--- parser3/src/main/pa_common.C 2003/04/07 15:32:57 1.143.2.21.2.23 +++ parser3/src/main/pa_common.C 2003/09/26 13:37:20 1.158 @@ -5,18 +5,22 @@ Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_COMMON_C="$Date: 2003/04/07 15:32:57 $"; +static const char* IDENT_COMMON_C="$Date: 2003/09/26 13:37:20 $"; #include "pa_common.h" #include "pa_exception.h" #include "pa_hash.h" #include "pa_globals.h" +#include "pa_request_charsets.h" +#include "pa_charsets.h" #define PA_HTTP #ifdef PA_HTTP #include "pa_vstring.h" #include "pa_vint.h" +#include "pa_vhash.h" +#include "pa_vtable.h" #ifdef CYGWIN #define _GNU_H_WINDOWS32_SOCKETS @@ -130,6 +134,8 @@ const String file_status_name(FILE_STATU #define HTTP_TIMEOUT_NAME "timeout" #define HTTP_HEADERS_NAME "headers" #define HTTP_ANY_STATUS_NAME "any-status" +#define HTTP_CHARSET_NAME "charset" +#define HTTP_TABLES_NAME "tables" // statics @@ -137,6 +143,8 @@ static const String http_method_name(HTT static const String http_timeout_name(HTTP_TIMEOUT_NAME); static const String http_headers_name(HTTP_HEADERS_NAME); static const String http_any_status_name(HTTP_ANY_STATUS_NAME); +static const String http_charset_name(HTTP_CHARSET_NAME); +static const String http_tables_name(HTTP_TABLES_NAME); // defines @@ -164,23 +172,27 @@ void fix_line_breaks(char *str, size_t& } else // \r, not \n = Macintosh bol=eol+1; } - // last piece without \r, including terminating 0 + // last piece without \r if(dest!=bol) memcpy(dest, bol, eob-bol); + str[length]=0; // terminating } -char* file_read_text(Charset& charset, - const String& file_spec, - bool fail_on_read_problem, - HashStringValue* params/*, HashStringValue* * out_fields*/) { +char* file_read_text(Request_charsets& charsets, + const String& file_spec, + bool fail_on_read_problem, + HashStringValue* params/*, HashStringValue* * out_fields*/) { File_read_result file= - file_read(charset, file_spec, true, params, fail_on_read_problem); + file_read(charsets, file_spec, true, params, fail_on_read_problem); return file.success?file.str:0; } #ifdef PA_HTTP //http request stuff +#undef CRLF +#define CRLF "\r\n" + static bool set_addr(struct sockaddr_in *addr, const char* host, const short port){ memset(addr, 0, sizeof(*addr)); addr->sin_family=AF_INET; @@ -204,7 +216,7 @@ static int http_read_response(String& re if(size<=0) break; response.append_strdup(buf, size, String::L_TAINTED); - if(!result && (EOLat=response.pos("\r\n", 2))!=STRING_NOT_FOUND) { // checking status in first response + if(!result && (EOLat=response.pos(CRLF, 2))!=STRING_NOT_FOUND) { // checking status in first response const String& status_line=response.mid(0, (size_t)EOLat); ArrayString astatus; size_t pos_after=0; @@ -229,18 +241,17 @@ static int http_read_response(String& re /* ********************** request *************************** */ #if defined(SIGALRM) && defined(HAVE_SIGSETJMP) && defined(HAVE_SIGLONGJMP) -# define WE_CAN_USE_ALARM +# define PA_USE_ALARM #endif -#ifdef WE_CAN_USE_ALARM +#ifdef PA_USE_ALARM static sigjmp_buf timeout_env; static void timeout_handler(int sig){ siglongjmp(timeout_env, 1); } #endif -static int http_request( - String& response, +static int http_request(String& response, const char* host, int port, const char* request, int timeout, @@ -250,23 +261,29 @@ static int http_request( 0, "zero hostname"); //never -#ifdef WE_CAN_USE_ALARM - signal(SIGALRM, timeout_handler); +#ifdef PA_USE_ALARM + signal(SIGALRM, timeout_handler); #endif int sock=-1; - try { - int result; -#ifdef WE_CAN_USE_ALARM - if(sigsetjmp(timeout_env, 1)) - throw Exception("http.timeout", - 0, - "timeout occured while retrieving document"); - else { - alarm(timeout); +#ifdef PA_USE_ALARM + if(sigsetjmp(timeout_env, 1)) { + // stupid gcc [2.95.4] generated bad code + // which failed to handle sigsetjmp+throw: crashed inside of pre-throw code. + // rewritten simplier [though duplicating closesocket code] + if(sock>=0) + closesocket(sock); + throw Exception("http.timeout", + origin_string, + "timeout occured while retrieving document"); + return 0; // never + } else { + alarm(timeout); #endif + try { + int result; struct sockaddr_in dest; - if(!set_addr(&dest, host, port)) + if(!set_addr(&dest, host, port)) throw Exception("http.host", 0, "can not resolve hostname \"%s\"", host); @@ -287,44 +304,84 @@ static int http_request( result=http_read_response(response, sock, fail_on_status_ne_200); closesocket(sock); -#ifdef WE_CAN_USE_ALARM +#ifdef PA_USE_ALARM alarm(0); - } #endif - return result; - } catch(...) { - if(sock>=0) - closesocket(sock); -#ifdef WE_CAN_USE_ALARM - alarm(0); + return result; + } catch(...) { +#ifdef PA_USE_ALARM + alarm(0); #endif - rethrow; + if(sock>=0) + closesocket(sock); + rethrow; + } +#ifdef PA_USE_ALARM } +#endif } #ifndef DOXYGEN struct Http_pass_header_info { - Charset* charset; + Request_charsets* charsets; String* request; bool user_agent_specified; }; #endif -static void http_pass_header(HashStringValue::key_type key, HashStringValue::value_type value, +static void http_pass_header(HashStringValue::key_type key, + HashStringValue::value_type value, Http_pass_header_info *info) { *info->request <charset, String::CC_UPPER)=="USER-AGENT") - info->user_agent_specified=true; + if(String(key, String::L_TAINTED).change_case(info->charsets->source(), String::CC_UPPER)=="USER-AGENT") + info->user_agent_specified=true; } + + +static Charset* detect_charset(Charset& source_charset, const String& content_type_value) { + const String::Body CONTENT_TYPE_VALUE= + content_type_value.change_case(source_charset, String::CC_UPPER); + // content-type: xxx/xxx; source_charset=WE-NEED-THIS + // content-type: xxx/xxx; source_charset="WE-NEED-THIS" + // content-type: xxx/xxx; source_charset="WE-NEED-THIS"; + size_t before_charseteq_pos=CONTENT_TYPE_VALUE.pos("CHARSET="); + if(before_charseteq_pos!=STRING_NOT_FOUND) { + size_t charset_begin=before_charseteq_pos+8/*CHARSET="*/; + size_t open_quote_pos=CONTENT_TYPE_VALUE.pos('"', charset_begin); + bool quoted=open_quote_pos==charset_begin; + if(quoted) + charset_begin++; // skip opening '"' + size_t charset_end=CONTENT_TYPE_VALUE.length(); + if(quoted) { + size_t close_quote_pos=CONTENT_TYPE_VALUE.pos('"', charset_begin); + if(close_quote_pos!=STRING_NOT_FOUND) + charset_end=close_quote_pos; + } else { + size_t delim_pos=CONTENT_TYPE_VALUE.pos(';', charset_begin); + if(delim_pos!=STRING_NOT_FOUND) + charset_end=delim_pos; + } + const String::Body CHARSET_NAME_BODY= + CONTENT_TYPE_VALUE.mid(charset_begin, charset_end); + + return &charsets.get(CHARSET_NAME_BODY); + } + + return 0; +} + + #ifndef DOXYGEN struct File_read_http_result { char *str; size_t length; HashStringValue* headers; }; #endif -static File_read_http_result file_read_http(Charset& charset, const String& file_spec, +/// @todo build .cookies field. use ^file.tables.SET-COOKIES.menu{ for now +static File_read_http_result file_read_http(Request_charsets& charsets, + const String& file_spec, HashStringValue *options=0) { File_read_http_result result; char host[MAX_STRING]; @@ -334,25 +391,7 @@ static File_read_http_result file_read_h int timeout=2; bool fail_on_status_ne_200=true; Value* vheaders=0; - - String& connect_string=*new String; - // not in ^sql{... L_SQL ...} spirit, but closer to ^file::load one - connect_string.append(file_spec, String::L_URI); // tainted pieces -> URI pieces - - const char* connect_string_cstr=connect_string.cstr(String::L_UNSPECIFIED); - const char* current=connect_string_cstr; - if(strncmp(current, "http://", 7)!=0) - throw Exception(0, - &connect_string, - "does not start with http://"); //never - current+=7; - - strncpy(host, current, sizeof(host)-1); host[sizeof(host)-1]=0; - char* host_uri=lsplit(host, '/'); - uri=host_uri?current+(host_uri-1-host):"/"; - char* port_cstr=lsplit(host, ':'); - char* error_pos=0; - port=port_cstr?strtol(port_cstr, &error_pos, 0):80; + Charset *asked_remote_charset=0; if(options) { int valid_options=0; @@ -371,59 +410,107 @@ static File_read_http_result file_read_h valid_options++; fail_on_status_ne_200=!vany_status->as_bool(); } + if(Value* vcharset_name=options->get(http_charset_name)) { + valid_options++; + asked_remote_charset=&::charsets.get(vcharset_name->as_string(). + change_case(charsets.source(), String::CC_UPPER)); + } if(valid_options!=options->count()) throw Exception("parser.runtime", 0, "invalid option passed"); - } + } + if(!asked_remote_charset) // defaulting to $request:charset + asked_remote_charset=&charsets.source(); - //making request - String request; - if(method) - request<is_string()) { // allow empty - if(HashStringValue *headers=vheaders->get_hash()) { - Http_pass_header_info info={&charset, &request}; - headers->for_each(http_pass_header, &info); - user_agent_specified=info.user_agent_specified; - } else - throw Exception("parser.runtime", - &connect_string, - "headers param must be hash"); - }; - if(!user_agent_specified) // defaulting - request << "user-agent: " DEFAULT_USER_AGENT "\n"; - request<<"\n"; + //preparing request + String& connect_string=*new String; + // not in ^sql{... L_SQL ...} spirit, but closer to ^file::load one + connect_string.append(file_spec, String::L_URI); // tainted pieces -> URI pieces + + const char* request_cstr; + { + // influence URLencoding of tainted pieces to String::L_URI lang + Temp_client_charset temp(charsets, *asked_remote_charset); + + const char* connect_string_cstr=connect_string.cstr(String::L_UNSPECIFIED); + + const char* current=connect_string_cstr; + if(strncmp(current, "http://", 7)!=0) + throw Exception(0, + &connect_string, + "does not start with http://"); //never + current+=7; + + strncpy(host, current, sizeof(host)-1); host[sizeof(host)-1]=0; + char* host_uri=lsplit(host, '/'); + uri=host_uri?current+(host_uri-1-host):"/"; + char* port_cstr=lsplit(host, ':'); + char* error_pos=0; + port=port_cstr?strtol(port_cstr, &error_pos, 0):80; + + + //making request + String request; + if(method) + request<is_string()) { // allow empty + if(HashStringValue *headers=vheaders->get_hash()) { + Http_pass_header_info info={&charsets, &request}; + headers->for_each(http_pass_header, &info); + user_agent_specified=info.user_agent_specified; + } else + throw Exception("parser.runtime", + &connect_string, + "headers param must be hash"); + }; + if(!user_agent_specified) // defaulting + request << "user-agent: " DEFAULT_USER_AGENT CRLF; + request << CRLF; + + request_cstr=request.cstr(String::L_UNSPECIFIED); + } + // recode those pieces which are not in String::L_URI lang + // [those violating HTTP standard, but widly used] + request_cstr=Charset::transcode( + String::C(request_cstr, strlen(request_cstr)), + charsets.source(), + *asked_remote_charset); //sending request String response; - const char* request_cstr=request.cstr(String::L_UNSPECIFIED); int status_code=http_request(response, host, port, request_cstr, timeout, fail_on_status_ne_200); //processing results - size_t pos=response.pos("\r\n\r\n", 4); - if(pos==STRING_NOT_FOUND || pos<1){ + size_t pos=response.pos(CRLF CRLF, 4); + if(pos==STRING_NOT_FOUND || pos<1) { throw Exception("http.response", &connect_string, "bad response from host - no headers found"); } const String& header_block=response.mid(0, pos); - const String& body=response.mid(pos+4, response.length()); + const String& raw_body=response.mid(pos+4, response.length()); ArrayString aheaders; result.headers=new HashStringValue; + VHash* vtables=new VHash; + result.headers->put(http_tables_name, vtables); + HashStringValue& tables=vtables->hash(); + size_t pos_after=0; - header_block.split(aheaders, pos_after, "\r\n"); + header_block.split(aheaders, pos_after, CRLF); //processing headers size_t aheaders_count=aheaders.count(); + Charset* real_remote_charset=0; // undetected, yet for(size_t i=1; iput( - line.mid(0, pos).change_case(charset, String::CC_UPPER), - new VString(line.mid(pos+2, line.length()))); + const String::Body HEADER_NAME= + line.mid(0, pos).change_case(charsets.source(), String::CC_UPPER); + const String& header_value=line.mid(pos+2, line.length()); + if(HEADER_NAME=="CONTENT-TYPE") + real_remote_charset=detect_charset(charsets.source(), header_value); + + // tables + { + Value *valready=(Value *)tables.get(HEADER_NAME); + bool existed=valready!=0; + Table *table; + if(existed) { + // second+ appearence + table=valready->get_table(); + } else { + // first appearence + Table::columns_type columns =new ArrayString(1); + *columns+=new String("value"); + table=new Table(columns); + } + // this string becomes next row + ArrayString& row=*new ArrayString(1); + row+=&header_value; + *table+=&row; + // not existed before? add it + if(!existed) + tables.put(HEADER_NAME, new VTable(table)); + } + + result.headers->put(HEADER_NAME, new VString(header_value)); } + // defaulting to used-asked charset [it's never empty!] + if(!real_remote_charset) + real_remote_charset=asked_remote_charset; // output response - result.str=body.cstrm(); - result.length=body.length(); + String::C real_body=Charset::transcode( + String::C(raw_body.cstrm()/*must be modifiable*/, raw_body.length()), + *real_remote_charset, + charsets.source()); + + result.str=const_cast(real_body.str); // hacking a little + result.length=real_body.length; result.headers->put(file_status_name, new VInt(status_code)); return result; } @@ -452,10 +573,10 @@ struct File_read_action_info { }; #endif static void file_read_action( - struct stat& finfo, - int f, - const String& file_spec, const char* fname, bool as_text, - void *context) { + struct stat& finfo, + int f, + const String& file_spec, const char* fname, bool as_text, + void *context) { File_read_action_info& info=*static_cast(context); if(size_t to_read_size=(size_t)finfo.st_size) { *info.data=new(PointerFreeGC) char[to_read_size+(as_text?1:0)]; @@ -476,14 +597,14 @@ static void file_read_action( return; } } -File_read_result file_read(Charset& charset, const String& file_spec, +File_read_result file_read(Request_charsets& charsets, const String& file_spec, bool as_text, HashStringValue *params, bool fail_on_read_problem) { File_read_result result={false}; #ifdef PA_HTTP if(file_spec.starts_with("http://")) { // fail on read problem - File_read_http_result http=file_read_http(charset, file_spec, params); + File_read_http_result http=file_read_http(charsets, file_spec, params); result.success=true; result.str=http.str; result.length=http.length; @@ -516,7 +637,7 @@ File_read_result file_read(Charset& char } #ifdef PA_SAFE_MODE -void check_safe_mode(stat finfo, const String& file_spec, const char* fname) { +void check_safe_mode(struct stat finfo, const String& file_spec, const char* fname) { if(finfo.st_uid/*foreign?*/!=geteuid() && finfo.st_gid/*foreign?*/!=getegid()) throw Exception("parser.runtime",