Add CSRF protection

Split HTML helpers into their own file
Move hex-related functions into their own file
2023-12-08 22:27:40 +11:00 · 2023-12-08 22:27:40 +11:00 · 2023-12-08 22:27:40 +11:00 · 2023-12-08 18:12:51 +11:00
21 changed files with 795 additions and 544 deletions
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -5,10 +5,14 @@ project(coyote C CXX)

 find_package(nlohmann_json REQUIRED)
 find_package(CURL REQUIRED)
+find_package(OpenSSL REQUIRED)
+
 set(HTTPLIB_REQUIRE_OPENSSL ON)
 add_subdirectory(thirdparty/httplib)
+
 set(LEXBOR_BUILD_SHARED OFF)
 add_subdirectory(thirdparty/lexbor)
+
 find_package(PkgConfig REQUIRED)
 pkg_check_modules(HIREDIS REQUIRED hiredis)

@ -29,7 +33,7 @@ list(APPEND FLAGS -Werror -Wall -Wextra -Wshadow -Wpedantic -Wno-gnu-anonymous-s
 add_link_options(${FLAGS})


-add_executable(${PROJECT_NAME} main.cpp numberhelper.cpp config.cpp settings.cpp models.cpp client.cpp servehelper.cpp timeutils.cpp hiredis_wrapper.cpp
+add_executable(${PROJECT_NAME} main.cpp numberhelper.cpp hex.cpp config.cpp settings.cpp models.cpp client.cpp servehelper.cpp htmlhelper.cpp timeutils.cpp openssl_wrapper.cpp hiredis_wrapper.cpp
    routes/home.cpp routes/css.cpp routes/user.cpp routes/status.cpp routes/tags.cpp routes/about.cpp routes/user_settings.cpp
    blankie/serializer.cpp blankie/escape.cpp)
 set_target_properties(${PROJECT_NAME}
@ -39,6 +43,6 @@ set_target_properties(${PROJECT_NAME}
        CXX_EXTENSIONS NO
 )
 target_include_directories(${PROJECT_NAME} PRIVATE thirdparty ${HIREDIS_INCLUDE_DIRS})
-target_link_libraries(${PROJECT_NAME} PRIVATE nlohmann_json::nlohmann_json CURL::libcurl httplib::httplib lexbor_static ${HIREDIS_LINK_LIBRARIES})
+target_link_libraries(${PROJECT_NAME} PRIVATE nlohmann_json::nlohmann_json CURL::libcurl OpenSSL::Crypto httplib::httplib lexbor_static ${HIREDIS_LINK_LIBRARIES})
 target_compile_definitions(${PROJECT_NAME} PRIVATE ${DEFINITIONS})
 target_compile_options(${PROJECT_NAME} PRIVATE ${FLAGS})
--- a/RUNNING.md
+++ b/RUNNING.md
@ -15,6 +15,7 @@ Copy `example_config.json` to a file with any name you like
 liking. Here's a list of what they are:
 - `bind_host` (string): What address to bind to
 - `bind_port` (zero or positive integer): What port to bind to
+- `hmac_key` (hex string): A secret key to be used; generate with `head -c32 /dev/urandom | basenc --base16`
 - `canonical_origin` (string or null): A fallback canonical origin if set, useful if you're, say, running Coyote behind Ngrok
 - `redis` (object)
  - `enabled` (boolean)
--- a/client.cpp
+++ b/client.cpp
@ -2,6 +2,7 @@
 #include <stdexcept>
 #include <nlohmann/json.hpp>

+#include "hex.h"
 #include "client.h"
 #include "models.h"
 #include "curlu_wrapper.h"
@ -11,8 +12,6 @@ MastodonClient mastodon_client;

 static void lowercase(std::string& str);
 static void handle_post_server(Post& post, const std::string& host);
-static std::string url_encode(const std::string& in);
-static inline void hexencode(char c, char out[2]);

 static void share_lock(CURL* curl, curl_lock_data data, curl_lock_access access, void* clientp);
 static void share_unlock(CURL* curl, curl_lock_data data, void* clientp);
@ -90,7 +89,7 @@ std::optional<Account> MastodonClient::get_account_by_username(std::string host,
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
    url.set(CURLUPART_PATH, "/api/v1/accounts/lookup");
-    url.set(CURLUPART_QUERY, "acct="s + url_encode(username));
+    url.set(CURLUPART_QUERY, "acct="s + percent_encode(username));
    try {
        Account account = this->_send_request("coyote:"s + host + ":@" + username, url);
        account.same_server = host == account.server;
@ -111,7 +110,7 @@ std::vector<Post> MastodonClient::get_pinned_posts(std::string host, const std::
    CurlUrl url;
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
-    url.set(CURLUPART_PATH, "/api/v1/accounts/"s + url_encode(account_id) + "/statuses");
+    url.set(CURLUPART_PATH, "/api/v1/accounts/"s + percent_encode(account_id) + "/statuses");
    url.set(CURLUPART_QUERY, "pinned=true");
    std::vector<Post> posts = this->_send_request("coyote:"s + host + ':' + account_id + ":pinned", url);

@ -129,7 +128,7 @@ std::vector<Post> MastodonClient::get_posts(const std::string& host, const std::
    CurlUrl url;
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
-    url.set(CURLUPART_PATH, "/api/v1/accounts/"s + url_encode(account_id) + "/statuses");
+    url.set(CURLUPART_PATH, "/api/v1/accounts/"s + percent_encode(account_id) + "/statuses");
    url.set(CURLUPART_QUERY, sorting_parameters[sorting_method]);
    if (max_id) {
        url.set(CURLUPART_QUERY, "max_id="s + std::move(*max_id), CURLU_URLENCODE | CURLU_APPENDQUERY);
@ -150,7 +149,7 @@ std::optional<Post> MastodonClient::get_post(const std::string& host, std::strin
    CurlUrl url;
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
-    url.set(CURLUPART_PATH, "/api/v1/statuses/"s + url_encode(std::move(id)));
+    url.set(CURLUPART_PATH, "/api/v1/statuses/"s + percent_encode(std::move(id)));
    try {
        Post post = this->_send_request(std::nullopt, url);
        handle_post_server(post, host);
@ -170,7 +169,7 @@ PostContext MastodonClient::get_post_context(const std::string& host, std::strin
    CurlUrl url;
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
-    url.set(CURLUPART_PATH, "/api/v1/statuses/"s + url_encode(std::move(id)) + "/context");
+    url.set(CURLUPART_PATH, "/api/v1/statuses/"s + percent_encode(std::move(id)) + "/context");
    PostContext context = this->_send_request(std::nullopt, url);

    for (Post& post : context.ancestors) {
@ -189,7 +188,7 @@ std::vector<Post> MastodonClient::get_tag_timeline(const std::string& host, cons
    CurlUrl url;
    url.set(CURLUPART_SCHEME, "https");
    url.set(CURLUPART_HOST, host);
-    url.set(CURLUPART_PATH, "/api/v1/timelines/tag/"s + url_encode(tag));
+    url.set(CURLUPART_PATH, "/api/v1/timelines/tag/"s + percent_encode(tag));
    if (max_id) {
        url.set(CURLUPART_QUERY, "max_id="s + std::move(*max_id), CURLU_URLENCODE | CURLU_APPENDQUERY);
    }
@ -313,42 +312,6 @@ static void handle_post_server(Post& post, const std::string& host) {
    }
 }

-static std::string url_encode(const std::string& in) {
-    std::string out;
-    char encoded[2];
-    size_t pos = 0;
-    size_t last_pos = 0;
-
-    out.reserve(in.size());
-    while ((pos = in.find_first_not_of("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789", pos)) != std::string::npos) {
-        out.append(in, last_pos, pos - last_pos);
-        hexencode(in[pos], encoded);
-        out += '%';
-        out.append(encoded, 2);
-        pos++;
-        last_pos = pos;
-    }
-
-    if (in.size() > last_pos) {
-        out.append(in, last_pos);
-    }
-
-    return out;
-}
-
-static inline void hexencode(char c, char out[2]) {
-    char nibble1 = (c >> 4) & 0xF;
-    char nibble2 = c & 0xF;
-
-    auto hexencode = [](char nibble) {
-        return static_cast<char>(nibble < 10
-            ? '0' + nibble
-            : 'A' + nibble - 10);
-    };
-    out[0] = hexencode(nibble1);
-    out[1] = hexencode(nibble2);
-}
-
 static void share_lock(CURL* curl, curl_lock_data data, curl_lock_access access, void* clientp) {
    (void)curl;
    (void)access;
--- a/config.cpp
+++ b/config.cpp
@ -1,6 +1,7 @@
 #include <stdexcept>
 #include <nlohmann/json.hpp>

+#include "hex.h"
 #include "file.h"
 #include "config.h"

@ -47,6 +48,7 @@ void from_json(const nlohmann::json& j, Config& conf) {
        throw std::invalid_argument("Invalid port to bind to: "s + std::to_string(conf.bind_port));
    }

+    conf.hmac_key = hex_decode(j.at("hmac_key").get_ref<const std::string&>());
    if (j.at("canonical_origin").is_string()) {
        conf.canonical_origin = j["canonical_origin"].get<std::string>();
    }
--- a/config.h
+++ b/config.h
@ -1,6 +1,7 @@
 #pragma once

 #include <string>
+#include <vector>
 #include <variant>
 #include <optional>
 #include <nlohmann/json_fwd.hpp>
@ -22,6 +23,7 @@ struct RedisConfig {
 struct Config {
    std::string bind_host = "127.0.0.1";
    int bind_port = 8080;
+    std::vector<char> hmac_key;
    std::optional<std::string> canonical_origin;
    std::optional<RedisConfig> redis_config;
 };
--- a/example_config.json
+++ b/example_config.json
@ -1,6 +1,7 @@
 {
    "bind_host": "127.0.0.1",
    "bind_port": 8080,
+    "hmac_key": "AA",
    "canonical_origin": null,
    "redis": {
        "enabled": true,
--- a/hex.cpp
+++ b/hex.cpp
@ -0,0 +1,82 @@
+#include <stdexcept>
+
+#include "hex.h"
+
+static void hex_encode(char c, char out[2]);
+static inline char hex_decode(char nibble1, char nibble2);
+
+
+std::string hex_encode(const char* in, size_t in_size) {
+    std::string out;
+
+    out.reserve(in_size * 2);
+    for (size_t i = 0; i < in_size; i++) {
+        char encoded[2];
+        hex_encode(in[i], encoded);
+        out.append(encoded, 2);
+    }
+
+    return out;
+}
+
+std::string percent_encode(std::string_view in) {
+    std::string out;
+    char encoded[2];
+    size_t pos = 0;
+    size_t last_pos = 0;
+
+    out.reserve(in.size());
+    while ((pos = in.find_first_not_of("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789", pos)) != std::string::npos) {
+        out.append(in, last_pos, pos - last_pos);
+        hex_encode(in[pos], encoded);
+        out += '%';
+        out.append(encoded, 2);
+        pos++;
+        last_pos = pos;
+    }
+
+    if (in.size() > last_pos) {
+        out.append(in, last_pos);
+    }
+
+    return out;
+}
+
+std::vector<char> hex_decode(std::string_view in) {
+    if (in.size() % 2 != 0) {
+        throw std::invalid_argument("hex_decode(): hex string with an odd size passed");
+    }
+    std::vector<char> out;
+
+    out.reserve(in.size() / 2);
+    for (size_t i = 0; i < in.size(); i += 2) {
+        out.push_back(hex_decode(in[i], in[i + 1]));
+    }
+
+    return out;
+}
+
+
+static void hex_encode(char c, char out[2]) {
+    char nibble1 = (c >> 4) & 0xF;
+    char nibble2 = c & 0xF;
+
+    auto hex_encode = [](char nibble) {
+        return static_cast<char>(nibble < 10
+            ? '0' + nibble
+            : 'A' + nibble - 10);
+    };
+    out[0] = hex_encode(nibble1);
+    out[1] = hex_encode(nibble2);
+}
+
+static inline char hex_decode(char nibble1, char nibble2) {
+    auto hex_decode = [](char nibble) {
+        if (nibble >= '0' && nibble <= '9') return nibble - '0';
+        if (nibble >= 'a' && nibble <= 'f') return nibble - 'a' + 10;
+        if (nibble >= 'A' && nibble <= 'F') return nibble - 'A' + 10;
+        throw std::invalid_argument("hex_decode(): invalid nibble");
+    };
+
+    return static_cast<char>((hex_decode(nibble1) << 4) | hex_decode(nibble2));
+}
--- a/hex.h
+++ b/hex.h
@ -0,0 +1,12 @@
+#pragma once
+
+#include <string>
+#include <vector>
+
+std::string hex_encode(const char* in, size_t in_size);
+inline std::string hex_encode(const std::vector<char>& in) {
+    return hex_encode(in.data(), in.size());
+}
+std::string percent_encode(std::string_view in);
+
+std::vector<char> hex_decode(std::string_view in);
--- a/htmlhelper.cpp
+++ b/htmlhelper.cpp
@ -0,0 +1,456 @@
+#include "models.h"
+#include "settings.h"
+#include "timeutils.h"
+#include "curlu_wrapper.h"
+#include "font_awesome.h"
+#include "blankie/escape.h"
+
+#include "htmlhelper.h"
+
+static inline void preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, lxb_dom_element_t* element);
+static inline void preprocess_link(const httplib::Request& req, const std::string& domain_name, lxb_dom_element_t* element);
+static inline bool should_fix_link(lxb_dom_element_t* element, const std::string& element_cls);
+static inline void get_text_content(lxb_dom_node_t* node, std::string& out);
+static inline lxb_dom_node_t* emojify(lxb_dom_node_t* child, const std::vector<Emoji>& emojis);
+static inline std::vector<lxb_dom_node*> emojify(lxb_dom_document_t* document, std::string str, const std::vector<Emoji>& emojis);
+
+struct PostStatus {
+    const char* icon_html;
+    Node info_node;
+};
+static Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool main_post, const std::optional<PostStatus>& post_status, const Post* reblogged = nullptr);
+static inline Element serialize_media(const Media& media);
+static inline Element serialize_poll(const httplib::Request& req, const Poll& poll);
+
+
+Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool pinned, bool main_post) {
+    using namespace std::string_literals;
+
+    if (post.reblog) {
+        PostStatus post_status = {
+            fa_retweet,
+            preprocess_html(req, post.account.emojis, post.account.display_name + " boosted"),
+        };
+        return serialize_post(req, server, *post.reblog, main_post, post_status, &post);
+    } else if (pinned) {
+        PostStatus post_status = {
+            fa_thumbtack,
+            blankie::html::HTMLString("Pinned post"),
+        };
+        return serialize_post(req, server, post, main_post, post_status);
+    } else if (post.in_reply_to_id && post.in_reply_to_account_id && post.account.id == *post.in_reply_to_account_id) {
+        PostStatus post_status = {
+            fa_reply,
+            preprocess_html(req, post.account.emojis, "Replied to "s + post.account.display_name),
+        };
+        return serialize_post(req, server, post, main_post, post_status);
+    } else {
+        return serialize_post(req, server, post, main_post, std::nullopt);
+    }
+}
+
+std::string get_text_content(lxb_dom_node_t* child) {
+    std::string out;
+    get_text_content(child, out);
+
+    if (!out.empty()) {
+        size_t remove_from = out.size();
+        while (remove_from && out[remove_from - 1] == '\n') {
+            remove_from--;
+        }
+        // Don't engulf everything, otherwise it crashes
+        // https://ruby.social/@CoralineAda/109951421922797743
+        if (out.size() > remove_from && remove_from != 0) {
+            out.erase(remove_from);
+        }
+    }
+    if (!out.empty()) {
+        size_t remove_to = 0;
+        while (out.size() > remove_to && out[remove_to] == '\n') {
+            remove_to++;
+        }
+        // Don't engulf everything, otherwise it crashes
+        // https://ruby.social/@CoralineAda/109951421922797743
+        if (out.size() > remove_to) {
+            out.erase(0, remove_to);
+        }
+    }
+
+    return out;
+}
+
+std::string get_text_content(blankie::html::HTMLString str) {
+    LXB::HTML::Document document(str.str);
+    return get_text_content(document.body());
+}
+
+blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, const blankie::html::HTMLString& str) {
+    LXB::HTML::Document document(str.str);
+    preprocess_html(req, domain_name, emojis, document.body_element());
+    return blankie::html::HTMLString(document.serialize());
+}
+
+blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::vector<Emoji>& emojis, const std::string& str) {
+    return preprocess_html(req, "", emojis, blankie::html::HTMLString(blankie::html::escape(str)));
+}
+
+
+static inline void preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, lxb_dom_element_t* element) {
+    const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(element, nullptr));
+
+    if (strncmp(tag_name, "A", 2) == 0) {
+        // Proprocess links
+        preprocess_link(req, domain_name, element);
+    }
+
+    // Walk through the element's children
+    lxb_dom_node_t* child = lxb_dom_node_first_child(lxb_dom_interface_node(element));
+    while (child) {
+        if (child->type == LXB_DOM_NODE_TYPE_ELEMENT) {
+            preprocess_html(req, domain_name, emojis, lxb_dom_interface_element(child));
+        } else if (child->type == LXB_DOM_NODE_TYPE_TEXT) {
+            child = emojify(child, emojis);
+        }
+
+        child = lxb_dom_node_next(child);
+    }
+}
+
+static std::regex mention_class_re("\\bmention\\b");
+static inline void preprocess_link(const httplib::Request& req, const std::string& domain_name, lxb_dom_element_t* element) {
+    using namespace std::string_literals;
+
+    // Remove target=...
+    lxb_status_t status = lxb_dom_element_remove_attribute(element, reinterpret_cast<const lxb_char_t*>("target"), 6);
+    if (status != LXB_STATUS_OK) {
+        throw LXB::Exception(status);
+    }
+
+    size_t href_c_len;
+    const lxb_char_t* href_c = lxb_dom_element_get_attribute(element, reinterpret_cast<const lxb_char_t*>("href"), 4, &href_c_len);
+    if (!href_c) {
+        return;
+    }
+    std::string href(reinterpret_cast<const char*>(href_c), href_c_len);
+
+    size_t cls_c_len;
+    const lxb_char_t* cls_c = lxb_dom_element_class(element, &cls_c_len);
+    std::string cls = cls_c ? std::string(reinterpret_cast<const char*>(cls_c), cls_c_len) : "";
+
+    try {
+        CurlUrl href_url;
+        href_url.set(CURLUPART_URL, get_origin(req));
+        href_url.set(CURLUPART_PATH, std::string(href_url.get(CURLUPART_PATH).get()) + req.path);
+        href_url.set(CURLUPART_URL, href);
+
+        CurlUrl instance_url_base;
+        instance_url_base.set(CURLUPART_SCHEME, "https");
+        instance_url_base.set(CURLUPART_HOST, domain_name);
+
+        // .mention is used in note and posts
+        // Instance base is used for link fields
+        if (std::regex_search(cls, mention_class_re) || starts_with(href_url, instance_url_base)) {
+            // Proxy this instance's URLs to Coyote
+            href = proxy_mastodon_url(req, std::move(href));
+
+            lxb_dom_element_set_attribute(element, reinterpret_cast<const lxb_char_t*>("href"), 4, reinterpret_cast<const lxb_char_t*>(href.data()), href.size());
+        }
+    } catch (const CurlUrlException& e) {
+        // example: <a href=""></a> on eldritch.cafe/about
+        if (e.code != CURLUE_MALFORMED_INPUT) {
+            throw;
+        }
+    }
+
+    if (should_fix_link(element, cls)) {
+        // Set the content of each <a> to its href
+        status = lxb_dom_node_text_content_set(lxb_dom_interface_node(element), reinterpret_cast<const lxb_char_t*>(href.data()), href.size());
+        if (status != LXB_STATUS_OK) {
+            throw LXB::Exception(status);
+        }
+    }
+}
+
+static std::regex unhandled_link_re("\\bunhandled-link\\b");
+static inline bool should_fix_link(lxb_dom_element_t* element, const std::string& element_cls) {
+    // https://vt.social/@LucydiaLuminous/111448085044245037
+    if (std::regex_search(element_cls, unhandled_link_re)) {
+        return true;
+    }
+
+    auto expected_element = [](lxb_dom_node_t* node, const char* expected_cls) {
+        if (!node || node->type != LXB_DOM_NODE_TYPE_ELEMENT) {
+            return false;
+        }
+        lxb_dom_element_t* span = lxb_dom_interface_element(node);
+
+        const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(span, nullptr));
+        if (strncmp(tag_name, "SPAN", 5) != 0) {
+            return false;
+        }
+
+        const lxb_char_t* cls = lxb_dom_element_get_attribute(span, reinterpret_cast<const lxb_char_t*>("class"), 5, nullptr);
+        return cls && strcmp(reinterpret_cast<const char*>(cls), expected_cls) == 0;
+    };
+
+    lxb_dom_node_t* child = lxb_dom_node_first_child(lxb_dom_interface_node(element));
+    if (!expected_element(child, "invisible")) {
+        return false;
+    }
+
+    child = lxb_dom_node_next(child);
+    if (!expected_element(child, "ellipsis") && !expected_element(child, "")) {
+        return false;
+    }
+
+    child = lxb_dom_node_next(child);
+    if (!expected_element(child, "invisible")) {
+        return false;
+    }
+
+    child = lxb_dom_node_next(child);
+    return child == nullptr;
+}
+
+static inline void get_text_content(lxb_dom_node_t* node, std::string& out) {
+    bool is_br = false, is_p = false;
+
+    if (node->type == LXB_DOM_NODE_TYPE_TEXT) {
+        size_t len;
+        const char* text = reinterpret_cast<const char*>(lxb_dom_node_text_content(node, &len));
+
+        out.append(text, len);
+    } else if (node->type == LXB_DOM_NODE_TYPE_ELEMENT) {
+        lxb_dom_element_t* element = lxb_dom_interface_element(node);
+        const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(element, nullptr));
+
+        is_p = strncmp(tag_name, "P", 2) == 0;
+        is_br = strncmp(tag_name, "BR", 3) == 0;
+    }
+
+    if (is_p || is_br) {
+        out.push_back('\n');
+    }
+
+    lxb_dom_node_t* child = lxb_dom_node_first_child(node);
+    while (child) {
+        get_text_content(child, out);
+
+        child = lxb_dom_node_next(child);
+    }
+
+    if (is_p) {
+        out.push_back('\n');
+    }
+}
+
+static inline lxb_dom_node_t* emojify(lxb_dom_node_t* child, const std::vector<Emoji>& emojis) {
+    std::vector<lxb_dom_node_t*> nodes = emojify(child->owner_document, get_text_content(child), emojis);
+
+    lxb_dom_node_insert_after(child, nodes[0]);
+    lxb_dom_node_destroy(child);
+    child = nodes[0];
+
+    for (size_t i = 1; i < nodes.size(); i++) {
+        lxb_dom_node_insert_after(child, nodes[i]);
+        child = nodes[i];
+    }
+
+    return child;
+}
+
+static std::regex shortcode_re(":([a-zA-Z0-9_]+):");
+static inline std::vector<lxb_dom_node_t*> emojify(lxb_dom_document_t* document, std::string str, const std::vector<Emoji>& emojis) {
+    std::string buf;
+    std::smatch sm;
+    std::vector<lxb_dom_node*> res;
+
+    while (std::regex_search(str, sm, shortcode_re)) {
+        buf += sm.prefix();
+
+        std::string group_0 = sm.str(0);
+        auto emoji = std::find_if(emojis.begin(), emojis.end(), [&](const Emoji& i) { return i.shortcode == sm.str(1); });
+        if (emoji != emojis.end()) {
+            res.push_back(lxb_dom_interface_node(lxb_dom_document_create_text_node(document, reinterpret_cast<const lxb_char_t*>(buf.data()), buf.size())));
+            buf.clear();
+
+            lxb_dom_element_t* img = lxb_dom_element_create(document, reinterpret_cast<const lxb_char_t*>("IMG"), 3, nullptr, 0, nullptr, 0, nullptr, 0, false);
+            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("class"), 5, reinterpret_cast<const lxb_char_t*>("custom_emoji"), 12);
+            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("alt"), 3, reinterpret_cast<const lxb_char_t*>(group_0.data()), group_0.size());
+            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("title"), 5, reinterpret_cast<const lxb_char_t*>(group_0.data()), group_0.size());
+            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("src"), 3, reinterpret_cast<const lxb_char_t*>(emoji->url.data()), emoji->url.size());
+            res.push_back(lxb_dom_interface_node(img));
+        } else {
+            buf += group_0;
+        }
+
+        str = sm.suffix();
+    }
+
+    if (!str.empty()) {
+        buf += std::move(str);
+    }
+    if (!buf.empty()) {
+        res.push_back(lxb_dom_interface_node(lxb_dom_document_create_text_node(document, reinterpret_cast<const lxb_char_t*>(buf.data()), buf.size())));
+    }
+
+    return res;
+}
+
+static Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool main_post, const std::optional<PostStatus>& post_status, const Post* reblogged) {
+    using namespace std::string_literals;
+
+    bool user_known = !post.account.id.empty();
+    bool user_ref_known = !post.account.username.empty() && !post.account.server.empty();
+    // `reblogged == nullptr` since a malicious server could take down the frontend
+    // by sending a post that is not a reblog with no account information
+    std::string post_url = user_known || reblogged == nullptr
+        ? get_origin(req) + '/' + server + "/@" + post.account.acct(false) + '/' + post.id + "#m"
+        : get_origin(req) + '/' + server + "/@" + reblogged->account.acct(false) + '/' + reblogged->id + "#m";
+
+    std::string time_title = post.edited_at < 0
+        ? full_time(post.created_at)
+        : "Created: "s + full_time(post.created_at) + "\nEdited: " + full_time(post.edited_at);
+    const char* time_badge = post.edited_at < 0 ? "" : " (edited)";
+
+    blankie::html::HTMLString preprocessed_html = preprocess_html(req, server, post.emojis, post.content);
+    // Workaround for https://vt.social/@a1ba@suya.place/110552480243348878#m
+    if (preprocessed_html.str.find("<p>") == std::string::npos) {
+        preprocessed_html.str.reserve(preprocessed_html.str.size() + 3 + 4);
+        preprocessed_html.str.insert(0, "<p>");
+        preprocessed_html.str.append("</p>");
+    }
+    Element contents("div", {{"class", "post-contents"}}, {std::move(preprocessed_html)});
+
+    Element post_attachments("div", {{"class", "post-attachments"}}, {});
+    post_attachments.nodes.reserve(post.media_attachments.size());
+    for (const Media& media : post.media_attachments) {
+        post_attachments.nodes.push_back(serialize_media(media));
+    }
+    contents.nodes.push_back(std::move(post_attachments));
+
+    if (post.poll) {
+        contents.nodes.push_back(serialize_poll(req, *post.poll));
+    }
+
+    if (post.sensitive) {
+        std::string spoiler_text = !post.spoiler_text.empty() ? post.spoiler_text : "See more";
+        contents = Element("details", {
+            Element("summary", {preprocess_html(req, post.emojis, std::move(spoiler_text))}),
+            std::move(contents),
+        });
+        if (UserSettings(req).auto_open_cw) {
+            contents.attributes.push_back({"open", ""});
+        }
+    }
+
+    Element div("div", {{"class", "post"}}, {
+        Element("div", {{"class", "post-header"}}, {
+            user_ref_known ? Element("a", {{"href", get_origin(req) + '/' + server + "/@" + post.account.acct(false)}}, {
+                !post.account.avatar_static.empty()
+                    ? Element("img", {{"class", "post-avatar"}, {"alt", "User profile picture"}, {"loading", "lazy"}, {"src", post.account.avatar_static}}, {})
+                    : Node(""),
+                Element("span", {
+                    Element("b", {preprocess_html(req, post.account.emojis, post.account.display_name)}),
+                    Element("br"), "@", post.account.acct(),
+                }),
+            }) : Element("b", {"Unknown user"}),
+            Element("a", {{"class", "post-time_header"}, {"href", std::move(post_url)}, {"title", time_title}}, {
+                Element("time", {{"datetime", to_rfc3339(post.created_at)}}, {relative_time(post.created_at, current_time()), time_badge}),
+            }),
+        }),
+
+        contents,
+    });
+    if (post_status) {
+        div.nodes.insert(div.nodes.begin(), Element("p", {
+            blankie::html::HTMLString(post_status->icon_html), " ", post_status->info_node,
+        }));
+    }
+    if (main_post) {
+        div.attributes = {{"class", "post main_post"}, {"id", "m"}};
+    }
+
+    return div;
+}
+
+static inline Element serialize_media(const Media& media) {
+    Element element = [&]() {
+        if (media.type == "image") {
+            return Element("a", {{"href", media.url}}, {
+                Element("img", {{"loading", "lazy"}, {"src", media.preview_url.value_or(media.url)}}, {}),
+            });
+        } else if (media.type == "video") {
+            Element video("video", {{"controls", ""}, {"src", media.url}}, {});
+            if (media.preview_url) {
+                video.attributes.push_back({"poster", *media.preview_url});
+            }
+            return video;
+        } else if (media.type == "audio") {
+            return Element("audio", {{"controls", ""}, {"src", media.url}}, {});
+        } else if (media.type == "gifv") {
+            // https://hachyderm.io/@Impossible_PhD/111444541628207638
+            Element video("video", {{"controls", ""}, {"loop", ""}, {"muted", ""}, {"autoplay", ""}, {"src", media.url}}, {});
+            if (media.preview_url) {
+                video.attributes.push_back({"poster", *media.preview_url});
+            }
+            return video;
+        } else if (media.type == "unknown") {
+            if (media.remote_url) {
+                // https://botsin.space/@lina@vt.social/111053598696451525
+                return Element("a", {{"class", "unknown_media"}, {"href", *media.remote_url}}, {"Media is not available from this instance, view externally"});
+            } else {
+                return Element("p", {{"class", "unknown_media"}}, {"Media is not available from this instance"});
+            }
+        } else {
+            return Element("p", {"Unsupported media type: ", media.type});
+        }
+    }();
+
+    if (media.description) {
+        element.attributes.push_back({"alt", *media.description});
+        element.attributes.push_back({"title", *media.description});
+    }
+
+    return element;
+}
+
+static inline Element serialize_poll(const httplib::Request& req, const Poll& poll) {
+    using namespace std::string_literals;
+
+    uint64_t voters_count = poll.voters_count >= 0 ? static_cast<uint64_t>(poll.voters_count) : poll.votes_count;
+    Element div("div");
+
+    auto pick_form = [](uint64_t count, const char* singular, const char* plural) {
+        return count == 1 ? singular : plural;
+    };
+
+    div.nodes.reserve(poll.options.size() + 1);
+    for (const PollOption& option : poll.options) {
+        std::string percentage = voters_count
+            ? std::to_string(option.votes_count * 100 / voters_count) + '%'
+            : "0%";
+
+        div.nodes.push_back(Element("div", {{"class", "poll-option"}, {"title", std::to_string(option.votes_count) + pick_form(option.votes_count, " vote", " votes")}}, {
+            Element("b", {{"class", "poll-percentage"}}, {percentage}), " ", preprocess_html(req, poll.emojis, option.title),
+            Element("object", {{"class", "poll-bar"}, {"width", percentage}}, {}),
+        }));
+    }
+
+    Element p("p", poll.voters_count >= 0
+        ? std::vector<Node>({std::to_string(voters_count), " ", pick_form(voters_count, "voter", "voters")})
+        : std::vector<Node>({std::to_string(poll.votes_count), " ", pick_form(poll.votes_count, "vote", "votes")})
+    );
+    if (poll.expired) {
+        p.nodes.push_back(" / ");
+        p.nodes.push_back(Element("time", {{"datetime", to_rfc3339(poll.expires_at)}, {"title", "Expired on "s + full_time(poll.expires_at)}}, {"Expired"}));
+    } else if (poll.expires_at >= 0) {
+        p.nodes.push_back(" / ");
+        p.nodes.push_back(Element("time", {{"datetime", to_rfc3339(poll.expires_at)}, {"title", full_time(poll.expires_at)}}, {
+            "Expires in ", relative_time(current_time(), poll.expires_at),
+        }));
+    }
+    div.nodes.push_back(std::move(p));
+
+    return div;
+}
--- a/htmlhelper.h
+++ b/htmlhelper.h
@ -0,0 +1,14 @@
+#pragma once
+
+#include "lxb_wrapper.h"
+
+#include "servehelper.h"
+struct Post; // forward declaration from models.h
+struct Emoji; // forward declaration from models.h
+
+Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool pinned = false, bool main_post = false);
+
+std::string get_text_content(lxb_dom_node_t* child);
+std::string get_text_content(blankie::html::HTMLString str);
+blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, const blankie::html::HTMLString& str);
+blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::vector<Emoji>& emojis, const std::string& str);
--- a/models.h
+++ b/models.h
@ -3,6 +3,7 @@
 #include <string>
 #include <vector>
 #include <memory>
+#include <optional>
 #include <nlohmann/json_fwd.hpp>

 #include "blankie/serializer.h"
--- a/openssl_wrapper.cpp
+++ b/openssl_wrapper.cpp
@ -0,0 +1,35 @@
+#include <memory>
+#include <stdexcept>
+
+#include <openssl/hmac.h>
+#include <openssl/rand.h>
+#include "openssl_wrapper.h"
+
+std::vector<char> secure_random_bytes(int num) {
+    if (num < 0) {
+        throw std::invalid_argument("secure_random_bytes(): num variable out of range (num < 0)");
+    }
+
+    std::vector<char> bytes(static_cast<size_t>(num), 0);
+    if (RAND_bytes(reinterpret_cast<unsigned char*>(bytes.data()), num) == 1) {
+        return bytes;
+    } else {
+        throw OpenSSLException(ERR_get_error());
+    }
+}
+
+std::array<char, 32> hmac_sha3_256(const std::vector<char>& key, const std::vector<char>& data) {
+    char hmac[32];
+    unsigned int md_len;
+
+    std::unique_ptr<EVP_MD, decltype(&EVP_MD_free)> md(EVP_MD_fetch(nullptr, "SHA3-256", nullptr), EVP_MD_free);
+    if (HMAC(md.get(), key.data(), static_cast<int>(key.size()), reinterpret_cast<const unsigned char*>(data.data()), data.size(), reinterpret_cast<unsigned char*>(hmac), &md_len)) {
+        if (md_len != 32) {
+            throw std::runtime_error("hmac_sha3_256(): HMAC() returned an unexpected size");
+        }
+
+        return std::to_array(hmac);
+    } else {
+        throw OpenSSLException(ERR_get_error());
+    }
+}
--- a/openssl_wrapper.h
+++ b/openssl_wrapper.h
@ -0,0 +1,24 @@
+#pragma once
+
+#include <array>
+#include <vector>
+#include <exception>
+
+#include <openssl/err.h>
+
+class OpenSSLException : public std::exception {
+public:
+    OpenSSLException(unsigned long e) {
+        ERR_error_string_n(e, this->_str, 1024);
+    }
+
+    const char* what() const noexcept {
+        return this->_str;
+    }
+
+private:
+    char _str[1024];
+};
+
+std::vector<char> secure_random_bytes(int num);
+std::array<char, 32> hmac_sha3_256(const std::vector<char>& key, const std::vector<char>& data);
--- a/routes/about.cpp
+++ b/routes/about.cpp
@ -1,5 +1,6 @@
 #include "routes.h"
 #include "../servehelper.h"
+#include "../htmlhelper.h"
 #include "../client.h"
 #include "../models.h"

--- a/routes/status.cpp
+++ b/routes/status.cpp
@ -1,6 +1,7 @@
 #include "routes.h"
 #include "../lxb_wrapper.h"
 #include "../servehelper.h"
+#include "../htmlhelper.h"
 #include "../client.h"
 #include "../models.h"

--- a/routes/tags.cpp
+++ b/routes/tags.cpp
@ -1,5 +1,6 @@
 #include "routes.h"
 #include "../servehelper.h"
+#include "../htmlhelper.h"
 #include "../client.h"
 #include "../models.h"

--- a/routes/user.cpp
+++ b/routes/user.cpp
@ -1,5 +1,6 @@
 #include "routes.h"
 #include "../servehelper.h"
+#include "../htmlhelper.h"
 #include "../client.h"
 #include "../models.h"
 #include "../timeutils.h"
--- a/routes/user_settings.cpp
+++ b/routes/user_settings.cpp
@ -1,23 +1,58 @@
 #include "routes.h"
+#include "../hex.h"
+#include "../config.h"
 #include "../servehelper.h"
 #include "../settings.h"
 #include "../timeutils.h"
 #include "../curlu_wrapper.h"
+#include "../openssl_wrapper.h"

-static void set_cookie(const httplib::Request& req, httplib::Response& res, const char* key, std::string_view value);
+static inline std::string generate_csrf_token(void);
+static inline bool validate_csrf_token(const httplib::Request& req, httplib::Response& res, std::string_view csrf_token, std::string_view query_csrf_token);
+static void set_cookie(const httplib::Request& req, httplib::Response& res, const char* key, std::string_view value, bool session = false);
+static bool safe_memcmp(const char* s1, const char* s2, size_t n);


 void user_settings_route(const httplib::Request& req, httplib::Response& res) {
    UserSettings settings;
+    Cookies cookies = parse_cookies(req);
+    std::string csrf_token;

    if (req.method == "POST") {
+        if (!cookies.contains("csrf-token")) {
+            res.status = 400;
+            serve_error(req, res, "400: Bad Request", "Missing CSRF token cookie, are cookies enabled?");
+            return;
+        }
+        csrf_token = cookies["csrf-token"];
+
+        auto query_csrf_token = req.params.find("csrf-token");
+        if (query_csrf_token == req.params.end()) {
+            res.status = 400;
+            serve_error(req, res, "400: Bad Request", "Missing CSRF token query parameter");
+            return;
+        }
+
+        if (!validate_csrf_token(req, res, csrf_token, query_csrf_token->second)) {
+            return;
+        }
+
        for (const auto& i : req.params) {
            settings.set(i.first, i.second);
        }

        set_cookie(req, res, "auto-open-cw", settings.auto_open_cw ? "true" : "false");
    } else {
-        settings.load_from_cookies(req);
+        for (auto &[name, value] : cookies) {
+            settings.set(name, value);
+        }
+
+        if (cookies.contains("csrf-token")) {
+            csrf_token = cookies["csrf-token"];
+        } else {
+            csrf_token = generate_csrf_token();
+            set_cookie(req, res, "csrf-token", csrf_token, true);
+        }
    }

    Element auto_open_cw_checkbox("input", {{"type", "checkbox"}, {"name", "auto-open-cw"}, {"value", "true"}}, {});
@ -33,6 +68,7 @@ void user_settings_route(const httplib::Request& req, httplib::Response& res) {
            }),

            Element("br"),
+            Element("input", {{"type", "hidden"}, {"name", "csrf-token"}, {"value", csrf_token}}, {}),
            Element("input", {{"type", "submit"}, {"value", "Save"}}, {}),
        }),
        Element("form", {{"class", "user_settings_page-form"}, {"method", "get"}, {"action", get_origin(req)}}, {
@ -49,16 +85,69 @@ void user_settings_route(const httplib::Request& req, httplib::Response& res) {
 }


-static void set_cookie(const httplib::Request& req, httplib::Response& res, const char* key, std::string_view value) {
+static inline std::string generate_csrf_token(void) {
+    std::vector<char> raw_token = secure_random_bytes(32);
+    std::array<char, 32> raw_token_hmac = hmac_sha3_256(config.hmac_key, raw_token);
+
+    return hex_encode(raw_token) + '.' + hex_encode(raw_token_hmac.data(), raw_token_hmac.size());
+}
+
+static inline bool validate_csrf_token(const httplib::Request& req, httplib::Response& res, std::string_view csrf_token, std::string_view query_csrf_token) {
+    if (csrf_token.size() != query_csrf_token.size() || !safe_memcmp(csrf_token.data(), query_csrf_token.data(), csrf_token.size())) {
+        res.status = 400;
+        serve_error(req, res, "400: Bad Request", "CSRF token cookie and CSRF token query parameter do not match");
+        return false;
+    }
+
+    if (csrf_token.size() != 64 + 1 + 64 || csrf_token[64] != '.') {
+        res.status = 400;
+        serve_error(req, res, "400: Bad Request", "CSRF token is in an unknown format");
+        return false;
+    }
+
+    std::vector<char> raw_token, raw_token_hmac;
+    try {
+        raw_token = hex_decode(csrf_token.substr(0, 64));
+        raw_token_hmac = hex_decode(csrf_token.substr(64 + 1, 64));
+    } catch (const std::exception& e) {
+        res.status = 400;
+        serve_error(req, res, "400: Bad Request", "Failed to parse CSRF token", e.what());
+        return false;
+    }
+
+    std::array<char, 32> our_raw_token_hmac = hmac_sha3_256(config.hmac_key, raw_token);
+    if (!safe_memcmp(raw_token_hmac.data(), our_raw_token_hmac.data(), 32)) {
+        res.status = 400;
+        serve_error(req, res, "400: Bad Request", "CSRF token HMAC is not correct");
+        return false;
+    }
+
+    return true;
+}
+
+static void set_cookie(const httplib::Request& req, httplib::Response& res, const char* key, std::string_view value, bool session) {
    CurlUrl origin;
    origin.set(CURLUPART_URL, get_origin(req));

    std::string header = std::string(key) + '=' + std::string(value)
-        + "; HttpOnly; SameSite=Strict; Domain=" + origin.get(CURLUPART_HOST).get() + "; Path=" + origin.get(CURLUPART_PATH).get()
-        + "; Expires=" + to_web_date(current_time() + 365 * 24 * 60 * 60);
+        + "; HttpOnly; SameSite=Lax; Domain=" + origin.get(CURLUPART_HOST).get() + "; Path=" + origin.get(CURLUPART_PATH).get();
+    if (!session) {
+        header += "; Expires=";
+        header += to_web_date(current_time() + 365 * 24 * 60 * 60);
+    }
    if (strcmp(origin.get(CURLUPART_SCHEME).get(), "https") == 0) {
        header += "; Secure";
    }

    res.set_header("Set-Cookie", header);
 }
+
+static bool safe_memcmp(const char* s1, const char* s2, size_t n) {
+    bool equal = true;
+
+    for (size_t i = 0; i < n; i++) {
+        equal &= s1[i] == s2[i];
+    }
+
+    return equal;
+}
--- a/servehelper.cpp
+++ b/servehelper.cpp
@ -4,31 +4,14 @@
 #include <FastHash.h>
 #include <curl/curl.h>

-#include "font_awesome.h"
 #include "config.h"
-#include "settings.h"
-#include "models.h"
-#include "timeutils.h"
 #include "servehelper.h"
-#include "lxb_wrapper.h"
 #include "curlu_wrapper.h"
 #include "routes/routes.h"
-#include "blankie/escape.h"

-static inline void preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, lxb_dom_element_t* element);
-static inline void preprocess_link(const httplib::Request& req, const std::string& domain_name, lxb_dom_element_t* element);
-static inline bool should_fix_link(lxb_dom_element_t* element, const std::string& element_cls);
-static inline void get_text_content(lxb_dom_node_t* node, std::string& out);
-static inline lxb_dom_node_t* emojify(lxb_dom_node_t* child, const std::vector<Emoji>& emojis);
-static inline std::vector<lxb_dom_node*> emojify(lxb_dom_document_t* document, std::string str, const std::vector<Emoji>& emojis);
+static inline void parse_cookies(std::string_view str, Cookies& cookies);
+static inline bool lowercase_compare(std::string_view lhs, std::string_view rhs);

-struct PostStatus {
-    const char* icon_html;
-    Node info_node;
-};
-static Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool main_post, const std::optional<PostStatus>& post_status, const Post* reblogged = nullptr);
-static inline Element serialize_media(const Media& media);
-static inline Element serialize_poll(const httplib::Request& req, const Poll& poll);


 void serve(const httplib::Request& req, httplib::Response& res, std::string title, Element element, Nodes extra_head) {
@ -169,434 +152,59 @@ bool should_send_304(const httplib::Request& req, uint64_t hash) {
    return pos != std::string::npos && (pos == 0 || header[pos - 1] != '/');
 }

-Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool pinned, bool main_post) {
+
+Cookies parse_cookies(const httplib::Request& req) {
+    Cookies cookies;
+
+    for (const auto& i : req.headers) {
+        if (lowercase_compare(i.first, "cookie")) {
+            parse_cookies(i.second, cookies);
+        }
+    }
+
+    return cookies;
+}
+
+
+
+static inline void parse_cookies(std::string_view str, Cookies& cookies) {
    using namespace std::string_literals;

-    if (post.reblog) {
-        PostStatus post_status = {
-            fa_retweet,
-            preprocess_html(req, post.account.emojis, post.account.display_name + " boosted"),
+    size_t offset = 0;
+    size_t new_offset = 0;
+    const char* delimiter = "; ";
+    size_t delimiter_len = strlen(delimiter);
+
+    while (offset < str.size()) {
+        new_offset = str.find(delimiter, offset);
+
+        std::string_view item = str.substr(offset, new_offset != std::string_view::npos ? new_offset - offset : std::string_view::npos);
+        size_t equal_offset = item.find('=');
+        if (equal_offset == std::string_view::npos) {
+            throw std::invalid_argument("invalid user setting item: "s + std::string(item));
+        }
+        cookies.insert({std::string(item.substr(0, equal_offset)), std::string(item.substr(equal_offset + 1))});
+
+        if (new_offset == std::string_view::npos) {
+            break;
+        }
+        offset = new_offset + delimiter_len;
+    }
+}
+
+static inline bool lowercase_compare(std::string_view lhs, std::string_view rhs) {
+    if (lhs.size() != rhs.size()) {
+        return false;
+    }
+
+    auto lower = [](char c) {
+        return c >= 'A' && c <= 'Z' ? c - 'A' + 'a' : c;
    };
-        return serialize_post(req, server, *post.reblog, main_post, post_status, &post);
-    } else if (pinned) {
-        PostStatus post_status = {
-            fa_thumbtack,
-            blankie::html::HTMLString("Pinned post"),
-        };
-        return serialize_post(req, server, post, main_post, post_status);
-    } else if (post.in_reply_to_id && post.in_reply_to_account_id && post.account.id == *post.in_reply_to_account_id) {
-        PostStatus post_status = {
-            fa_reply,
-            preprocess_html(req, post.account.emojis, "Replied to "s + post.account.display_name),
-        };
-        return serialize_post(req, server, post, main_post, post_status);
-    } else {
-        return serialize_post(req, server, post, main_post, std::nullopt);
+    for (size_t i = 0; i < lhs.size(); i++) {
+        if (lower(lhs[i]) != lower(rhs[i])) {
+            return false;
        }
    }

-std::string get_text_content(lxb_dom_node_t* child) {
-    std::string out;
-    get_text_content(child, out);
-
-    if (!out.empty()) {
-        size_t remove_from = out.size();
-        while (remove_from && out[remove_from - 1] == '\n') {
-            remove_from--;
-        }
-        // Don't engulf everything, otherwise it crashes
-        // https://ruby.social/@CoralineAda/109951421922797743
-        if (out.size() > remove_from && remove_from != 0) {
-            out.erase(remove_from);
-        }
-    }
-    if (!out.empty()) {
-        size_t remove_to = 0;
-        while (out.size() > remove_to && out[remove_to] == '\n') {
-            remove_to++;
-        }
-        // Don't engulf everything, otherwise it crashes
-        // https://ruby.social/@CoralineAda/109951421922797743
-        if (out.size() > remove_to) {
-            out.erase(0, remove_to);
-        }
-    }
-
-    return out;
-}
-
-std::string get_text_content(blankie::html::HTMLString str) {
-    LXB::HTML::Document document(str.str);
-    return get_text_content(document.body());
-}
-
-blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, const blankie::html::HTMLString& str) {
-    LXB::HTML::Document document(str.str);
-    preprocess_html(req, domain_name, emojis, document.body_element());
-    return blankie::html::HTMLString(document.serialize());
-}
-
-blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::vector<Emoji>& emojis, const std::string& str) {
-    return preprocess_html(req, "", emojis, blankie::html::HTMLString(blankie::html::escape(str)));
-}
-
-
-static inline void preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, lxb_dom_element_t* element) {
-    const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(element, nullptr));
-
-    if (strncmp(tag_name, "A", 2) == 0) {
-        // Proprocess links
-        preprocess_link(req, domain_name, element);
-    }
-
-    // Walk through the element's children
-    lxb_dom_node_t* child = lxb_dom_node_first_child(lxb_dom_interface_node(element));
-    while (child) {
-        if (child->type == LXB_DOM_NODE_TYPE_ELEMENT) {
-            preprocess_html(req, domain_name, emojis, lxb_dom_interface_element(child));
-        } else if (child->type == LXB_DOM_NODE_TYPE_TEXT) {
-            child = emojify(child, emojis);
-        }
-
-        child = lxb_dom_node_next(child);
-    }
-}
-
-static std::regex mention_class_re("\\bmention\\b");
-static inline void preprocess_link(const httplib::Request& req, const std::string& domain_name, lxb_dom_element_t* element) {
-    using namespace std::string_literals;
-
-    // Remove target=...
-    lxb_status_t status = lxb_dom_element_remove_attribute(element, reinterpret_cast<const lxb_char_t*>("target"), 6);
-    if (status != LXB_STATUS_OK) {
-        throw LXB::Exception(status);
-    }
-
-    size_t href_c_len;
-    const lxb_char_t* href_c = lxb_dom_element_get_attribute(element, reinterpret_cast<const lxb_char_t*>("href"), 4, &href_c_len);
-    if (!href_c) {
-        return;
-    }
-    std::string href(reinterpret_cast<const char*>(href_c), href_c_len);
-
-    size_t cls_c_len;
-    const lxb_char_t* cls_c = lxb_dom_element_class(element, &cls_c_len);
-    std::string cls = cls_c ? std::string(reinterpret_cast<const char*>(cls_c), cls_c_len) : "";
-
-    try {
-        CurlUrl href_url;
-        href_url.set(CURLUPART_URL, get_origin(req));
-        href_url.set(CURLUPART_PATH, std::string(href_url.get(CURLUPART_PATH).get()) + req.path);
-        href_url.set(CURLUPART_URL, href);
-
-        CurlUrl instance_url_base;
-        instance_url_base.set(CURLUPART_SCHEME, "https");
-        instance_url_base.set(CURLUPART_HOST, domain_name);
-
-        // .mention is used in note and posts
-        // Instance base is used for link fields
-        if (std::regex_search(cls, mention_class_re) || starts_with(href_url, instance_url_base)) {
-            // Proxy this instance's URLs to Coyote
-            href = proxy_mastodon_url(req, std::move(href));
-
-            lxb_dom_element_set_attribute(element, reinterpret_cast<const lxb_char_t*>("href"), 4, reinterpret_cast<const lxb_char_t*>(href.data()), href.size());
-        }
-    } catch (const CurlUrlException& e) {
-        // example: <a href=""></a> on eldritch.cafe/about
-        if (e.code != CURLUE_MALFORMED_INPUT) {
-            throw;
-        }
-    }
-
-    if (should_fix_link(element, cls)) {
-        // Set the content of each <a> to its href
-        status = lxb_dom_node_text_content_set(lxb_dom_interface_node(element), reinterpret_cast<const lxb_char_t*>(href.data()), href.size());
-        if (status != LXB_STATUS_OK) {
-            throw LXB::Exception(status);
-        }
-    }
-}
-
-static std::regex unhandled_link_re("\\bunhandled-link\\b");
-static inline bool should_fix_link(lxb_dom_element_t* element, const std::string& element_cls) {
-    // https://vt.social/@LucydiaLuminous/111448085044245037
-    if (std::regex_search(element_cls, unhandled_link_re)) {
    return true;
 }
-
-    auto expected_element = [](lxb_dom_node_t* node, const char* expected_cls) {
-        if (!node || node->type != LXB_DOM_NODE_TYPE_ELEMENT) {
-            return false;
-        }
-        lxb_dom_element_t* span = lxb_dom_interface_element(node);
-
-        const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(span, nullptr));
-        if (strncmp(tag_name, "SPAN", 5) != 0) {
-            return false;
-        }
-
-        const lxb_char_t* cls = lxb_dom_element_get_attribute(span, reinterpret_cast<const lxb_char_t*>("class"), 5, nullptr);
-        return cls && strcmp(reinterpret_cast<const char*>(cls), expected_cls) == 0;
-    };
-
-    lxb_dom_node_t* child = lxb_dom_node_first_child(lxb_dom_interface_node(element));
-    if (!expected_element(child, "invisible")) {
-        return false;
-    }
-
-    child = lxb_dom_node_next(child);
-    if (!expected_element(child, "ellipsis") && !expected_element(child, "")) {
-        return false;
-    }
-
-    child = lxb_dom_node_next(child);
-    if (!expected_element(child, "invisible")) {
-        return false;
-    }
-
-    child = lxb_dom_node_next(child);
-    return child == nullptr;
-}
-
-static inline void get_text_content(lxb_dom_node_t* node, std::string& out) {
-    bool is_br = false, is_p = false;
-
-    if (node->type == LXB_DOM_NODE_TYPE_TEXT) {
-        size_t len;
-        const char* text = reinterpret_cast<const char*>(lxb_dom_node_text_content(node, &len));
-
-        out.append(text, len);
-    } else if (node->type == LXB_DOM_NODE_TYPE_ELEMENT) {
-        lxb_dom_element_t* element = lxb_dom_interface_element(node);
-        const char* tag_name = reinterpret_cast<const char*>(lxb_dom_element_tag_name(element, nullptr));
-
-        is_p = strncmp(tag_name, "P", 2) == 0;
-        is_br = strncmp(tag_name, "BR", 3) == 0;
-    }
-
-    if (is_p || is_br) {
-        out.push_back('\n');
-    }
-
-    lxb_dom_node_t* child = lxb_dom_node_first_child(node);
-    while (child) {
-        get_text_content(child, out);
-
-        child = lxb_dom_node_next(child);
-    }
-
-    if (is_p) {
-        out.push_back('\n');
-    }
-}
-
-static inline lxb_dom_node_t* emojify(lxb_dom_node_t* child, const std::vector<Emoji>& emojis) {
-    std::vector<lxb_dom_node_t*> nodes = emojify(child->owner_document, get_text_content(child), emojis);
-
-    lxb_dom_node_insert_after(child, nodes[0]);
-    lxb_dom_node_destroy(child);
-    child = nodes[0];
-
-    for (size_t i = 1; i < nodes.size(); i++) {
-        lxb_dom_node_insert_after(child, nodes[i]);
-        child = nodes[i];
-    }
-
-    return child;
-}
-
-static std::regex shortcode_re(":([a-zA-Z0-9_]+):");
-static inline std::vector<lxb_dom_node_t*> emojify(lxb_dom_document_t* document, std::string str, const std::vector<Emoji>& emojis) {
-    std::string buf;
-    std::smatch sm;
-    std::vector<lxb_dom_node*> res;
-
-    while (std::regex_search(str, sm, shortcode_re)) {
-        buf += sm.prefix();
-
-        std::string group_0 = sm.str(0);
-        auto emoji = std::find_if(emojis.begin(), emojis.end(), [&](const Emoji& i) { return i.shortcode == sm.str(1); });
-        if (emoji != emojis.end()) {
-            res.push_back(lxb_dom_interface_node(lxb_dom_document_create_text_node(document, reinterpret_cast<const lxb_char_t*>(buf.data()), buf.size())));
-            buf.clear();
-
-            lxb_dom_element_t* img = lxb_dom_element_create(document, reinterpret_cast<const lxb_char_t*>("IMG"), 3, nullptr, 0, nullptr, 0, nullptr, 0, false);
-            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("class"), 5, reinterpret_cast<const lxb_char_t*>("custom_emoji"), 12);
-            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("alt"), 3, reinterpret_cast<const lxb_char_t*>(group_0.data()), group_0.size());
-            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("title"), 5, reinterpret_cast<const lxb_char_t*>(group_0.data()), group_0.size());
-            lxb_dom_element_set_attribute(img, reinterpret_cast<const lxb_char_t*>("src"), 3, reinterpret_cast<const lxb_char_t*>(emoji->url.data()), emoji->url.size());
-            res.push_back(lxb_dom_interface_node(img));
-        } else {
-            buf += group_0;
-        }
-
-        str = sm.suffix();
-    }
-
-    if (!str.empty()) {
-        buf += std::move(str);
-    }
-    if (!buf.empty()) {
-        res.push_back(lxb_dom_interface_node(lxb_dom_document_create_text_node(document, reinterpret_cast<const lxb_char_t*>(buf.data()), buf.size())));
-    }
-
-    return res;
-}
-
-static Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool main_post, const std::optional<PostStatus>& post_status, const Post* reblogged) {
-    using namespace std::string_literals;
-
-    bool user_known = !post.account.id.empty();
-    bool user_ref_known = !post.account.username.empty() && !post.account.server.empty();
-    // `reblogged == nullptr` since a malicious server could take down the frontend
-    // by sending a post that is not a reblog with no account information
-    std::string post_url = user_known || reblogged == nullptr
-        ? get_origin(req) + '/' + server + "/@" + post.account.acct(false) + '/' + post.id + "#m"
-        : get_origin(req) + '/' + server + "/@" + reblogged->account.acct(false) + '/' + reblogged->id + "#m";
-
-    std::string time_title = post.edited_at < 0
-        ? full_time(post.created_at)
-        : "Created: "s + full_time(post.created_at) + "\nEdited: " + full_time(post.edited_at);
-    const char* time_badge = post.edited_at < 0 ? "" : " (edited)";
-
-    blankie::html::HTMLString preprocessed_html = preprocess_html(req, server, post.emojis, post.content);
-    // Workaround for https://vt.social/@a1ba@suya.place/110552480243348878#m
-    if (preprocessed_html.str.find("<p>") == std::string::npos) {
-        preprocessed_html.str.reserve(preprocessed_html.str.size() + 3 + 4);
-        preprocessed_html.str.insert(0, "<p>");
-        preprocessed_html.str.append("</p>");
-    }
-    Element contents("div", {{"class", "post-contents"}}, {std::move(preprocessed_html)});
-
-    Element post_attachments("div", {{"class", "post-attachments"}}, {});
-    post_attachments.nodes.reserve(post.media_attachments.size());
-    for (const Media& media : post.media_attachments) {
-        post_attachments.nodes.push_back(serialize_media(media));
-    }
-    contents.nodes.push_back(std::move(post_attachments));
-
-    if (post.poll) {
-        contents.nodes.push_back(serialize_poll(req, *post.poll));
-    }
-
-    if (post.sensitive) {
-        std::string spoiler_text = !post.spoiler_text.empty() ? post.spoiler_text : "See more";
-        contents = Element("details", {
-            Element("summary", {preprocess_html(req, post.emojis, std::move(spoiler_text))}),
-            std::move(contents),
-        });
-        if (UserSettings(req).auto_open_cw) {
-            contents.attributes.push_back({"open", ""});
-        }
-    }
-
-    Element div("div", {{"class", "post"}}, {
-        Element("div", {{"class", "post-header"}}, {
-            user_ref_known ? Element("a", {{"href", get_origin(req) + '/' + server + "/@" + post.account.acct(false)}}, {
-                !post.account.avatar_static.empty()
-                    ? Element("img", {{"class", "post-avatar"}, {"alt", "User profile picture"}, {"loading", "lazy"}, {"src", post.account.avatar_static}}, {})
-                    : Node(""),
-                Element("span", {
-                    Element("b", {preprocess_html(req, post.account.emojis, post.account.display_name)}),
-                    Element("br"), "@", post.account.acct(),
-                }),
-            }) : Element("b", {"Unknown user"}),
-            Element("a", {{"class", "post-time_header"}, {"href", std::move(post_url)}, {"title", time_title}}, {
-                Element("time", {{"datetime", to_rfc3339(post.created_at)}}, {relative_time(post.created_at, current_time()), time_badge}),
-            }),
-        }),
-
-        contents,
-    });
-    if (post_status) {
-        div.nodes.insert(div.nodes.begin(), Element("p", {
-            blankie::html::HTMLString(post_status->icon_html), " ", post_status->info_node,
-        }));
-    }
-    if (main_post) {
-        div.attributes = {{"class", "post main_post"}, {"id", "m"}};
-    }
-
-    return div;
-}
-
-static inline Element serialize_media(const Media& media) {
-    Element element = [&]() {
-        if (media.type == "image") {
-            return Element("a", {{"href", media.url}}, {
-                Element("img", {{"loading", "lazy"}, {"src", media.preview_url.value_or(media.url)}}, {}),
-            });
-        } else if (media.type == "video") {
-            Element video("video", {{"controls", ""}, {"src", media.url}}, {});
-            if (media.preview_url) {
-                video.attributes.push_back({"poster", *media.preview_url});
-            }
-            return video;
-        } else if (media.type == "audio") {
-            return Element("audio", {{"controls", ""}, {"src", media.url}}, {});
-        } else if (media.type == "gifv") {
-            // https://hachyderm.io/@Impossible_PhD/111444541628207638
-            Element video("video", {{"controls", ""}, {"loop", ""}, {"muted", ""}, {"autoplay", ""}, {"src", media.url}}, {});
-            if (media.preview_url) {
-                video.attributes.push_back({"poster", *media.preview_url});
-            }
-            return video;
-        } else if (media.type == "unknown") {
-            if (media.remote_url) {
-                // https://botsin.space/@lina@vt.social/111053598696451525
-                return Element("a", {{"class", "unknown_media"}, {"href", *media.remote_url}}, {"Media is not available from this instance, view externally"});
-            } else {
-                return Element("p", {{"class", "unknown_media"}}, {"Media is not available from this instance"});
-            }
-        } else {
-            return Element("p", {"Unsupported media type: ", media.type});
-        }
-    }();
-
-    if (media.description) {
-        element.attributes.push_back({"alt", *media.description});
-        element.attributes.push_back({"title", *media.description});
-    }
-
-    return element;
-}
-
-static inline Element serialize_poll(const httplib::Request& req, const Poll& poll) {
-    using namespace std::string_literals;
-
-    uint64_t voters_count = poll.voters_count >= 0 ? static_cast<uint64_t>(poll.voters_count) : poll.votes_count;
-    Element div("div");
-
-    auto pick_form = [](uint64_t count, const char* singular, const char* plural) {
-        return count == 1 ? singular : plural;
-    };
-
-    div.nodes.reserve(poll.options.size() + 1);
-    for (const PollOption& option : poll.options) {
-        std::string percentage = voters_count
-            ? std::to_string(option.votes_count * 100 / voters_count) + '%'
-            : "0%";
-
-        div.nodes.push_back(Element("div", {{"class", "poll-option"}, {"title", std::to_string(option.votes_count) + pick_form(option.votes_count, " vote", " votes")}}, {
-            Element("b", {{"class", "poll-percentage"}}, {percentage}), " ", preprocess_html(req, poll.emojis, option.title),
-            Element("object", {{"class", "poll-bar"}, {"width", percentage}}, {}),
-        }));
-    }
-
-    Element p("p", poll.voters_count >= 0
-        ? std::vector<Node>({std::to_string(voters_count), " ", pick_form(voters_count, "voter", "voters")})
-        : std::vector<Node>({std::to_string(poll.votes_count), " ", pick_form(poll.votes_count, "vote", "votes")})
-    );
-    if (poll.expired) {
-        p.nodes.push_back(" / ");
-        p.nodes.push_back(Element("time", {{"datetime", to_rfc3339(poll.expires_at)}, {"title", "Expired on "s + full_time(poll.expires_at)}}, {"Expired"}));
-    } else if (poll.expires_at >= 0) {
-        p.nodes.push_back(" / ");
-        p.nodes.push_back(Element("time", {{"datetime", to_rfc3339(poll.expires_at)}, {"title", full_time(poll.expires_at)}}, {
-            "Expires in ", relative_time(current_time(), poll.expires_at),
-        }));
-    }
-    div.nodes.push_back(std::move(p));
-
-    return div;
-}
--- a/servehelper.h
+++ b/servehelper.h
@ -1,17 +1,16 @@
 #pragma once

 #include <optional>
+#include <unordered_map>

 #include <httplib/httplib.h>
 #include "blankie/serializer.h"
-#include "lxb_wrapper.h"
-struct Post; // forward declaration from models.h
-struct Emoji; // forward declaration from models.h
 class CurlUrl; // forward declaration from curlu_wrapper.h

 using Element = blankie::html::Element;
 using Node = blankie::html::Node;
 using Nodes = std::vector<Node>;
+using Cookies = std::unordered_map<std::string, std::string>;

 void serve(const httplib::Request& req, httplib::Response& res, std::string title, Element element, Nodes extra_head = {});
 void serve_error(const httplib::Request& req, httplib::Response& res,
@ -23,9 +22,4 @@ std::string get_origin(const httplib::Request& req);
 std::string proxy_mastodon_url(const httplib::Request& req, const std::string& url_str);
 bool should_send_304(const httplib::Request& req, uint64_t hash);

-Element serialize_post(const httplib::Request& req, const std::string& server, const Post& post, bool pinned = false, bool main_post = false);
-
-std::string get_text_content(lxb_dom_node_t* child);
-std::string get_text_content(blankie::html::HTMLString str);
-blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::string& domain_name, const std::vector<Emoji>& emojis, const blankie::html::HTMLString& str);
-blankie::html::HTMLString preprocess_html(const httplib::Request& req, const std::vector<Emoji>& emojis, const std::string& str);
+Cookies parse_cookies(const httplib::Request& req);
--- a/settings.cpp
+++ b/settings.cpp
@ -1,9 +1,8 @@
 #include <string>
 #include <stdexcept>
 #include "settings.h"
+#include "servehelper.h"

-static void set_settings(std::string_view str, const char* delimiter, UserSettings& settings);
-static inline bool lowercase_compare(std::string_view lhs, std::string_view rhs);
 static bool parse_bool(std::string_view value);


@ -14,54 +13,14 @@ void UserSettings::set(std::string_view key, std::string_view value) {
 }

 void UserSettings::load_from_cookies(const httplib::Request& req) {
-    for (const auto& i : req.headers) {
-        if (lowercase_compare(i.first, "cookie")) {
-            set_settings(i.second, "; ", *this);
-        }
+    Cookies cookies = parse_cookies(req);
+
+    for (auto &[name, value] : cookies) {
+        this->set(name, value);
    }
 }


-static void set_settings(std::string_view str, const char* delimiter, UserSettings& settings) {
-    using namespace std::string_literals;
-    size_t offset = 0;
-    size_t new_offset = 0;
-    size_t delimiter_len = strlen(delimiter);
-
-    while (offset < str.size()) {
-        new_offset = str.find(delimiter, offset);
-
-        std::string_view item = str.substr(offset, new_offset != std::string_view::npos ? new_offset - offset : std::string_view::npos);
-        size_t equal_offset = item.find('=');
-        if (equal_offset == std::string_view::npos) {
-            throw std::invalid_argument("invalid user setting item: "s + std::string(item));
-        }
-        settings.set(item.substr(0, equal_offset), item.substr(equal_offset + 1));
-
-        if (new_offset == std::string_view::npos) {
-            break;
-        }
-        offset = new_offset + delimiter_len;
-    }
-}
-
-static inline bool lowercase_compare(std::string_view lhs, std::string_view rhs) {
-    if (lhs.size() != rhs.size()) {
-        return false;
-    }
-
-    auto lower = [](char c) {
-        return c >= 'A' && c <= 'Z' ? c - 'A' + 'a' : c;
-    };
-    for (size_t i = 0; i < lhs.size(); i++) {
-        if (lower(lhs[i]) != lower(rhs[i])) {
-            return false;
-        }
-    }
-
-    return true;
-}
-
 static bool parse_bool(std::string_view value) {
    using namespace std::string_literals;
Author	SHA1	Message	Date
blankie	9b21060b3a	Add CSRF protection	2023-12-08 22:27:40 +11:00
blankie	973a0eada2	Split HTML helpers into their own file	2023-12-08 22:27:40 +11:00
blankie	d1711a040b	Move hex-related functions into their own file	2023-12-08 22:27:40 +11:00
blankie	9b6f6a4bcc	Relax SameSite	2023-12-08 18:12:51 +11:00