http地址支持相对路径 (#2223)

* 部分hls中会出现类似[../]或者[./]的相对地址, 因此写了一个解析的方法

* 有时候上游http协议不规范会导致直接输出eof, 特别是在自动重试机制开启下, 容易造成定时器没办法初始化从而拉不到流但是session一直存在,所以需要做一个判断,如果没数据直接报错

src/Common/Parser.cpp

@@ -159,7 +159,81 @@ StrCaseMap Parser::parseArgs(const string &str, const char *pair_delim, const ch
     }
     return ret;
 }
+std::string Parser::merge_url(const string &base_url, const string &path) {
+    //以base_url为基础, 合并path路径生成新的url, path支持相对路径和绝对路径
+    if (base_url.empty()) {
+        return path;
+    }
+    if (path.empty()) {
+        return base_url;
+    }
+    // 如果包含协议，则直接返回
+    if (path.find("://") != string::npos) {
+        return path;
+    }
 
+    string protocol = "http://";
+    size_t protocol_end = base_url.find("://");
+    if (protocol_end != string::npos) {
+        protocol = base_url.substr(0, protocol_end + 3);
+    }
+    // 如果path以"//"开头，则直接拼接协议
+    if (path.find("//") == 0) {
+        return protocol + path.substr(2);
+    }
+    string host;
+    size_t pos = 0;
+    if (protocol_end != string::npos) {
+        pos = base_url.find('/', protocol_end + 3);
+        host = base_url.substr(0, pos);
+        if (pos == string::npos) {
+            pos = base_url.size();
+        } else {
+            pos++;
+        }
+    }
+    // 如果path以"/"开头，则直接拼接协议和主机
+    if (path[0] == '/') {
+        return host + path;
+    }
+    vector<string> path_parts;
+    size_t next_pos = 0;
+    if (!host.empty()) {
+        path_parts.emplace_back(host);
+    }
+    while ((next_pos = base_url.find('/', pos)) != string::npos) {
+        path_parts.emplace_back(base_url.substr(pos, next_pos - pos));
+        pos = next_pos + 1;
+    }
+    pos = 0;
+    while ((next_pos = path.find('/', pos)) != string::npos) {
+        string part = path.substr(pos, next_pos - pos);
+        if (part == "..") {
+            if (!path_parts.empty() && !path_parts.back().empty()) {
+                if (path_parts.size() > 1 || protocol_end == string::npos) {
+                    path_parts.pop_back();
+                }
+            }
+        } else if (part != "." && !part.empty()) {
+            path_parts.emplace_back(part);
+        }
+        pos = next_pos + 1;
+    }
+
+    string part = path.substr(pos);
+    if (part != ".." && part != "." && !part.empty()) {
+        path_parts.emplace_back(part);
+    }
+    stringstream final_url;
+    for (size_t i = 0; i < path_parts.size(); ++i) {
+        if (i == 0) {
+            final_url << path_parts[i];
+        } else {
+            final_url << '/' << path_parts[i];
+        }
+    }
+    return final_url.str();
+}
 void RtspUrl::parse(const string &strUrl) {
     auto schema = FindField(strUrl.data(), nullptr, "://");
     bool is_ssl = strcasecmp(schema.data(), "rtsps") == 0;
@@ -238,4 +312,4 @@ static onceToken token([](){
 });
 #endif
 
-}//namespace mediakit
+}//namespace mediakit

src/Common/Parser.h

@@ -105,6 +105,8 @@ public:
     //解析?后面的参数
     static StrCaseMap parseArgs(const std::string &str, const char *pair_delim = "&", const char *key_delim = "=");
 
+    static std::string merge_url(const std::string &base_url, const std::string &path);
+
 private:
     std::string _strMethod;
     std::string _strUrl;

src/Http/HlsParser.cpp

@@ -37,15 +37,7 @@ bool HlsParser::parse(const string &http_url, const string &m3u8) {
 
         if ((_is_m3u8_inner || extinf_dur != 0) && line[0] != '#') {
             segment.duration = extinf_dur;
-            if (line.find("http://") == 0 || line.find("https://") == 0) {
-                segment.url = line;
-            } else {
-                if (line.find("/") == 0) {
-                    segment.url = http_url.substr(0, http_url.find("/", 8)) + line;
-                } else {
-                    segment.url = http_url.substr(0, http_url.rfind("/") + 1) + line;
-                }
-            }
+            segment.url = Parser::merge_url(http_url, line);
             if (!_is_m3u8_inner) {
                 //ts按照先后顺序排序
                 ts_map.emplace(index++, segment);

src/Http/HttpClient.cpp

@@ -182,8 +182,8 @@ void HttpClient::onErr(const SockException &ex) {
 
 ssize_t HttpClient::onRecvHeader(const char *data, size_t len) {
     _parser.Parse(data);
-    if (_parser.Url() == "302" || _parser.Url() == "301") {
-        auto new_url = _parser["Location"];
+    if (_parser.Url() == "302" || _parser.Url() == "301" || _parser.Url() == "303") {
+        auto new_url = Parser::merge_url(_url, _parser["Location"]);
         if (new_url.empty()) {
             throw invalid_argument("未找到Location字段(跳转url)");
         }
@@ -206,7 +206,11 @@ ssize_t HttpClient::onRecvHeader(const char *data, size_t len) {
                 onResponseBody(data, len);
             } else {
                 _total_body_size = _recved_body_size;
-                onResponseCompleted_l(SockException(Err_success, "success"));
+                if (_recved_body_size > 0) {
+                    onResponseCompleted_l(SockException(Err_success, "success"));
+                }else{
+                    onResponseCompleted_l(SockException(Err_other, "no body"));
+                }
             }
         });
         //后续为源源不断的body