--- parser3/src/main/pa_common.C 2003/03/24 17:00:49 1.143.2.21.2.12 +++ 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: 2003/03/24 17:00:49 $"; +static const char* IDENT_COMMON_C="$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 +#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 @@ -61,6 +65,9 @@ int PASCAL closesocket(SOCKET); #endif // locking constants +//#define PA_DEBUG_NO_LOCKING + +#ifdef PA_DEBUG_NO_LOCKING #ifdef HAVE_FLOCK @@ -105,6 +112,14 @@ static int unlock(int fd) { FLOCK(F_TLOC #endif #endif +#else +static int lock_shared_blocking(int fd) { return 0; } +static int lock_exclusive_blocking(int fd) { return 0; } +static int lock_exclusive_nonblocking(int fd) { return 0; } +static int unlock(int fd) { return 0; } + +#endif + // defines for globals #define FILE_STATUS_NAME "status" @@ -119,6 +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 @@ -126,6 +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 @@ -133,16 +152,13 @@ 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) @@ -152,27 +168,31 @@ void fix_line_breaks(char* buf, size_t& if(&eol[1]sin_family=AF_INET; @@ -187,7 +207,7 @@ static bool set_addr(struct sockaddr_in return true; } -static int http_read_responseString& response, int sock, bool fail_on_status_ne_200){ +static int http_read_response(String& response, int sock, bool fail_on_status_ne_200){ int result=0; size_t EOLat=0; while(true) { @@ -195,17 +215,18 @@ static int http_read_responseString& res ssize_t size=recv(sock, buf, MAX_STRING, 0); if(size<=0) break; - response.APPEND_TAINTED(buf, size, "remote HTTP server response", 0); - if(!result && (EOLat=response.pos("\r\n", 2))!=STRING_NOT_FOUND) { // checking status in first response + response.append_strdup(buf, size, String::L_TAINTED); + 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; - 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(1); + 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"); } } @@ -220,132 +241,157 @@ static int http_read_responseString& res /* ********************** 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, - const String* origin_string, - const char* host, int port, - const char* request, - int timeout, - bool fail_on_status_ne_200) { +static int http_request(String& response, + const char* host, int port, + const char* request, + int timeout, + 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); +#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", - origin_string, - "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", - origin_string, + 0, "can not resolve hostname \"%s\"", host); if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) throw Exception("http.connect", - origin_string, + 0, "can not make socket: %s (%d)", strerror(errno), errno); if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) throw Exception("http.connect", - origin_string, + 0, "can not connect to host \"%s\": %s (%d)", host, strerror(errno), errno); size_t request_size=strlen(request); if(send(sock, request, request_size, 0)!=(ssize_t)request_size) throw Exception("http.connect", - origin_string, + 0, "error sending request: %s (%d)", strerror(errno), errno); 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, - Http_pass_header_info *info) { - *(info->request)<<*key<<": " - << *attributed_meaning_to_string(*info->value, String::L_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; +} + + #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) { +/// @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]; const char* uri; int port; - const char* method(0); + const char* method=0; int timeout=2; bool fail_on_status_ne_200=true; - Value* vheaders(0); - - const 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; + Value* vheaders=0; + Charset *asked_remote_charset=0; if(options) { int valid_options=0; @@ -364,74 +410,158 @@ 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(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"; + //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, - connect_string, host, port, request_cstr, + 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, + &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=HashStringValue* (new HashStringValue); - header_block->split(aheaders, 0, "\r\n", 2); + 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, CRLF); //processing headers - for(int i=1; ipos(": ", 2); + size_t aheaders_count=aheaders.count(); + Charset* real_remote_charset=0; // undetected, yet + for(size_t i=1; icstr().get()); - - result.headers->put( - line->mid(0, pos)->change_case(charset, String::CC_UPPER), - Value*(new VString(line->mid(pos+2, line->length())))); + &connect_string, + "bad response from host - bad header \"%s\"", line.cstr()); + 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=new(PointerFreeGC) char[1/*as text puts terminating zero later*/+(result.length=body->length())]; - body->store_to(result.str, String::L_AS_IS); - result.headers->put(file_status_name, Value*(new VInt(status_code))); + 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; } @@ -443,10 +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)]; @@ -467,17 +597,17 @@ 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://", 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, params); result.success=true; - result.str=http.data; - result.length=http.size; + result.str=http.str; + result.length=http.length; result.headers=http.headers; } else { #endif @@ -493,8 +623,7 @@ File_read_result file_read(Charset& char // UTF-8 signature: EF BB BF if(result.length>=3) { char *in=(char *)result.str; - if((in[0] == '\xEF') && (in[1] == '\xBB') && - (in[2] == '\xBF')) { + if(strncmp(in, "\xEB\xBB\xBF", 3)==0) { result.str=in+3; result.length-=3;// skip prefix } } @@ -507,6 +636,23 @@ File_read_result file_read(Charset& char 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, @@ -539,12 +685,7 @@ 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); @@ -570,7 +711,7 @@ bool file_read_action_under_lock(const S static void create_dir_for_file(const String& file_spec) { size_t pos_after=1; size_t pos_before; - while((pos_before=file_spec.pos("/", pos_after))!=STRING_NOT_FOUND) { + 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; } @@ -659,7 +800,7 @@ void file_write( // throws nothing! [this is required in file_move & file_delete] static void rmdir(const String& file_spec, size_t pos_after) { size_t pos_before; - if((pos_before=file_spec.pos("/", pos_after))!=STRING_NOT_FOUND) + 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::L_FILE_SPEC)); @@ -749,7 +890,7 @@ bool file_stat(const String& file_spec, time_t& rctime, bool fail_on_read_problem) { const char* fname=file_spec.cstr(String::L_FILE_SPEC); - struct stat finfo; + struct stat finfo; if(stat(fname, &finfo)!=0) if(fail_on_read_problem) throw Exception("file.missing",