--- parser3/src/main/pa_common.C 2003/03/20 11:11:55 1.143.2.21.2.6 +++ parser3/src/main/pa_common.C 2004/12/10 08:37:53 1.201 @@ -1,22 +1,27 @@ /** @file Parser: commonly functions. - Copyright(c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com) + Copyright(c) 2001-2004 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_COMMON_C="$Date: 2003/03/20 11:11:55 $"; +static const char * const IDENT_COMMON_C="$Date: 2004/12/10 08:37:53 $"; #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 +#define PA_HTTP #ifdef PA_HTTP #include "pa_vstring.h" #include "pa_vint.h" +#include "pa_vhash.h" +#include "pa_vtable.h" +#include "pa_socks.h" #ifdef CYGWIN #define _GNU_H_WINDOWS32_SOCKETS @@ -60,50 +65,9 @@ int PASCAL closesocket(SOCKET); # endif #endif -// locking constants - -#ifdef HAVE_FLOCK - -static int lock_shared_blocking(int fd) { return flock(fd, LOCK_SH); } -static int lock_exclusive_blocking(int fd) { return flock(fd, LOCK_EX); } -static int lock_exclusive_nonblocking(int fd) { return flock(fd, LOCK_EX || LOCK_NB); } -static int unlock(int fd) { return flock(fd, LOCK_UN); } - -#else -#ifdef HAVE__LOCKING - -#define FLOCK(operation) lseek(fd, 0, SEEK_SET); return _locking(fd, operation, 1) -static int lock_shared_blocking(int fd) { FLOCK(_LK_LOCK); } -static int lock_exclusive_blocking(int fd) { FLOCK(_LK_LOCK); } -static int lock_exclusive_nonblocking(int fd) { FLOCK(_LK_NBLCK); } -static int unlock(int fd) { FLOCK(_LK_UNLCK); } - -#else -#ifdef HAVE_FCNTL - -#define FLOCK(cmd, arg) struct flock ls={arg, SEEK_SET}; return fcntl(fd, cmd, &ls) -static int lock_shared_blocking(int fd) { FLOCK(F_SETLKW, F_RDLCK); } -static int lock_exclusive_blocking(int fd) { FLOCK(F_SETLKW, F_WRLCK); } -static int lock_exclusive_nonblocking(int fd) { FLOCK(F_SETLK, F_RDLCK); } -static int unlock(int fd) { FLOCK(F_SETLK, F_UNLCK); } - -#else -#ifdef HAVE_LOCKF - -#define FLOCK(fd, operation) lseek(fd, 0, SEEK_SET); return lockf(fd, operation, 1) -static int lock_shared_blocking(int fd) { FLOCK(F_LOCK); } // on intel solaris man doesn't have doc on shared blocking -static int lock_exclusive_blocking(int fd) { FLOCK(F_LOCK); } -static int lock_exclusive_nonblocking(int fd) { FLOCK(F_TLOCK); } -static int unlock(int fd) { FLOCK(F_TLOCK); } - -#else - -#error unable to find file locking func - -#endif -#endif -#endif -#endif +# ifndef INADDR_NONE +# define INADDR_NONE ((ulong) -1) +# endif // defines for globals @@ -113,19 +77,18 @@ static int unlock(int fd) { FLOCK(F_TLOC const String file_status_name(FILE_STATUS_NAME); -// defines for statics +// defines #define HTTP_METHOD_NAME "method" +#define HTTP_FORM_NAME "form" +#define HTTP_BODY_NAME "body" #define HTTP_TIMEOUT_NAME "timeout" #define HTTP_HEADERS_NAME "headers" #define HTTP_ANY_STATUS_NAME "any-status" - -// statics - -static const String http_method_name(HTTP_METHOD_NAME); -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); +#define HTTP_CHARSET_NAME "charset" +#define HTTP_TABLES_NAME "tables" +#define HTTP_USER "user" +#define HTTP_PASSWORD "password" // defines @@ -133,16 +96,13 @@ static const String http_any_status_name // functions -void fix_line_breaks(char* buf, size_t& size) { - if(size==0) - return; - +void fix_line_breaks(char *str, size_t& length) { //_asm int 3; - const char* const eob=buf+size; - char* dest=buf; + const char* const eob=str+length; + char* dest=str; // fix DOS: \r\n -> \n // fix Macintosh: \r -> \n - char* bol=buf; + char* bol=str; while(char* eol=(char*)memchr(bol, '\r', eob -bol)) { size_t len=eol-bol; if(dest!=bol) @@ -152,60 +112,82 @@ void fix_line_breaks(char* buf, size_t& if(&eol[1]sin_family=AF_INET; addr->sin_port=htons(port); if(host) { - if(struct hostent *hostIP=gethostbyname(host)) - memcpy(&addr->sin_addr, hostIP->h_addr, hostIP->h_length); - else - return false; + ulong packed_ip=inet_addr(host); + if(packed_ip!=INADDR_NONE) + memcpy(&addr->sin_addr, &packed_ip, sizeof(packed_ip)); + else { + struct hostent *hostIP=gethostbyname(host); + if(hostIP) + memcpy(&addr->sin_addr, hostIP->h_addr, hostIP->h_length); + else + return false; + } } else addr->sin_addr.s_addr=INADDR_ANY; return true; } -static int http_read_responseString& response, int sock, bool fail_on_status_ne_200){ +static int http_read_response(char*& response, size_t& response_size, int sock, bool fail_on_status_ne_200){ + response=(char*)pa_malloc_atomic(1/*terminator*/); // setting memory block type + response[(response_size=0)]=0; int result=0; - ssize_t EOLat=0; + char* EOLat=0; while(true) { - char *buf=new(PointerFreeGC) char[MAX_STRING]; - ssize_t size=recv(sock, buf, MAX_STRING, 0); - if(size<=0) + char buf[MAX_STRING*10]; + ssize_t received_size=recv(sock, buf, sizeof(buf), 0); + if(received_size<=0) { + if(int no=pa_socks_errno()) + throw Exception("http.timeout", + 0, + "error receiving response: %s (%d)", pa_socks_strerr(no), no); break; - response.APPEND_TAINTED(buf, size, "remote HTTP server response", 0); - if(!result && (EOLat=response.pos("\r\n", 2))>=0) { // checking status in first response - const String& status_line=response.mid(0, (size_t)EOLat); + } + response=(char*)pa_realloc(response, response_size+received_size+1/*terminator*/); + memcpy(response+response_size, buf, received_size); + response_size+=received_size; + response[response_size]=0; + + if(!result && (EOLat=strstr(response, "\n"))) { // checking status in first response + const String status_line(pa_strdup(response, EOLat-response)); ArrayString astatus; - status_line->split(astatus, 0, " ", 1); - const String& status_code=astatus.get(1); - result=status_code->as_int(); + size_t pos_after=0; + status_line.split(astatus, pos_after, " "); + const String& status_code=*astatus.get(astatus.count()>1?1:0); + result=status_code.as_int(); if(fail_on_status_ne_200 && result!=200) throw Exception("http.status", - status_code, + &status_code, "invalid HTTP response status"); } } @@ -214,224 +196,492 @@ static int http_read_responseString& res else throw Exception("http.response", 0, - "bad response from host - no status found (size=%lu)", response.size()); + "bad response from host - no status found (size=%u)", response_size); } /* ********************** 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){ +static void timeout_handler(int /*sig*/){ siglongjmp(timeout_env, 1); } #endif -static int http_request( - String& response, - const String* origin_string, - const char* host, int port, - const char* request, - int timeout, - bool fail_on_status_ne_200) { +static int http_request(char*& response, size_t& response_size, + const char* host, short port, + const char* request, + int timeout_secs, + bool fail_on_status_ne_200) { if(!host) throw Exception("http.host", - origin_string, + 0, "zero hostname"); //never -#ifdef WE_CAN_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", - origin_string, - "timeout occured while retrieving document"); - else { - alarm(timeout); + volatile // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning] + int sock=-1; +#ifdef PA_USE_ALARM + signal(SIGALRM, timeout_handler); +#endif +#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 [athough duplicating closesocket code] + if(sock>=0) + closesocket(sock); + throw Exception("http.timeout", + 0, + "timeout occured while retrieving document"); + return 0; // never + } else { + alarm(timeout_secs); #endif + try { + int result; struct sockaddr_in dest; - if(!set_addr(&dest, host, port)) + if(!set_addr(&dest, host, port)) throw Exception("http.host", - origin_string, + 0, "can not resolve hostname \"%s\"", host); - if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) + if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) { + int no=pa_socks_errno(); throw Exception("http.connect", - origin_string, - "can not make socket: %s (%d)", strerror(errno), errno); - if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) + 0, + "can not make socket: %s (%d)", pa_socks_strerr(no), no); + } + + // To enable SO_DONTLINGER (that is, disable SO_LINGER) + // l_onoff should be set to zero and setsockopt should be called + linger dont_linger={0,0}; + setsockopt(sock, SOL_SOCKET, SO_LINGER, (const char *)&dont_linger, sizeof(dont_linger)); + +#ifdef WIN32 +// SO_*TIMEO can be defined in .h but not implemlemented in protocol, +// failing subsequently with Option not supported by protocol (99) message +// could not suppress that, so leaving this only for win32 + int timeout_ms=timeout_secs*1000; + setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms)); + setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms)); +#endif + + if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) { + int no=pa_socks_errno(); throw Exception("http.connect", - origin_string, - "can not connect to host \"%s\": %s (%d)", host, strerror(errno), errno); + 0, + "can not connect to host \"%s\": %s (%d)", host, pa_socks_strerr(no), no); + } size_t request_size=strlen(request); - if(send(sock, request, request_size, 0)!=(ssize_t)request_size) - throw Exception("http.connect", - origin_string, - "error sending request: %s (%d)", strerror(errno), errno); + if(send(sock, request, request_size, 0)!=(ssize_t)request_size) { + int no=pa_socks_errno(); + throw Exception("http.timeout", + 0, + "error sending request: %s (%d)", pa_socks_strerr(no), no); + } - result=http_read_response(response, sock, fail_on_status_ne_200); + result=http_read_response(response, response_size, 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, - Http_pass_header_info *info) { - *(info->request)<<*key<<": " - << *attributed_meaning_to_string(*info->value, String::UL_HTTP_HEADER, false) - <<"\n"; +static void http_pass_header(HashStringValue::key_type key, + HashStringValue::value_type value, + Http_pass_header_info *info) { + *info->request <change_case(*info->*info->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; +} + +static const String* basic_authorization_field(const char* user, const char* pass) { + if(!user&& !pass) + return 0; + + String combined; + if(user) + combined<key, *row->get(0), info->result); +} +static void form_value2string( + HashStringValue::key_type key, + HashStringValue::value_type value, + String* result) +{ + if(const String* svalue=value->get_string()) + form_string_value2string(key, *svalue, *result); + else if(Table* tvalue=value->get_table()) { + Form_table_value2string_info info(key, *result); + tvalue->for_each(form_table_value2string, &info); + } else + throw Exception(0, + new String(key, String::L_TAINTED), + "is %s, "HTTP_FORM_NAME" option value must either string or table", value->type()); +} +static const char* form2string(HashStringValue& form) { + String string; + form.for_each(form_value2string, &string); + return string.cstr(String::L_UNSPECIFIED); } #ifndef DOXYGEN struct File_read_http_result { - char *data; size_t size; + char *str; size_t length; HashStringValue* headers; }; #endif -static File_read_http_result file_read_httpCharset& charset, const String& file_spec, - HashStringValue *options=0) { +static void find_headers_end(char* p, + char*& headers_end_at, + char*& raw_body) +{ + raw_body=p; + // \n\n + // \r\n\r\n + while((p=strchr(p, '\n'))) { + headers_end_at=++p; // \n>.< + if(*p=='\r') // \r\n>\r?<\n + p++; + if(*p=='\n') { // \r\n\r>\n?< + raw_body=p+1; + return; + } + } + headers_end_at=0; +} + +/// @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, + bool as_text, + HashStringValue *options=0) { File_read_http_result result; char host[MAX_STRING]; const char* uri; - int port; - const char* method(0); - int timeout=2; + short port; + const char* method="GET"; bool method_is_get; + HashStringValue* form=0; + const char* body_cstr=0; + int timeout_secs=2; bool fail_on_status_ne_200=true; - ValuePtr vheaders(0); - - const String& connect_string(new String()); - // not in ^sql{... UL_SQL ...} spirit, but closer to ^file::load one - connect_string->append(*file_spec, String::UL_URI); // tainted pieces -> URI pieces - - const char* connect_string_cstr=connect_string->cstr(String::UL_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; + Value* vheaders=0; + Charset *asked_remote_charset=0; + const char* user_cstr=0; + const char* password_cstr=0; if(options) { int valid_options=0; - if(ValuePtr vmethod=options->get(http_method_name)) { + if(Value* vmethod=options->get(HTTP_METHOD_NAME)) { valid_options++; - method=vmethod->as_string(&pool)->cstr(); + method=vmethod->as_string().cstr(); } - if(ValuePtr vtimeout=options->get(http_timeout_name)) { + if(Value* vform=options->get(HTTP_FORM_NAME)) { valid_options++; - timeout=vtimeout->as_int(); - } - if(vheaders=options->get(http_headers_name)) { + form=vform->get_hash(); + } + if(Value* vbody=options->get(HTTP_BODY_NAME)) { valid_options++; - } - if(ValuePtr vany_status=options->get(http_any_status_name)) { + body_cstr=vbody->as_string().cstr(String::L_UNSPECIFIED); + } + if(Value* vtimeout=options->get(HTTP_TIMEOUT_NAME)) { + valid_options++; + timeout_secs=vtimeout->as_int(); + } + if((vheaders=options->get(HTTP_HEADERS_NAME))) { + valid_options++; + } + if(Value* vany_status=options->get(HTTP_ANY_STATUS_NAME)) { 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(Value* vuser=options->get(HTTP_USER)) { + valid_options++; + user_cstr=vuser->as_string().cstr(); + } + if(Value* vpassword=options->get(HTTP_PASSWORD)) { + valid_options++; + password_cstr=vpassword->as_string().cstr(); } 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(connect_string)) { - 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"; + method_is_get=strcmp(method, "GET")==0; + if(method_is_get && body_cstr) + throw Exception("parser.runtime", + 0, + "you can not use $."HTTP_BODY_NAME" option with method GET"); + + //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 + + String request_head_and_body; + { + // 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?(short)strtol(port_cstr, &error_pos, 0):80; + + if(strchr(uri, '?') && form) + throw Exception("parser.runtime", + 0, + "use either uri with ?params or $."HTTP_FORM_NAME" option"); + + //making request head + String head; + head << method; + head << " " << uri; + if(form) + if(method_is_get) + head << "?" << form2string(*form); + head <<" HTTP/1.0" CRLF + "host: "<< host << CRLF; + if(form && !method_is_get) { + head << "content-type: application/x-www-form-urlencoded" CRLF; + body_cstr = form2string(*form); + } + + // http://www.ietf.org/rfc/rfc2617.txt + if(const String* authorization_field_value=basic_authorization_field(user_cstr, password_cstr)) + head<<"authorization: "<<*authorization_field_value<is_string()) { // allow empty + if(HashStringValue *headers=vheaders->get_hash()) { + Http_pass_header_info info={&charsets, &head, false}; + 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 + head << "user-agent: " DEFAULT_USER_AGENT CRLF; + + if(body_cstr) { + // recode those pieces which are not in String::L_URI lang + // [those violating HTTP standard, but widly used] + body_cstr=Charset::transcode( + String::C(body_cstr, strlen(body_cstr)), + charsets.source(), + *asked_remote_charset); + + head << "content-length: " << format(strlen(body_cstr), "%u") << CRLF; + } + + const char* head_cstr=head.cstr(String::L_UNSPECIFIED); + + // recode those pieces which are not in String::L_URI lang + // [those violating HTTP standard, but widly used] + head_cstr=Charset::transcode( + String::C(head_cstr, strlen(head_cstr)), + charsets.source(), + *asked_remote_charset); + + // head + end of header + request_head_and_body << head_cstr << CRLF; + // body + if(body_cstr) + request_head_and_body << body_cstr; + } //sending request - String response; - const char* request_cstr=request.cstr(String::UL_UNSPECIFIED); - int status_code=http_request(response, - connect_string, host, port, request_cstr, - timeout, fail_on_status_ne_200); + char* response; + size_t response_size; + int status_code=http_request(response, response_size, + host, port, request_head_and_body.cstr(), + timeout_secs, fail_on_status_ne_200); //processing results - int pos=response.pos("\r\n\r\n", 4); - if(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.size()); - - ArrayString aheaders; - result.headers=HashStringValue* (new HashStringValue); - header_block->split(aheaders, 0, "\r\n", 2); - - //processing headers - for(int i=1; ipos(": ", 2); - if(pos<1) - throw Exception("http.response", - connect_string, - "bad response from host - bad header \"%s\"", line->cstr().get()); - - result.headers->put( - line->mid(0, pos)->change_case(charset, String::CC_UPPER), - ValuePtr(new VString(line->mid(pos+2, line->size())))); + char* raw_body; size_t raw_body_size; + char* headers_end_at; + find_headers_end(response, + headers_end_at, + raw_body); + raw_body_size=response_size-(raw_body-response); + + result.headers=new HashStringValue; + VHash* vtables=new VHash; + result.headers->put(HTTP_TABLES_NAME, vtables); + Charset* real_remote_charset=0; // undetected, yet + + if(headers_end_at) { + *headers_end_at=0; + const String header_block(String::C(response, headers_end_at-response), true); + + ArrayString aheaders; + HashStringValue& tables=vtables->hash(); + + size_t pos_after=0; + header_block.split(aheaders, pos_after, "\n"); + + //processing headers + size_t aheaders_count=aheaders.count(); + for(size_t i=1; iget_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)); + } } // output response - result.data=new(PointerFreeGC) char[1/*as text puts terminating zero later*/+(result.size=body->size())]; - body->store_to(result.data, String::UL_AS_IS); - result.headers->put(file_status_name, ValuePtr(new VInt(status_code))); + String::C real_body=String::C(raw_body, raw_body_size); + if(as_text && raw_body_size) { // must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below + // defaulting to used-asked charset [it's never empty!] + if(!real_remote_charset) + real_remote_charset=asked_remote_charset; + real_body=Charset::transcode(real_body, *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; } @@ -440,22 +690,31 @@ static File_read_http_result file_read_h #ifndef DOXYGEN struct File_read_action_info { char **data; size_t *data_size; + char* buf; size_t offset; size_t count; }; #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)]; + size_t to_read_size=info.count; + if(!to_read_size) + to_read_size=(size_t)finfo.st_size; + assert( !(info.buf && as_text) ); + if(to_read_size) { + if(info.offset) + lseek(f, info.offset, SEEK_SET); + *info.data=info.buf + ? info.buf + : new(PointerFreeGC) char[to_read_size+(as_text?1:0)]; *info.data_size=(size_t)read(f, *info.data, to_read_size); if(ssize_t(*info.data_size)<0 || *info.data_size>to_read_size) throw Exception(0, &file_spec, - "read failed: actually read %lu bytes count not in [0..%lu] valid range", + "read failed: actually read %u bytes count not in [0..%u] valid range", *info.data_size, to_read_size); } else { // empty file if(as_text) { @@ -467,21 +726,28 @@ 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; + bool fail_on_read_problem, + char* buf, size_t offset, size_t count) { + File_read_result result={false, 0, 0, 0}; #ifdef PA_HTTP - if(file_spec.starts_with("http://", 7)) { + 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, as_text, params); result.success=true; - result.data=http.data; - result.size=http.size; + result.str=http.str; + result.length=http.length; result.headers=http.headers; } else { #endif - File_read_action_info info={&result.data, &result.size}; + if(params && params->count()) + throw Exception("parser.runtime", + 0, + "invalid option passed"); + + File_read_action_info info={&result.str, &result.length, + buf, offset, count}; result.success=file_read_action_under_lock(file_spec, "read", file_read_action, &info, as_text, fail_on_read_problem); @@ -491,27 +757,41 @@ File_read_result file_read(Charset& char if(result.success && as_text) { // UTF-8 signature: EF BB BF - if(result.size>=3) { - char *in=(char *)result.data; - if((in[0] == '\xEF') && (in[1] == '\xBB') && - (in[2] == '\xBF')) { - result.data=in+3; result.size-=3;// skip prefix + if(result.length>=3) { + char *in=(char *)result.str; + if(strncmp(in, "\xEF\xBB\xBF", 3)==0) { + result.str=in+3; result.length-=3;// skip prefix } } - fix_line_breaks((char *)(result.data), result.size); - // note: after fixing - ((char*&)(result.data))[result.size]=0; + fix_line_breaks((char *)(result.str), result.length); } return result; } +#ifdef PA_SAFE_MODE +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", + &file_spec, + "parser is in safe mode: " + "reading files of foreign group and user disabled " + "[recompile parser with --disable-safe-mode configure option], " + "actual filename '%s', " + "fuid(%d)!=euid(%d) or fgid(%d)!=egid(%d)", + fname, + finfo.st_uid, geteuid(), + finfo.st_gid, getegid()); +} +#endif + bool file_read_action_under_lock(const String& file_spec, const char* action_name, File_read_action action, void *context, bool as_text, bool fail_on_read_problem) { - const char* fname=file_spec.cstr(String::UL_FILE_SPEC); + const char* fname=file_spec.cstr(String::L_FILE_SPEC); int f; // first open, next stat: @@ -525,7 +805,7 @@ bool file_read_action_under_lock(const S // later: it seems, even this does not help sometimes if((f=open(fname, O_RDONLY|(as_text?_O_TEXT:_O_BINARY)))>=0) { try { - if(lock_shared_blocking(f)!=0) + if(pa_lock_shared_blocking(f)!=0) throw Exception("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", @@ -539,23 +819,18 @@ bool file_read_action_under_lock(const S strerror(errno), errno, fname); #ifdef PA_SAFE_MODE - if(finfo.st_uid/*foreign?*/!=geteuid() - && finfo.st_gid/*foreign?*/!=getegid()) - throw Exception("parser.runtime", - &file_spec, - "parser is in safe mode: reading files of foreign group and user disabled [recompile parser with --disable-safe-mode configure option], actual filename '%s'", - fname); + check_safe_mode(finfo, file_spec, fname); #endif action(finfo, f, file_spec, fname, as_text, context); } catch(...) { - unlock(f);close(f); + pa_unlock(f);close(f); if(fail_on_read_problem) rethrow; return false; } - unlock(f);close(f); + pa_unlock(f);close(f); return true; } else { if(fail_on_read_problem) @@ -569,9 +844,9 @@ bool file_read_action_under_lock(const S static void create_dir_for_file(const String& file_spec) { size_t pos_after=1; - int pos_before; - while((pos_before=file_spec.pos("/", pos_after))>=0) { - mkdir(file_spec.mid(0, pos_before).cstr(String::UL_FILE_SPEC), 0775); + size_t pos_before; + while((pos_before=file_spec.pos('/', pos_after))!=STRING_NOT_FOUND) { + mkdir(file_spec.mid(0, pos_before).cstr(String::L_FILE_SPEC), 0775); pos_after=pos_before+1; } } @@ -583,7 +858,7 @@ bool file_write_action_under_lock( bool do_append, bool do_block, bool fail_on_lock_problem) { - const char* fname=file_spec.cstr(String::UL_FILE_SPEC); + const char* fname=file_spec.cstr(String::L_FILE_SPEC); int f; if(access(fname, W_OK)!=0) // no create_dir_for_file(file_spec); @@ -592,7 +867,7 @@ bool file_write_action_under_lock( O_CREAT|O_RDWR |(as_text?_O_TEXT:_O_BINARY) |(do_append?O_APPEND:PA_O_TRUNC), 0664))>=0) { - if((do_block?lock_exclusive_blocking(f):lock_exclusive_nonblocking(f))!=0) { + if((do_block?pa_lock_exclusive_blocking(f):pa_lock_exclusive_nonblocking(f))!=0) { Exception e("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", @@ -610,7 +885,7 @@ bool file_write_action_under_lock( if(!do_append) ftruncate(f, lseek(f, 0, SEEK_CUR)); // one can not use O_TRUNC, read lower #endif - unlock(f);close(f); + pa_unlock(f);close(f); rethrow; } @@ -618,7 +893,7 @@ bool file_write_action_under_lock( if(!do_append) ftruncate(f, lseek(f, 0, SEEK_CUR)); // O_TRUNC truncates even exclusevely write-locked file [thanks to Igor Milyakov for discovering] #endif - unlock(f);close(f); + pa_unlock(f);close(f); return true; } else throw Exception(errno==EACCES?"file.access":0, @@ -630,13 +905,13 @@ bool file_write_action_under_lock( #ifndef DOXYGEN struct File_write_action_info { - const char* data; size_t size; + const char* str; size_t length; }; #endif static void file_write_action(int f, void *context) { File_write_action_info& info=*static_cast(context); - if(info.size) { - int written=write(f, info.data, info.size); + if(info.length) { + int written=write(f, info.str, info.length); if(written<0) throw Exception(0, 0, @@ -658,16 +933,16 @@ void file_write( // throws nothing! [this is required in file_move & file_delete] static void rmdir(const String& file_spec, size_t pos_after) { - int pos_before; - if((pos_before=file_spec.pos("/", pos_after))>=0) + size_t pos_before; + if((pos_before=file_spec.pos('/', pos_after))!=STRING_NOT_FOUND) rmdir(file_spec, pos_before+1); - rmdir(file_spec.mid(0, pos_after-1/* / */).cstr(String::UL_FILE_SPEC)); + rmdir(file_spec.mid(0, pos_after-1/* / */).cstr(String::L_FILE_SPEC)); } -bool file_delete(const String& file_spec, bool fail_on_read_problem) { - const char* fname=file_spec.cstr(String::UL_FILE_SPEC); +bool file_delete(const String& file_spec, bool fail_on_problem) { + const char* fname=file_spec.cstr(String::L_FILE_SPEC); if(unlink(fname)!=0) - if(fail_on_read_problem) + if(fail_on_problem) throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, &file_spec, "unlink failed: %s (%d), actual filename '%s'", @@ -679,8 +954,8 @@ bool file_delete(const String& file_spec return true; } void file_move(const String& old_spec, const String& new_spec) { - const char* old_spec_cstr=old_spec.cstr(String::UL_FILE_SPEC); - const char* new_spec_cstr=new_spec.cstr(String::UL_FILE_SPEC); + const char* old_spec_cstr=old_spec.cstr(String::L_FILE_SPEC); + const char* new_spec_cstr=new_spec.cstr(String::L_FILE_SPEC); create_dir_for_file(new_spec); @@ -703,12 +978,12 @@ bool entry_exists(const char* fname, str } bool entry_exists(const String& file_spec) { - const char* fname=file_spec.cstr(String::UL_FILE_SPEC); + const char* fname=file_spec.cstr(String::L_FILE_SPEC); return entry_exists(fname, 0); } static bool entry_readable(const String& file_spec, bool need_dir) { - char* fname=file_spec.cstrm(String::UL_FILE_SPEC); + char* fname=file_spec.cstrm(String::L_FILE_SPEC); if(need_dir) { size_t size=strlen(fname); while(size) { @@ -732,14 +1007,14 @@ bool file_readable(const String& file_sp bool dir_readable(const String& file_spec) { return entry_readable(file_spec, true); } -const String* file_readable(String& path, String& name) { +const String* file_readable(const String& path, const String& name) { String& result=*new String(path); result << "/"; result << name; return file_readable(result)?&result:0; } bool file_executable(const String& file_spec) { - return access(file_spec.cstr(String::UL_FILE_SPEC), X_OK)==0; + return access(file_spec.cstr(String::L_FILE_SPEC), X_OK)==0; } bool file_stat(const String& file_spec, @@ -748,8 +1023,8 @@ bool file_stat(const String& file_spec, time_t& rmtime, time_t& rctime, bool fail_on_read_problem) { - const char* fname=file_spec.cstr(String::UL_FILE_SPEC); - struct stat finfo; + const char* fname=file_spec.cstr(String::L_FILE_SPEC); + struct stat finfo; if(stat(fname, &finfo)!=0) if(fail_on_read_problem) throw Exception("file.missing", @@ -827,6 +1102,7 @@ const char* format(double value, char* f size_t stdout_write(const void *buf, size_t size) { #ifdef WIN32 + size_t to_write = size; do{ int chunk_written=fwrite(buf, 1, min((size_t)8*0x400, size), stdout); if(chunk_written<=0) @@ -835,7 +1111,7 @@ size_t stdout_write(const void *buf, siz buf=((const char*)buf)+chunk_written; } while(size>0); - return size; + return to_write-size; #else return fwrite(buf, 1, size, stdout); #endif @@ -848,11 +1124,11 @@ char* unescape_chars(const char* cp, int EscapeFirst, EscapeSecond } escapeState=EscapeRest; - int escapedValue=0; + uchar escapedValue=0; int srcPos=0; int dstPos=0; while(srcPos < len) { - int ch=cp[srcPos]; + uchar ch=(uchar)cp[srcPos]; switch(escapeState) { case EscapeRest: if(ch=='%') { @@ -864,7 +1140,7 @@ char* unescape_chars(const char* cp, int } break; case EscapeFirst: - escapedValue=hex_value[ch] << 4; + escapedValue=(uchar)(hex_value[ch] << 4); escapeState=EscapeSecond; break; case EscapeSecond: @@ -905,8 +1181,8 @@ bool StrEqNc(const char* s1, const char* return !strict; } else if(!(*s2)) return !strict; - if(isalpha(*s1)) { - if(tolower(*s1) !=tolower(*s2)) + if(isalpha((unsigned char)*s1)) { + if(tolower((unsigned char)*s1) !=tolower((unsigned char)*s2)) return false; } else if((*s1) !=(*s2)) return false; @@ -958,6 +1234,11 @@ int __vsnprintf(char* b, size_t s, const // note: on win32& maybe somewhere else // vsnprintf do not writes terminating 0 in 'buffer full' case, reducing --s; + + // clients do not check for negative 's', feature: ignore such prints + if((ssize_t)s<0) + return 0; + #if _MSC_VER /* win32: @@ -983,7 +1264,7 @@ int __vsnprintf(char* b, size_t s, const if(r<0) r=0; - else if(r>s) + else if((size_t)r>s) r=s; #endif b[r]=0; @@ -998,18 +1279,178 @@ int __snprintf(char* b, size_t s, const return r; } -int pa_sleep(unsigned long secs, unsigned long usecs) { - for (; usecs >= 1000000; ++secs, usecs -= 1000000); +/* mime64 functions are from libgmime[http://spruce.sourceforge.net/gmime/] lib */ +/* + * Authors: Michael Zucchi + * Jeffrey Stedfast + * + * Copyright 2000 Helix Code, Inc. (www.helixcode.com) + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Street #330, Boston, MA 02111-1307, USA. + * + */ +static char *base64_alphabet = + "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; + +/** + * g_mime_utils_base64_encode_step: + * @in: input stream + * @inlen: length of the input + * @out: output string + * @state: holds the number of bits that are stored in @save + * @save: leftover bits that have not yet been encoded + * + * Base64 encodes a chunk of data. Performs an 'encode step', only + * encodes blocks of 3 characters to the output at a time, saves + * left-over state in state and save (initialise to 0 on first + * invocation). + * + * Returns the number of bytes encoded. + **/ +static size_t +g_mime_utils_base64_encode_step (const unsigned char *in, size_t inlen, unsigned char *out, int *state, int *save) +{ + register const unsigned char *inptr; + register unsigned char *outptr; + + if (inlen <= 0) + return 0; + + inptr = in; + outptr = out; + + if (inlen + ((unsigned char *)save)[0] > 2) { + const unsigned char *inend = in + inlen - 2; + register int c1 = 0, c2 = 0, c3 = 0; + register int already; + + already = *state; + + switch (((char *)save)[0]) { + case 1: c1 = ((unsigned char *)save)[1]; goto skip1; + case 2: c1 = ((unsigned char *)save)[1]; + c2 = ((unsigned char *)save)[2]; goto skip2; + } + + /* yes, we jump into the loop, no i'm not going to change it, its beautiful! */ + while (inptr < inend) { + c1 = *inptr++; + skip1: + c2 = *inptr++; + skip2: + c3 = *inptr++; + *outptr++ = base64_alphabet [c1 >> 2]; + *outptr++ = base64_alphabet [(c2 >> 4) | ((c1 & 0x3) << 4)]; + *outptr++ = base64_alphabet [((c2 & 0x0f) << 2) | (c3 >> 6)]; + *outptr++ = base64_alphabet [c3 & 0x3f]; + /* this is a bit ugly ... */ + if ((++already) >= 19) { + *outptr++ = '\n'; + already = 0; + } + } + + ((unsigned char *)save)[0] = 0; + inlen = 2 - (inptr - inend); + *state = already; + } + + //d(printf ("state = %d, inlen = %d\n", (int)((char *)save)[0], inlen)); + + if (inlen > 0) { + register char *saveout; + + /* points to the slot for the next char to save */ + saveout = & (((char *)save)[1]) + ((char *)save)[0]; + + /* inlen can only be 0 1 or 2 */ + switch (inlen) { + case 2: *saveout++ = *inptr++; + case 1: *saveout++ = *inptr++; + } + ((char *)save)[0] += inlen; + } + + /*d(printf ("mode = %d\nc1 = %c\nc2 = %c\n", + (int)((char *)save)[0], + (int)((char *)save)[1], + (int)((char *)save)[2]));*/ + + return (outptr - out); +} -#ifdef WIN32 - Sleep(secs * 1000 + usecs / 1000); - return 0; -#else - struct timeval t; - t.tv_sec = secs; - t.tv_usec = usecs; - return (select(0, NULL, NULL, NULL, &t) == -1 ? errno : 0); -#endif +/** + * g_mime_utils_base64_encode_close: + * @in: input stream + * @inlen: length of the input + * @out: output string + * @state: holds the number of bits that are stored in @save + * @save: leftover bits that have not yet been encoded + * + * Base64 encodes the input stream to the output stream. Call this + * when finished encoding data with g_mime_utils_base64_encode_step to + * flush off the last little bit. + * + * Returns the number of bytes encoded. + **/ +static size_t +g_mime_utils_base64_encode_close (const unsigned char *in, size_t inlen, unsigned char *out, int *state, int *save) +{ + unsigned char *outptr = out; + int c1, c2; + + if (inlen > 0) + outptr += g_mime_utils_base64_encode_step (in, inlen, outptr, state, save); + + c1 = ((unsigned char *)save)[1]; + c2 = ((unsigned char *)save)[2]; + + switch (((unsigned char *)save)[0]) { + case 2: + outptr[2] = base64_alphabet [(c2 & 0x0f) << 2]; + goto skip; + case 1: + outptr[2] = '='; + skip: + outptr[0] = base64_alphabet [c1 >> 2]; + outptr[1] = base64_alphabet [c2 >> 4 | ((c1 & 0x3) << 4)]; + outptr[3] = '='; + outptr += 4; + break; + } + + *outptr++ = 0; + + *save = 0; + *state = 0; + + return (outptr - out); } +char* pa_base64(const char *in, size_t len) +{ + /* wont go to more than 2x size (overly conservative) */ + char* result=new(PointerFreeGC) char[len * 2 + 6]; + int state=0; + int save=0; +#ifndef NDEBUG + size_t filled= +#endif + g_mime_utils_base64_encode_close ((const unsigned char*)in, len, + (unsigned char*)result, &state, &save); + assert(filled <= len * 2 + 6); + return result; +}