From e4b38653876821e2c7687decc1f27b1f2a4563e2 Mon Sep 17 00:00:00 2001 From: yhirose Date: Sat, 17 Oct 2026 09:00:41 -0400 Subject: [PATCH] Revert "Encode the characters RFC 3986 doesn't allow in a client path" This reverts commit 28f3922. Keep the path encoding settled in #2082 (#2170) and #2608 until a concrete report needs more: user input belongs in encode_path_component(), and an embedded NUL is invalid input. --- README.md | 2 +- httplib.h | 54 ++++++++++++++++++++++------------------------------ test/test.cc | 52 -------------------------------------------------- 3 files changed, 24 insertions(+), 84 deletions(-) diff --git a/README.md b/README.md index 4e21bf18..b66a672c 100644 --- a/README.md +++ b/README.md @@ -1471,7 +1471,7 @@ auto res = cli.Get("/already%20encoded/path"); // Use pre-encoded paths ``` - `set_path_encode(bool on)` - Controls automatic encoding of special characters in URL paths - - `true` (default): Percent-encodes characters not allowed in a URL path (spaces, control characters, `<`, `|`, non-ASCII bytes, etc.) and `+`, `,`, `;`, `'`. `%` is kept, so existing escapes aren't encoded twice + - `true` (default): Automatically encodes spaces, plus signs, newlines, and other special characters - `false`: Sends paths as-is without encoding (useful for pre-encoded URLs) ### Performance Note for Local Connections diff --git a/httplib.h b/httplib.h index 1159c630..3548e720 100644 --- a/httplib.h +++ b/httplib.h @@ -5999,24 +5999,32 @@ inline size_t FileStat::size() const { return ret_ >= 0 ? static_cast(st_.st_size) : 0; } -// Percent-encode every byte outside RFC 3986 pchar, keeping '/' and '%' so that -// escapes pass through. '+', ',', ';' and '\'' are pchar but encoded too, as -// some servers give '+' and ';' a special meaning. inline std::string encode_path(const std::string &s) { - static const char hex[] = "0123456789ABCDEF"; std::string result; result.reserve(s.size()); - for (auto c : s) { - // The '\0' check keeps strchr from matching the terminator. - if (is_ascii_alnum(c) || - (c != '\0' && std::strchr("-._~!$&()*=:@/%", c) != nullptr)) { - result += c; - } else { - auto uc = static_cast(c); - result += '%'; - result += hex[uc >> 4]; - result += hex[uc & 0x0f]; + for (size_t i = 0; s[i]; i++) { + switch (s[i]) { + case ' ': result += "%20"; break; + case '+': result += "%2B"; break; + case '\'': result += "%27"; break; + case ',': result += "%2C"; break; + // case ':': result += "%3A"; break; // ok? probably... + case ';': result += "%3B"; break; + default: + auto c = static_cast(s[i]); + // Control characters (incl. CR/LF) and non-ASCII bytes are not allowed + // in a request-target as-is. + if (c < 0x20 || c == 0x7f || c >= 0x80) { + result += '%'; + char hex[4]; + auto len = snprintf(hex, sizeof(hex) - 1, "%02X", c); + assert(len == 2); + result.append(hex, static_cast(len)); + } else { + result += s[i]; + } + break; } } @@ -9315,23 +9323,7 @@ inline std::string encode_request_target(const std::string &uri, query_part = target.substr(query_pos + 1); } - std::string result; - if (path_encode) { - // Encode only the path: an absolute-form target (sent to a proxy) keeps - // its "http://host", and a CONNECT "host:port" has no path. - size_t path_pos = 0; - if (!path_part.empty() && path_part[0] != '/') { - auto scheme_end = path_part.find("://"); - path_pos = scheme_end == std::string::npos - ? path_part.size() - : (std::min)(path_part.find('/', scheme_end + 3), - path_part.size()); - } - result = path_part.substr(0, path_pos); - result += encode_path(path_part.substr(path_pos)); - } else { - result = std::move(path_part); - } + auto result = path_encode ? encode_path(path_part) : std::move(path_part); if (!query_part.empty()) { // When path encoding is disabled the caller has supplied an already-encoded diff --git a/test/test.cc b/test/test.cc index 9227579c..cdbd4167 100644 --- a/test/test.cc +++ b/test/test.cc @@ -4457,10 +4457,6 @@ TEST(PathUrlEncodeTest, StreamingMatchesBufferedTarget) { {"/a b?x=1", "/a%20b?x=1"}, {"/\xE6\x97\xA5\xE6\x9C\xAC", "/%E6%97%A5%E6%9C%AC"}, {"/a,b;c+d", "/a%2Cb%3Bc%2Bd"}, - // Characters RFC 3986 doesn't allow in a path. - {"/\"<>\\^`{|}[]", "/%22%3C%3E%5C%5E%60%7B%7C%7D%5B%5D"}, - // Escapes and the rest of pchar are kept. - {"/a%20b/!$&()*=:@~", "/a%20b/!$&()*=:@~"}, }; for (const auto &c : cases) { @@ -4576,57 +4572,9 @@ TEST(PathUrlEncodeTest, ControlCharsInPathAreEncoded) { auto res = cli.Get("/a\tb\x1b\x7f"); ASSERT_TRUE(res) << "Error: " << to_string(res.error()); EXPECT_EQ("/a%09b%1B%7F", target); - - // An embedded NUL doesn't cut the path short. - res = cli.Get(std::string("/a\0b", 4)); - ASSERT_TRUE(res) << "Error: " << to_string(res.error()); - EXPECT_EQ("/a%00b", target); } } -TEST(PathUrlEncodeTest, HostInProxyTargetIsNotEncoded) { - // Only the path of a target sent to a proxy is encoded, so the brackets of - // an IPv6 host stay. - Server proxy; - - std::string target; - proxy.set_pre_routing_handler([&](const Request &req, Response &res) { - target = req.target; - res.status = StatusCode::Forbidden_403; - return Server::HandlerResponse::Handled; - }); - - auto port = proxy.bind_to_any_port(HOST); - auto thread = std::thread([&]() { proxy.listen_after_bind(); }); - auto se = detail::scope_exit([&] { - proxy.stop(); - thread.join(); - ASSERT_FALSE(proxy.is_running()); - }); - - proxy.wait_until_ready(); - - { - Client cli("::1", 8080); - cli.set_proxy(HOST, port); - - auto res = cli.Get("/a b"); - ASSERT_TRUE(res) << "Error: " << to_string(res.error()); - EXPECT_EQ("http://[::1]:8080/a%20b", target); - } - -#ifdef CPPHTTPLIB_SSL_ENABLED - { - SSLClient cli("::1", 8443); - cli.set_proxy(HOST, port); - - auto res = cli.Get("/"); - EXPECT_EQ(Error::ProxyConnection, res.error()); - EXPECT_EQ("[::1]:8443", target); - } -#endif -} - TEST(PathUrlEncodeTest, StreamingCRLFRejectedWhenPathEncodeDisabled) { // Nothing may reach the wire: a raw CR/LF target would split the request // line and inject headers.