST(sanitize_path_trailing_spaces)
{
using lt::aux::sanitize_append_path_element;
std::string path;
sanitize_append_path_element(path, "a");
sanitize_append_path_element(path, "abc ");
sanitize_append_path_element(path, "c");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "a" SEPARATOR "abc" SEPARATOR "c");
#else
TEST_EQUAL(path, "a" SEPARATOR "abc " SEPARATOR "c");
#endif
path.clear();
sanitize_append_path_element(path, "abc ");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "abc");
#else
TEST_EQUAL(path, "abc ");
#endif
path.clear();
sanitize_append_path_element(path, "abc ");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "abc");
#else
TEST_EQUAL(path, "abc ");
#endif
}
TORRENT_TEST(sanitize_path)
{
using lt::aux::sanitize_append_path_element;
std::string path;
sanitize_append_path_element(path, "\0\0\xed\0\x80");
TEST_EQUAL(path, "_");
path.clear();
sanitize_append_path_element(path, "/a/");
sanitize_append_path_element(path, "b");
sanitize_append_path_element(path, "c");
TEST_EQUAL(path, "a" SEPARATOR "b" SEPARATOR "c");
path.clear();
sanitize_append_path_element(path, "a...b");
TEST_EQUAL(path, "a...b");
path.clear();
sanitize_append_path_element(path, "a");
sanitize_append_path_element(path, "..");
sanitize_append_path_element(path, "c");
TEST_EQUAL(path, "a" SEPARATOR "c");
path.clear();
sanitize_append_path_element(path, "a");
sanitize_append_path_element(path, "..");
TEST_EQUAL(path, "a");
path.clear();
sanitize_append_path_element(path, "/..");
sanitize_append_path_element(path, ".");
sanitize_append_path_element(path, "c");
TEST_EQUAL(path, "c");
path.clear();
sanitize_append_path_element(path, "dev:");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "dev_");
#else
TEST_EQUAL(path, "dev:");
#endif
path.clear();
sanitize_append_path_element(path, "c:");
sanitize_append_path_element(path, "b");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "c_" SEPARATOR "b");
#else
TEST_EQUAL(path, "c:" SEPARATOR "b");
#endif
path.clear();
sanitize_append_path_element(path, "c:");
sanitize_append_path_element(path, ".");
sanitize_append_path_element(path, "c");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "c_" SEPARATOR "c");
#else
TEST_EQUAL(path, "c:" SEPARATOR "c");
#endif
path.clear();
sanitize_append_path_element(path, "\\c");
sanitize_append_path_element(path, ".");
sanitize_append_path_element(path, "c");
TEST_EQUAL(path, "c" SEPARATOR "c");
path.clear();
sanitize_append_path_element(path, "\b");
TEST_EQUAL(path, "_");
path.clear();
sanitize_append_path_element(path, "\b");
sanitize_append_path_element(path, "filename");
TEST_EQUAL(path, "_" SEPARATOR "filename");
path.clear();
sanitize_append_path_element(path, "filename");
sanitize_append_path_element(path, "\b");
TEST_EQUAL(path, "filename" SEPARATOR "_");
path.clear();
sanitize_append_path_element(path, "abc");
sanitize_append_path_element(path, "");
TEST_EQUAL(path, "abc" SEPARATOR "_");
path.clear();
sanitize_append_path_element(path, "abc");
sanitize_append_path_element(path, " ");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "abc");
#else
TEST_EQUAL(path, "abc" SEPARATOR " ");
#endif
path.clear();
sanitize_append_path_element(path, "");
sanitize_append_path_element(path, "abc");
TEST_EQUAL(path, "_" SEPARATOR "abc");
path.clear();
sanitize_append_path_element(path, "\b?filename=4");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "__filename=4");
#else
TEST_EQUAL(path, "_?filename=4");
#endif
path.clear();
sanitize_append_path_element(path, "filename=4");
TEST_EQUAL(path, "filename=4");
// valid 2-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xc2\xa1");
TEST_EQUAL(path, "filename\xc2\xa1");
// truncated 2-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xc2");
TEST_EQUAL(path, "filename_");
// valid 3-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xe2\x9f\xb9");
TEST_EQUAL(path, "filename\xe2\x9f\xb9");
// truncated 3-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xe2\x9f");
TEST_EQUAL(path, "filename_");
// truncated 3-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xe2");
TEST_EQUAL(path, "filename_");
// valid 4-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xf0\x9f\x92\x88");
TEST_EQUAL(path, "filename\xf0\x9f\x92\x88");
// truncated 4-byte sequence
path.clear();
sanitize_append_path_element(path, "filename\xf0\x9f\x92");
TEST_EQUAL(path, "filename_");
// 5-byte utf-8 sequence (not allowed)
path.clear();
sanitize_append_path_element(path, "filename\xf8\x9f\x9f\x9f\x9f" "foobar");
TEST_EQUAL(path, "filename_foobar");
// redundant (overlong) 2-byte sequence
// ascii code 0x2e encoded with a leading 0
path.clear();
sanitize_append_path_element(path, "filename\xc0\xae");
TEST_EQUAL(path, "filename_");
// redundant (overlong) 3-byte sequence
// ascii code 0x2e encoded with two leading 0s
path.clear();
sanitize_append_path_element(path, "filename\xe0\x80\xae");
TEST_EQUAL(path, "filename_");
// redundant (overlong) 4-byte sequence
// ascii code 0x2e encoded with three leading 0s
path.clear();
sanitize_append_path_element(path, "filename\xf0\x80\x80\xae");
TEST_EQUAL(path, "filename_");
// a filename where every character is filtered is not replaced by an understcore
path.clear();
sanitize_append_path_element(path, "//\\");
TEST_EQUAL(path, "");
// make sure suspicious unicode characters are filtered out
path.clear();
// that's utf-8 for U+200e LEFT-TO-RIGHT MARK
sanitize_append_path_element(path, "foo\xe2\x80\x8e" "bar");
TEST_EQUAL(path, "foobar");
// make sure suspicious unicode characters are filtered out
path.clear();
// that's utf-8 for U+202b RIGHT-TO-LEFT EMBEDDING
sanitize_append_path_element(path, "foo\xe2\x80\xab" "bar");
TEST_EQUAL(path, "foobar");
}
TORRENT_TEST(sanitize_path_zeroes)
{
using lt::aux::sanitize_append_path_element;
std::string path;
sanitize_append_path_element(path, "\0foo");
TEST_EQUAL(path, "_");
path.clear();
sanitize_append_path_element(path, "\0\0\0\0");
TEST_EQUAL(path, "_");
}
TORRENT_TEST(sanitize_path_colon)
{
using lt::aux::sanitize_append_path_element;
std::string path;
sanitize_append_path_element(path, "foo:bar");
#ifdef TORRENT_WINDOWS
TEST_EQUAL(path, "foo_bar");
#else
TEST_EQUAL(path, "foo:bar");
#endif
}
TORRENT_TEST(verify_encoding)
{
using aux::verify_encoding;
// verify_encoding
std::string test = "\b?filename=4";
TEST_CHECK(verify_encoding(test));
TEST_CHECK(test == "\b?filename=4");
test = "filename=4";
TEST_CHECK(verify_encoding(test));
TEST_CHECK(test == "filename=4");
// valid 2-byte sequence
test = "filename\xc2\xa1";
TEST_CHECK(verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename\xc2\xa1");
// truncated 2-byte sequence
test = "filename\xc2";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// valid 3-byte sequence
test = "filename\xe2\x9f\xb9";
TEST_CHECK(verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename\xe2\x9f\xb9");
// truncated 3-byte sequence
test = "filename\xe2\x9f";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// truncated 3-byte sequence
test = "filename\xe2";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// valid 4-byte sequence
test = "filename\xf0\x9f\x92\x88";
TEST_CHECK(verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename\xf0\x9f\x92\x88");
// truncated 4-byte sequence
test = "filename\xf0\x9f\x92";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// 5-byte utf-8 sequence (not allowed)
test = "filename\xf8\x9f\x9f\x9f\x9f""foobar";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_foobar");
// redundant (overlong) 2-byte sequence
// ascii code 0x2e encoded with a leading 0
test = "filename\xc0\xae";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// redundant (overlong) 3-byte sequence
// ascii code 0x2e encoded with two leading 0s
test = "filename\xe0\x80\xae";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// redundant (overlong) 4-byte sequence
// ascii code 0x2e encoded with three leading 0s
test = "filename\xf0\x80\x80\xae";
TEST_CHECK(!verify_encoding(test));
std::printf("%s\n", test.c_str());
TEST_CHECK(test == "filename_");
// missing byte header
test = "filename\xed\0\x80";
TEST_CHECK(!verify_encoding(test));
fprintf(stdout, "%s\n", test.c_str());
TEST_CHECK(test == "filename_");
}
namespace {
void sanity_check(std::shared_ptr<torrent_info> const& ti)
{
// construct a piece_picker to get some more test coverage. Perhaps
// loading the torrent is fine, but if we can't construct a piece_picker
// for it, it's still no good.
piece_picker pp(ti->total_size(), ti->piece_length());
TEST_CHECK(ti->piece_length() < std::numeric_limits<int>::max() / 2);
TEST_EQUAL(ti->v1(), ti->info_hashes().has_v1());
TEST_EQUAL(ti->v2(), ti->info_hashes().has_v2());
}
}
TORRENT_TEST(parse_torrents)
{
// test torrent parsing
entry info;
info["pieces"] = "aaaaaaaaaaaaaaaaaaaa";
info["name.utf-8"] = "test1";
info["name"] = "test__";
info["piece length"] = 16 * 1024;
info["length"] = 3245;
entry torrent;
torrent["info"] = info;
std::vector<char> buf;
bencode(std::back_inserter(buf), torrent);
torrent_info ti1(buf, from_span);
std::cout << ti1.name() << std::endl;
TEST_CHECK(ti1.name() == "test1");
#ifdef TORRENT_WINDOWS
info["name.utf-8"] = "c:/test1/test2/test3";
#else
info["name.utf-8"] = "/test1/test2/test3";
#endif
torrent["info"] = info;
buf.clear();
bencode(std::back_inserter(buf), torrent);
torrent_info ti2(buf, from_span);
std::cout << ti2.name() << std::endl;
#ifdef TORRENT_WINDOWS
TEST_EQUAL(ti2.name(), "c_test1test2test3");
#else
TEST_EQUAL(ti2.name(), "test1test2test3");
#endif
info["name.utf-8"] = "test2/../test3/.././../../test4";
torrent["info"] = info;
buf.clear();
bencode(std::back_inserter(buf), torrent);
torrent_info ti3(buf, from_span);
std::cout << ti3.name() << std::endl;
TEST_EQUAL(ti3.name(), "test2..test3.......test4");
std::string root_dir = parent_path(current_working_directory());
for (auto const& t : test_torrents)
{
std::printf("loading %s\n", t.file);
std::string filename = combine_path(combine_path(root_dir, "test_torrents")
, t.file);
error_code ec;
auto ti = std::make_shared<torrent_info>(filename, ec);
TEST_CHECK(!ec);
if (ec) std::printf(" -> failed %s\n", ec.message().c_str());
sanity_check(ti);
add_torrent_params atp = load_torrent_file(filename);
TEST_CHECK(atp.info_hashes == ti->info_hashes());
sanity_check(atp.ti);
// trackers are loaded into atp.trackers
TEST_CHECK(atp.ti->trackers().empty());
// web seeds are loaded into atp.trackers
TEST_CHECK(atp.ti->web_seeds().empty());
// piece layers are loaded into atp.merkle_trees and
// atp.merkle_trees_mask
TEST_CHECK(!atp.ti->v2_piece_hashes_verified());
if (t.test) t.test(ti.get());
file_storage const& fs = ti->files();
for (file_index_t const idx : fs.file_range())
{
piece_index_t const first = ti->map_file(idx, 0, 0).piece;
piece_index_t const last = ti->map_file(idx, std::max(fs.file_size(idx)-1, std::int64_t(0)), 0).piece;
file_flags_t const flags = fs.file_flags(idx);
sha1_hash const ih = fs.hash(idx);
std::printf(" %11" PRId64 " %c%c%c%c [ %4d, %4d ] %7u %s %s %s%s\n"
, fs.file_size(idx)
, (flags & file_storage::flag_pad_file)?'p':'-'
, (flags & file_storage::flag_executable)?'x':'-'
, (flags & file_storage::flag_hidden)?'h':'-'
, (flags & file_storage::flag_symlink)?'l':'-'
, static_cast<int>(first), static_cast<int>(last)
, std::uint32_t(fs.mtime(idx))
, ih != sha1_hash(nullptr) ? aux::to_hex(ih).c_str() : ""
, fs.file_path(idx).c_str()
, flags & file_storage::flag_symlink ? "-> ": ""
, flags & file_storage::flag_symlink ? fs.symlink(idx).c_str() : "");
}
}
}
TORRENT_TEST(parse_invalid_torrents)
{
std::string const root_dir = parent_path(current_working_directory());
for (auto const& e : test_error_torrents)
{
error_code ec;
std::printf("loading %s\n", e.file);
std::string const filename = combine_path(combine_path(root_dir, "test_torrents")
, e.file);
auto ti = std::make_shared<torrent_info>(filename, ec);
std::printf("E: \"%s\"\nexpected: \"%s\"\n", ec.message().c_str()
, e.error.message().c_str());
TEST_EQUAL(ec.message(), e.error.message());
TEST_EQUAL(ti->is_valid(), false);
try
{
add_torrent_params atp = load_torrent_file(filename);
}
catch (system_error const& err)
{
std::printf("E: \"%s\"\nexpected: \"%s\"\n", err.code().message().c_str()
, e.error.message().c_str());
TEST_EQUAL(err.code().message(), e.error.message());
}
}
}
namespace {
struct file_t
{
std::string filename;
int size;
file_flags_t flags;
string_view expected_filename;
};
std::vector<lt::aux::vector<file_t, lt::file_index_t>> const test_cases
{
{
{"test/temporary.txt", 0x4000, {}, "test/temporary.txt"},
{"test/Temporary.txt", 0x4000, {}, "test/Temporary.1.txt"},
{"test/TeMPorArY.txT", 0x4000, {}, "test/TeMPorArY.2.txT"},
// a file with the same name in a seprate directory is fine
{"test/test/TEMPORARY.TXT", 0x4000, {}, "test/test/TEMPORARY.TXT"},
},
{
{"test/b.exe", 0x4000, {}, "test/b.exe"},
// duplicate of b.exe
{"test/B.ExE", 0x4000, {}, "test/B.1.ExE"},
// duplicate of b.exe
{"test/B.exe", 0x4000, {}, "test/B.2.exe"},
{"test/filler", 0x4000, {}, "test/filler"},
},
{
{"test/a/b/c/d/e/f/g/h/i/j/k/l/m", 0x4000, {}, "test/a/b/c/d/e/f/g/h/i/j/k/l/m"},
{"test/a", 0x4000, {}, "test/a.1"},
{"test/a/b", 0x4000, {}, "test/a/b.1"},
{"test/a/b/c", 0x4000, {}, "test/a/b/c.1"},
{"test/a/b/c/d", 0x4000, {}, "test/a/b/c/d.1"},
{"test/a/b/c/d/e", 0x4000, {}, "test/a/b/c/d/e.1"},
{"test/a/b/c/d/e/f", 0x4000, {}, "test/a/b/c/d/e/f.1"},
{"test/a/b/c/d/e/f/g", 0x4000, {}, "test/a/b/c/d/e/f/g.1"},
{"test/a/b/c/d/e/f/g/h", 0x4000, {}, "test/a/b/c/d/e/f/g/h.1"},
{"test/a/b/c/d/e/f/g/h/i", 0x4000, {}, "test/a/b/c/d/e/f/g/h/i.1"},
{"test/a/b/c/d/e/f/g/h/i/j", 0x4000, {}, "test/a/b/c/d/e/f/g/h/i/j.1"},
},
{
// it doesn't matter whether the file comes before the directory,
// directories take precedence
{"test/a", 0x4000, {}, "test/a.1"},
{"test/a/b", 0x4000, {}, "test/a/b"},
},
{
{"test/A/tmp", 0x4000, {}, "test/A/tmp"},
// a file may not have the same name as a directory
{"test/a", 0x4000, {}, "test/a.1"},
// duplicate of directory a
{"test/A", 0x4000, {}, "test/A.2"},
{"test/filler", 0x4000, {}, "test/filler"},
},
{
// a subset of this path collides with the next filename
{"test/long/path/name/that/collides", 0x4000, {}, "test/long/path/name/that/collides"},
// so this file needs to be renamed, to not collide with the path name
{"test/long/path", 0x4000, {}, "test/long/path.1"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/filler-2", 0x4000, {}, "test/filler-2"},
},
{
// pad files are allowed to collide, as long as they have the same size
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234"},
{"test/filler-2", 0x4000, {}, "test/filler-2"},
},
{
// pad files of different sizes are NOT allowed to collide
{"test/.pad/1234", 0x8000, file_storage::flag_pad_file, "test/.pad/1234"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234.1"},
{"test/filler-2", 0x4000, {}, "test/filler-2"},
},
{
// pad files are NOT allowed to collide with normal files
{"test/.pad/1234", 0x4000, {}, "test/.pad/1234"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234.1"},
{"test/filler-2", 0x4000, {}, "test/filler-2"},
},
{
// normal files are NOT allowed to collide with pad files
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/.pad/1234", 0x4000, {}, "test/.pad/1234.1"},
{"test/filler-2", 0x4000, {}, "test/filler-2"},
},
{
// pad files are NOT allowed to collide with directories
{"test/.pad/1234", 0x4000, file_storage::flag_pad_file, "test/.pad/1234.1"},
{"test/filler-1", 0x4000, {}, "test/filler-1"},
{"test/.pad/1234/filler-2", 0x4000, {}, "test/.pad/1234/filler-2"},
},
};
void test_resolve_duplicates(aux::vector<file_t, file_index_t> const& test)
{
file_storage fs;
for (auto const& f : test) fs.add_file(f.filename, f.size, f.flags);
// This test creates torrents with duplicate (identical) filenames, which
// isn't supported by v2 torrents, so we can only test this with v1 torrents
lt::create_torrent t(fs, 0x4000, create_torrent::v1_only);
for (auto const i : fs.piece_range())
t.set_hash(i, sha1_hash::max());
std::vector<char> tmp;
std::back_insert_iterator<std::vector<char>> out(tmp);
entry tor = t.generate();
bencode(out, tor);
torrent_info ti(tmp, from_span);
for (auto const i : fs.file_range())
{
std::string p = ti.files().file_path(i);
convert_path_to_posix(p);
std::printf("%s == %s\n", p.c_str(), test[i].expected_filename.to_string().c_str());
TEST_EQUAL(p, test[i].expected_filename);
}
}
} // anonymous namespace
TORRENT_TEST(resolve_duplicates)
{
for (auto const& t : test_cases)
test_resolve_duplicates(t);
}
TORRENT_TEST(empty_file)
{
error_code ec;
auto ti = std::make_shared<torrent_info>("", ec, from_span);
TEST_CHECK(ec);
}
TORRENT_TEST(empty_file2)
{
try
{
auto ti = std::make_shared<torrent_info>("", from_span);
TEST_ERROR("expected exception thrown");
}
catch (system_error const& e)
{
std::printf("Expected error: %s\n", e.code().message().c_str());
}
}
TORRENT_TEST(load_torrent_empty_file)
{
try
{
auto atp = load_torrent_buffer({});
TEST_ERROR("expected exception thrown");
}
catch (system_error const& e)
{
std::printf("Expected error: %s\n", e.code().message().c_str());
}
}
TORRENT_TEST(copy)
{
using namespace lt;
std::shared_ptr<torrent_info> a = std::make_shared<torrent_info>(
combine_path(parent_path(current_working_directory())
, combine_path("test_torrents", "sample.torrent")));
aux::vector<char const*, file_index_t> expected_files =
{
"sample/text_file2.txt",
"sample/.____padding_file/0",
"sample/text_file.txt",
};
aux::vector<sha1_hash, file_index_t> file_hashes =
{
sha1_hash(nullptr),
sha1_hash(nullptr),
sha1_hash("abababababababababab")
};
file_storage const& fs = a->files();
for (auto const i : fs.file_range())
{
std::string p = fs.file_path(i);
convert_path_to_posix(p);
TEST_EQUAL(p, expected_files[i]);
std::printf("%s\n", p.c_str());
TEST_EQUAL(a->files().hash(i), file_hashes[i]);
}
// copy the torrent_info object
std::shared_ptr<torrent_info> b = std::make_shared<torrent_info>(*a);
a.reset();
TEST_EQUAL(b->num_files(), 3);
file_storage const& fs2 = b->files();
for (auto const i : fs2.file_range())
{
std::string p = fs2.file_path(i);
convert_path_to_posix(p);
TEST_EQUAL(p, expected_files[i]);
std::printf("%s\n", p.c_str());
TEST_EQUAL(fs2.hash(i), file_hashes[i]);
}
}
struct A
{
int val;
};
TORRENT_TEST(copy_ptr)
{
copy_ptr<A> a(new A{4});
copy_ptr<A> b(a);
TEST_EQUAL(a->val, b->val);
TEST_CHECK(&*a != &*b);
a->val = 5;
TEST_EQUAL(b->val, 4);
}
TORRENT_TEST(torrent_info_with_hashes_roundtrip)
{
std::string const root_dir = parent_path(current_working_directory());
std::string const filename = combine_path(combine_path(root_dir, "test_torrents"), "v2_only.torrent");
error_code ec;
std::vector<char> data;
TEST_CHECK(load_file(filename, data, ec) == 0);
auto ti = std::make_shared<torrent_info>(data, ec, from_span);
TEST_CHECK(!ec);
if (ec) std::printf(" loading(\"%s\") -> failed %s\n", filename.c_str()
, ec.message().c_str());
TEST_CHECK(ti->v2());
TEST_CHECK(!ti->v1());
TEST_EQUAL(ti->v2_piece_hashes_verified(), true);
add_torrent_params atp;
atp.ti = ti;
atp.save_path = ".";
session ses;
torrent_handle h = ses.add_torrent(atp);
TEST_CHECK(ti->v2());
TEST_CHECK(!ti->v1());
{
auto ti2 = h.torrent_file();
TEST_CHECK(ti2->v2());
TEST_CHECK(!ti2->v1());
TEST_EQUAL(ti2->v2_piece_hashes_verified(), false);
}
ti = h.torrent_file_with_hashes();
TEST_CHECK(ti->v2());
TEST_CHECK(!ti->v1());
TEST_EQUAL(ti->v2_piece_hashes_verified(), true);
std::vector<char> out_buffer = serialize(*ti);
TEST_EQUAL(out_buffer, data);
}
TORRENT_TEST(write_torrent_file_session_roundtrip)
{
std::string const root_dir = combine_path(parent_path(current_working_directory()), "test_torrents");
auto const files = {
"base.torrent",
"empty_path.torrent",
"parent_path.torrent",
"hidden_parent_path.torrent",
"single_multi_file.torrent",
"slash_path.torrent",
"slash_path2.torrent",
"slash_path3.torrent",
"backslash_path.torrent",
"httpseed.torrent",
"long_name.torrent",
"duplicate_files.torrent",
"pad_file.torrent",
"creation_date.torrent",
"no_creation_date.torrent",
"url_seed.torrent",
"url_seed_multi_single_file.torrent",
"empty_path_multi.torrent",
"invalid_name2.torrent",
"invalid_name3.torrent",
"symlink1.torrent",
"symlink2.torrent",
"unordered.torrent",
"symlink_zero_size.torrent",
"pad_file_no_path.torrent",
"large.torrent",
"absolute_filename.torrent",
"invalid_filename.torrent",
"invalid_filename2.torrent",
"overlapping_symlinks.torrent",
"v2.torrent",
"v2_multipiece_file.torrent",
"v2_only.torrent",
"v2_invalid_filename.torrent",
"v2_multiple_files.torrent",
"v2_symlinks.torrent",
"v2_hybrid.torrent",
"empty-files-1.torrent",
"empty-files-2.torrent",
"empty-files-3.torrent",
"empty-files-4.torrent",
"empty-files-5.torrent",
"similar.torrent",
"collection.torrent",
"collection2.torrent",
"similar.torrent",
"similar2.torrent",
"dht_nodes.torrent",
};
for (auto const& name : files)
{
std::string const filename = combine_path(root_dir, name);
std::printf("loading(\"%s\")\n", name);
error_code ec;
std::vector<char> data;
TEST_CHECK(load_file(filename, data, ec) == 0);
auto ti = std::make_shared<torrent_info>(data, ec, from_span);
TEST_CHECK(!ec);
if (ec) std::printf(" -> failed %s\n", ec.message().c_str());
add_torrent_params atp;
atp.ti = ti;
atp.save_path = ".";
session ses;
torrent_handle h = ses.add_torrent(atp);
h.save_resume_data(torrent_handle::save_info_dict);
alert const* a = wait_for_alert(ses, save_resume_data_alert::alert_type);
TORRENT_ASSERT(a);
{
auto const& p = static_cast<save_resume_data_alert const*>(a)->params;
entry e = write_torrent_file(p, write_flags::include_dht_nodes);
std::vector<char> out_buffer;
bencode(std::back_inserter(out_buffer), e);
TEST_CHECK(out_buffer == write_torrent_file_buf(p, write_flags::include_dht_nodes));
if (out_buffer != data)
{
std::cout << "GOT:\n";
for (char b : out_buffer)
std::cout << (std::isprint(std::uint8_t(b)) ? b : '.');
std::cout << '\n';
std::cout << "EXPECTED:\n";
for (char b : data)
std::cout << (std::isprint(std::uint8_t(b)) ? b : '.');
std::cout << '\n';
}
TEST_CHECK(out_buffer == data);
}
{
add_torrent_params p = load_torrent_file(filename);
entry e = write_torrent_file(p, write_flags::include_dht_nodes);
std::vector<char> out_buffer;
bencode(std::back_inserter(out_buffer), e);
TEST_CHECK(out_buffer == write_torrent_file_buf(p, write_flags::include_dht_nodes));
if (out_buffer != data)
{
std::cout << "GOT:\n";
for (char b : out_buffer)
std::cout << (std::isprint(std::uint8_t(b)) ? b : '.');
std::cout << '\n';
std::cout << "EXPECTED:\n";
for (char b : data)
std::cout << (std::isprint(std::uint8_t(b)) ? b : '.');
std::cout << '\n';
}
TEST_CHECK(out_buffer == data);
}
}
}