--- parser3/src/main/pa_common.C 2005/11/22 11:12:40 1.210 +++ parser3/src/main/pa_common.C 2016/12/29 15:55:40 1.299 @@ -1,7 +1,7 @@ /** @file Parser: commonly functions. - Copyright(c) 2001-2005 ArtLebedev Group (http://www.artlebedev.com) + Copyright (c) 2000-2015 Art. Lebedev Studio (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) * BASE64 part @@ -26,47 +26,32 @@ * */ -static const char * const IDENT_COMMON_C="$Date: 2005/11/22 11:12:40 $"; - #include "pa_common.h" #include "pa_exception.h" #include "pa_hash.h" #include "pa_globals.h" -#include "pa_request_charsets.h" #include "pa_charsets.h" +#include "pa_http.h" +#include "pa_request_charsets.h" +#include "pcre.h" +#include "pa_request.h" -#define PA_HTTP +#include "pa_idna.h" +#include "pa_convert_utf.h" -#ifdef PA_HTTP -#include "pa_vstring.h" -#include "pa_vint.h" -#include "pa_vhash.h" -#include "pa_vtable.h" -#include "pa_socks.h" - -#ifdef CYGWIN -#define _GNU_H_WINDOWS32_SOCKETS -// for PASCAL +#ifdef _MSC_VER #include -// SOCKET -typedef u_int SOCKET; -int PASCAL closesocket(SOCKET); -#else -# if defined(WIN32) -# include -# else -# define closesocket close -# endif +#include #endif +#ifdef _MSC_VER +#define pa_mkdir(path, mode) _mkdir(path) #else - -# if defined(WIN32) -# include -# endif - +#define pa_mkdir(path, mode) mkdir(path, mode) #endif +volatile const char * IDENT_PA_COMMON_C="$Id: pa_common.C,v 1.299 2016/12/29 15:55:40 moko Exp $" IDENT_PA_COMMON_H IDENT_PA_HASH_H IDENT_PA_ARRAY_H IDENT_PA_STACK_H; + // some maybe-undefined constants #ifndef _O_TEXT @@ -86,10 +71,6 @@ int PASCAL closesocket(SOCKET); # endif #endif -# ifndef INADDR_NONE -# define INADDR_NONE ((ulong) -1) -# endif - // defines for globals #define FILE_STATUS_NAME "status" @@ -98,63 +79,20 @@ int PASCAL closesocket(SOCKET); const String file_status_name(FILE_STATUS_NAME); -// 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" -#define HTTP_CHARSET_NAME "charset" -#define HTTP_TABLES_NAME "tables" -#define HTTP_USER "user" -#define HTTP_PASSWORD "password" - -// defines - -#define DEFAULT_USER_AGENT "parser3" - // functions -void fix_line_breaks(char *str, size_t& length) { - //_asm int 3; - const char* const eob=str+length; - char* dest=str; - // fix DOS: \r\n -> \n - // fix Macintosh: \r -> \n - char* bol=str; - while(char* eol=(char*)memchr(bol, '\r', eob -bol)) { - size_t len=eol-bol; - if(dest!=bol) - memcpy(dest, bol, len); - dest+=len; - *dest++='\n'; - - if(&eol[1]sin_family=AF_INET; - addr->sin_port=htons(port); - if(host) { - 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; -} - -size_t guess_content_length(char* buf) { - char* ptr; - if((ptr=strstr(buf, "Content-Length:"))) // Apache - goto found; - if((ptr=strstr(buf, "content-length:"))) // Parser 3 - goto found; - if((ptr=strstr(buf, "Content-length:"))) // maybe 1 - goto found; - if((ptr=strstr(buf, "CONTENT-LENGTH:"))) // maybe 2 - goto found; - return 0; -found: - char *error_pos; - size_t result=(size_t)strtol(ptr+15/*strlen("CONTENT-LENGTH:")*/, &error_pos, 0); - - const size_t reasonable_initial_max=0x400*0x400*10 /*10M*/; - if(result>reasonable_initial_max) // sanity check - return reasonable_initial_max; - return 0;//result; -} - -static int http_read_response(char*& response, size_t& response_size, int sock, bool fail_on_status_ne_200) { - int result=0; - // fetching some to local buffer, guessing on possible content-length - response_size=0x400*20; // initial size if content-length could not be determined - const size_t preview_size=0x400*20; - char preview_buf[preview_size+1/*terminator*/]; // 20K buffer to preview headers - ssize_t received_size=recv(sock, preview_buf, preview_size, 0); - if(received_size==0) - goto done; - if(received_size<0) { - if(int no=pa_socks_errno()) - throw Exception("http.timeout", - 0, - "error receiving response header: %s (%d)", pa_socks_strerr(no), no); - goto done; - } - // detecting response_size - { - preview_buf[received_size]=0; // terminator - if(size_t content_length=guess_content_length(preview_buf)) - response_size=preview_size+content_length; // a little more than needed, will adjust response_size by actual received size later - } - - // [gcc is happier this way, see goto above] - { - // allocating initial buf - response=(char*)pa_malloc_atomic(response_size+1/*terminator*/); // just setting memory block type - char* ptr=response; - size_t todo_size=response_size; - // coping part of already received body - memcpy(ptr, preview_buf, received_size); - ptr+=received_size; - todo_size-=received_size; - - // we use terminator byte for two purposes here: - // 1. we return there zero always, not knowing: maybe they would want to create String form $file.body? - // invariant: all Strings should have zero-terminated buffers - // 2. we use that out-of-size byte to detect if our content-length guess was wrong - // when recv gets more than we expected - // a) we know that the content-length guess was wrong - // b) we have space to put the first byte of extra data - // c) we use less code to detect normal situation: on last while-cycle recv expected to just return 0 - while(true) { - received_size=recv(sock, ptr, todo_size+1/*there is always a place for terminator*/, 0); - if(received_size==0) { - response_size-=todo_size; // in case we received less than expected, cut down the reported size - break; - } - if(received_size<0) { - if(int no=pa_socks_errno()) - throw Exception("http.timeout", - 0, - "error receiving response body: %s (%d)", pa_socks_strerr(no), no); - break; - } - // they've touched the terminator? - if((size_t)received_size>todo_size) - { - // that means that our guessed response_size was not big enough - const size_t grow_chunk_size=0x400*0x400; // 1M - response_size+=grow_chunk_size; - size_t ptr_offset=ptr-response; - response=(char*)pa_realloc(response, response_size+1/*terminator*/); - ptr=response+ptr_offset; - todo_size+=grow_chunk_size; - } - // can't do this before realloc: we need 1?1:0); - result=status_code.as_int(); - - if(fail_on_status_ne_200 && result!=200) - throw Exception("http.status", - &status_code, - "invalid HTTP response status"); - } - } - } -done: - if(result) - { - response[response_size]=0; - return result; - } - else - throw Exception("http.response", - 0, - "bad response from host - no status found (size=%u)", response_size); -} - -/* ********************** request *************************** */ - -#if defined(SIGALRM) && defined(HAVE_SIGSETJMP) && defined(HAVE_SIGLONGJMP) -# define PA_USE_ALARM -#endif - -#ifdef PA_USE_ALARM -static sigjmp_buf timeout_env; -static void timeout_handler(int /*sig*/){ - siglongjmp(timeout_env, 1); -} -#endif - -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", - 0, - "zero hostname"); //never - - 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)) - throw Exception("http.host", - 0, - "can not resolve hostname \"%s\"", host); - - if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) { - int no=pa_socks_errno(); - throw Exception("http.connect", - 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", - 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) { - 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, response_size, sock, fail_on_status_ne_200); - closesocket(sock); -#ifdef PA_USE_ALARM - alarm(0); -#endif - return result; - } catch(...) { -#ifdef PA_USE_ALARM - alarm(0); -#endif - if(sock>=0) - closesocket(sock); - rethrow; - } -#ifdef PA_USE_ALARM - } -#endif -} - #ifndef DOXYGEN -struct Http_pass_header_info { - Request_charsets* charsets; - String* request; - bool user_agent_specified; -}; +struct File_read_action_info { + char **data; size_t *data_size; + char* buf; size_t offset; size_t count; +}; #endif -static void http_pass_header(HashStringValue::key_type key, - HashStringValue::value_type value, - Http_pass_header_info *info) { - *info->request <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); +static void file_read_action(struct stat& finfo, int f, const String& file_spec, void *context) { + File_read_action_info& info=*static_cast(context); + size_t to_read_size=info.count; + if(!to_read_size) + to_read_size=(size_t)finfo.st_size; + if(to_read_size) { + if(info.offset) + lseek(f, info.offset, SEEK_SET); + *info.data=info.buf ? info.buf : (char *)pa_malloc_atomic(to_read_size+1); + ssize_t result=read(f, *info.data, to_read_size); + if(result<0) + throw Exception("file.read", &file_spec, "read failed: %s (%d)", strerror(errno), errno); + *info.data_size=result; + } else { // empty file + // for both, text and binary: for text we need that terminator, for binary we need nonzero pointer to be able to save such files + *info.data=(char *)pa_malloc_atomic(1); + *(char*)(*info.data)=0; + *info.data_size=0; + return; } - - 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 *str; size_t length; - HashStringValue* headers; -}; -#endif -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; - } +File_read_result file_read(Request_charsets& charsets, const String& file_spec, + bool as_text, HashStringValue *params, + bool fail_on_read_problem, + char* buf, size_t offset, size_t count, bool transcode_text_result) { + File_read_result result={false, 0, 0, 0}; + if(params){ + int valid_options=pa_get_valid_file_options_count(*params); + if(valid_options!=params->count()) + throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION); } - 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; - 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; - Value* vheaders=0; - Charset *asked_remote_charset=0; - const char* user_cstr=0; - const char* password_cstr=0; - - if(options) { - int valid_options=get_valid_file_options_count(*options); - - if(Value* vmethod=options->get(HTTP_METHOD_NAME)) { - valid_options++; - method=vmethod->as_string().cstr(); - } - if(Value* vform=options->get(HTTP_FORM_NAME)) { - valid_options++; - form=vform->get_hash(); - } - if(Value* vbody=options->get(HTTP_BODY_NAME)) { - valid_options++; - 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(); - - 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); + File_read_action_info info={&result.str, &result.length, buf, offset, count}; - head << "content-length: " << format(strlen(body_cstr), "%u") << CRLF; - } + result.success=file_read_action_under_lock(file_spec, "read", file_read_action, &info, as_text, fail_on_read_problem); - const char* head_cstr=head.cstr(String::L_UNSPECIFIED); + if(as_text){ + if(result.success){ + Charset* asked_charset=0; + if(params) + if(Value* vcharset_name=params->get(PA_CHARSET_NAME)) + asked_charset=&pa_charsets.get(vcharset_name->as_string()); - // 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 - 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 - 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(); + asked_charset=pa_charsets.checkBOM(result.str, result.length, asked_charset); - 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.str=const_cast(body.str); // hacking a little + result.length=body.length; } - - result.headers->put(HEADER_NAME, new VString(header_value)); } + if(result.length) + fix_line_breaks(result.str, result.length); } - // output response - 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; } -#endif +File_read_result file_load(Request& r, const String& file_spec, + bool as_text, HashStringValue *params, + bool fail_on_read_problem, + char* buf, size_t offset, size_t count, bool transcode_text_result) { -#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) { - File_read_action_info& info=*static_cast(context); - 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 %u bytes count not in [0..%u] valid range", - *info.data_size, to_read_size); - } else { // empty file - // for both, text and binary: for text we need that terminator, for binary we need nonzero pointer to be able to save such files - *info.data=new(PointerFreeGC) char[1]; - *(char*)(*info.data)=0; - *info.data_size=0; - return; - } -} -File_read_result file_read(Request_charsets& charsets, const String& file_spec, - bool as_text, HashStringValue *params, - 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://")) { if(offset || count) - throw Exception("parser.runtime", - 0, - "offset and load options are not supported for HTTP:// file load"); + throw Exception(PARSER_RUNTIME, 0, "offset and load options are not supported for HTTP:// file load"); // fail on read problem - File_read_http_result http=file_read_http(charsets, file_spec, as_text, params); + File_read_http_result http=pa_internal_file_read_http(r, file_spec, as_text, params, transcode_text_result); result.success=true; result.str=http.str; result.length=http.length; result.headers=http.headers; - } else { -#endif - if(params) { - int valid_options=get_valid_file_options_count(*params); - if(valid_options!=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); -#ifdef PA_HTTP - } -#endif - - if(result.success && as_text) { - // UTF-8 signature: EF BB BF - 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.str), result.length); - } + } else + result= + file_read(r.charsets, file_spec, as_text, params, fail_on_read_problem, buf, offset, count, transcode_text_result); return result; } + #ifdef PA_SAFE_MODE -void check_safe_mode(struct 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", - &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 + 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() + ); +} +#else +void check_safe_mode(struct stat, const String&, const char*) { +} +#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::L_FILE_SPEC); + const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC); int f; // first open, next stat: @@ -923,41 +234,32 @@ bool file_read_action_under_lock(const S // they delay update till open, so we would receive "!^test[" string // if would do stat, next open. // later: it seems, even this does not help sometimes - if((f=open(fname, O_RDONLY|(as_text?_O_TEXT:_O_BINARY)))>=0) { + if((f=open(fname, O_RDONLY|(as_text?_O_TEXT:_O_BINARY)))>=0) { try { if(pa_lock_shared_blocking(f)!=0) - throw Exception("file.lock", - &file_spec, - "shared lock failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); + throw Exception("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); struct stat finfo; - if(stat(fname, &finfo)!=0) + if(pa_fstat(f, &finfo)!=0) throw Exception("file.missing", // hardly possible: we just opened it OK - &file_spec, - "stat failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); + &file_spec, "stat failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); -#ifdef PA_SAFE_MODE check_safe_mode(finfo, file_spec, fname); -#endif - action(finfo, f, file_spec, fname, as_text, context); + action(finfo, f, file_spec, context); } catch(...) { pa_unlock(f);close(f); if(fail_on_read_problem) rethrow; - return false; + return false; } pa_unlock(f);close(f); return true; - } else { + } else { if(fail_on_read_problem) - throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, - &file_spec, - "%s failed: %s (%d), actual filename '%s'", - action_name, strerror(errno), errno, fname); + throw Exception(errno==EACCES ? "file.access" : (errno==ENOENT || errno==ENOTDIR || errno==ENODEV) ? "file.missing" : 0, + &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname); return false; } } @@ -966,19 +268,21 @@ void create_dir_for_file(const String& f size_t pos_after=1; 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); + pa_mkdir(file_spec.mid(0, pos_before).taint_cstr(String::L_FILE_SPEC), 0775); pos_after=pos_before+1; } } bool file_write_action_under_lock( const String& file_spec, - const char* action_name, File_write_action action, void *context, + const char* action_name, + File_write_action action, + void *context, bool as_text, bool do_append, bool do_block, bool fail_on_lock_problem) { - const char* fname=file_spec.cstr(String::L_FILE_SPEC); + const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC); int f; if(access(fname, W_OK)!=0) // no create_dir_for_file(file_spec); @@ -988,10 +292,7 @@ bool file_write_action_under_lock( |(as_text?_O_TEXT:_O_BINARY) |(do_append?O_APPEND:PA_O_TRUNC), 0664))>=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'", - strerror(errno), errno, fname); + Exception e("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); close(f); if(fail_on_lock_problem) throw e; @@ -999,7 +300,12 @@ bool file_write_action_under_lock( } try { - action(f, context); +#if (defined(HAVE_FCHMOD) && defined(PA_SAFE_MODE)) + struct stat finfo; + if(pa_fstat(f, &finfo)==0 && finfo.st_mode & 0111) + fchmod(f, finfo.st_mode & 0666/*clear executable bits*/); // backward: ignore errors if any +#endif + action(f, context); } catch(...) { #ifdef HAVE_FTRUNCATE if(!do_append) @@ -1016,208 +322,401 @@ bool file_write_action_under_lock( pa_unlock(f);close(f); return true; } else - throw Exception(errno==EACCES?"file.access":0, - &file_spec, - "%s failed: %s (%d), actual filename '%s'", - action_name, strerror(errno), errno, fname); + throw Exception(errno==EACCES ? "file.access" : 0, &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname); // here should be nothing, see rethrow above } #ifndef DOXYGEN struct File_write_action_info { - const char* str; size_t length; + 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.length) { - int written=write(f, info.str, info.length); + ssize_t written=write(f, info.str, info.length); if(written<0) - throw Exception(0, - 0, - "write failed: %s (%d)", strerror(errno), errno); + throw Exception("file.write", 0, "write failed: %s (%d)", strerror(errno), errno); + if((size_t)written!=info.length) + throw Exception("file.write", 0, "write failed: %u of %u bytes written", written, info.length); } } + void file_write( - const String& file_spec, - const char* data, size_t size, + Request_charsets& charsets, + const String& file_spec, + const char* data, + size_t size, bool as_text, - bool do_append) { + bool do_append, + Charset* asked_charset) { + + if(as_text && asked_charset){ + String::C body=String::C(data, size); + body=Charset::transcode(body, charsets.source(), *asked_charset); + data=body.str; + size=body.length; + }; + File_write_action_info info={data, size}; + file_write_action_under_lock( file_spec, - "write", file_write_action, &info, + "write", + file_write_action, + &info, as_text, do_append); } +static size_t get_dir(char* fname, size_t helper_length){ + bool dir=false; + size_t pos=0; + for(pos=helper_length; pos; pos--){ + char c=fname[pos-1]; + if(c=='/' || c=='\\'){ + fname[pos-1]=0; + dir=true; + } else if(dir) break; + } + return pos; +} + +static bool entry_readable(char* fname, bool need_dir) { + if(need_dir){ + size_t size=strlen(fname); + while(size) { + char c=fname[size-1]; + if(c=='/' || c=='\\') + fname[--size]=0; + else + break; + } + } + + struct stat finfo; + if(access(fname, R_OK)==0 && entry_exists(fname, &finfo)) { + bool is_dir=(finfo.st_mode&S_IFDIR) != 0; + return is_dir==need_dir; + } + return false; +} + +static bool entry_readable(const String& file_spec, bool need_dir) { + return entry_readable(file_spec.taint_cstrm(String::L_FILE_SPEC), need_dir); +} + // 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) - rmdir(file_spec, pos_before+1); - - rmdir(file_spec.mid(0, pos_after-1/* / */).cstr(String::L_FILE_SPEC)); + char* dir_spec=file_spec.taint_cstrm(String::L_FILE_SPEC); + size_t length=strlen(dir_spec); + while( (length=get_dir(dir_spec, length)) && (length > pos_after) ){ +#ifdef _MSC_VER + if(!entry_readable(dir_spec, true)) + break; + DWORD attrs=GetFileAttributes(dir_spec); + if( + (attrs==INVALID_FILE_ATTRIBUTES) + || !(attrs & FILE_ATTRIBUTE_DIRECTORY) + || (attrs & FILE_ATTRIBUTE_REPARSE_POINT) + ) + break; +#endif + if( rmdir(dir_spec) ) + break; + }; } -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) + +bool file_delete(const String& file_spec, bool fail_on_problem, bool keep_empty_dirs) { + const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC); + if(unlink(fname)!=0) { if(fail_on_problem) - throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, - &file_spec, - "unlink failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); + throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, + &file_spec, "unlink failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); else return false; + } + + if(!keep_empty_dirs) + rmdir(file_spec, 1); - rmdir(file_spec, 1); return true; } -void file_move(const String& old_spec, const String& new_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); + +void file_move(const String& old_spec, const String& new_spec, bool keep_empty_dirs) { + const char* old_spec_cstr=old_spec.taint_cstr(String::L_FILE_SPEC); + const char* new_spec_cstr=new_spec.taint_cstr(String::L_FILE_SPEC); create_dir_for_file(new_spec); if(rename(old_spec_cstr, new_spec_cstr)!=0) - throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0, - &old_spec, - "rename failed: %s (%d), actual filename '%s' to '%s'", - strerror(errno), errno, old_spec_cstr, new_spec_cstr); + throw Exception(errno==EACCES ? "file.access" : errno==ENOENT ? "file.missing" : 0, + &old_spec, "rename failed: %s (%d), actual filename '%s' to '%s'", strerror(errno), errno, old_spec_cstr, new_spec_cstr); - rmdir(old_spec, 1); + if(!keep_empty_dirs) + rmdir(old_spec, 1); } bool entry_exists(const char* fname, struct stat *afinfo) { struct stat lfinfo; - bool result=stat(fname, &lfinfo)==0; + bool result=pa_stat(fname, &lfinfo)==0; if(afinfo) *afinfo=lfinfo; return result; } bool entry_exists(const String& file_spec) { - const char* fname=file_spec.cstr(String::L_FILE_SPEC); + const char* fname=file_spec.taint_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::L_FILE_SPEC); - if(need_dir) { - size_t size=strlen(fname); - while(size) { - char c=fname[size-1]; - if(c=='/' || c=='\\') - fname[--size]=0; - else - break; - } - } - struct stat finfo; - if(access(fname, R_OK)==0 && entry_exists(fname, &finfo)) { - bool is_dir=(finfo.st_mode&S_IFDIR) != 0; - return is_dir==need_dir; - } - return false; -} -bool file_readable(const String& file_spec) { +bool file_exist(const String& file_spec) { return entry_readable(file_spec, false); } -bool dir_readable(const String& file_spec) { + +bool dir_exists(const String& file_spec) { return entry_readable(file_spec, true); } -const String* file_readable(const String& path, const String& name) { + +const String* file_exist(const String& path, const String& name) { String& result=*new String(path); - result << "/"; + if(path.last_char() != '/') + result << "/"; result << name; - return file_readable(result)?&result:0; + return file_exist(result)?&result:0; } + bool file_executable(const String& file_spec) { - return access(file_spec.cstr(String::L_FILE_SPEC), X_OK)==0; + return access(file_spec.taint_cstr(String::L_FILE_SPEC), X_OK)==0; } -bool file_stat(const String& file_spec, - size_t& rsize, - time_t& ratime, - time_t& rmtime, - time_t& rctime, - bool fail_on_read_problem) { - const char* fname=file_spec.cstr(String::L_FILE_SPEC); +bool file_stat(const String& file_spec, uint64_t& rsize, time_t& ratime, time_t& rmtime, time_t& rctime, bool fail_on_read_problem) { + const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC); struct stat finfo; - if(stat(fname, &finfo)!=0) + if(pa_stat(fname, &finfo)!=0) { if(fail_on_read_problem) - throw Exception("file.missing", - &file_spec, - "getting file size failed: %s (%d), real filename '%s'", - strerror(errno), errno, fname); + throw Exception("file.missing", &file_spec, "getting file size failed: %s (%d), real filename '%s'", strerror(errno), errno, fname); else return false; + } rsize=finfo.st_size; - ratime=finfo.st_atime; - rmtime=finfo.st_mtime; - rctime=finfo.st_ctime; + ratime=(time_t)finfo.st_atime; + rmtime=(time_t)finfo.st_mtime; + rctime=(time_t)finfo.st_ctime; return true; } +/** + String related functions +*/ + +bool capitalized(const char* s){ + bool upper=true; + for(const char* c=s; *c; c++){ + if(*c != (upper ? toupper((unsigned char)*c) : tolower((unsigned char)*c))) + return false; + upper=strchr("-_ ", *c) != 0; + } + return true; +} + +const char* capitalize(const char* s){ + if(!s || capitalized(s)) + return s; + + char* result=pa_strdup(s); + if(result){ + bool upper=true; + for(char* c=result; *c; c++){ + *c=upper ? (char)toupper((unsigned char)*c) : (char)tolower((unsigned char)*c); + upper=strchr("-_ ", *c) != 0; + } + } + return (const char*)result; +} + +char *str_lower(const char *s, size_t helper_length){ + char *result=pa_strdup(s, helper_length); + for(char* c=result; *c; c++) + *c=(char)tolower((unsigned char)*c); + return result; +} + +char *str_upper(const char *s, size_t helper_length){ + char *result=pa_strdup(s, helper_length); + for(char* c=result; *c; c++) + *c=(char)toupper((unsigned char)*c); + return result; +} + +void fix_line_breaks(char *str, size_t& length) { + //_asm int 3; + const char* const eob=str+length; + char* dest=str; + // fix DOS: \r\n -> \n + // fix Macintosh: \r -> \n + char* bol=str; + while(char* eol=(char*)memchr(bol, '\r', eob -bol)) { + size_t len=eol-bol; + if(dest!=bol) + memmove(dest, bol, len); + dest+=len; + *dest++='\n'; + + if(&eol[1]='0' && c<='9'){ + if(state == Flags) state=Width; // no more flags + break; + } else if(c=='d' || c=='i'){ + result=FormatInt; + } else if(strchr("feEgG", c)!=0){ + result=FormatDouble; + } else if(strchr("uoxX", c)!=0){ + result=FormatUInt; + } else { + return FormatInvalid; // invalid char + } + state=Done; + break; + case Done: + return FormatInvalid; // no chars allowed after 'type' + } + } + return result; +} + + +const char* format(double value, const char* fmt) { + char local_buf[MAX_NUMBER]; + int size=-1; + + if(fmt && strlen(fmt)){ + switch(format_type(fmt)){ + case FormatDouble: + size=snprintf(local_buf, sizeof(local_buf), fmt, value); + break; + case FormatInt: size=snprintf(local_buf, sizeof(local_buf), fmt, (int)value); - else - size=snprintf(local_buf, sizeof(local_buf), fmt, value); - else - size=snprintf(local_buf, sizeof(local_buf), "%d", (int)value); - - return pa_strdup(local_buf, size); + break; + case FormatUInt: + size=snprintf(local_buf, sizeof(local_buf), fmt, (uint)value); + break; + case FormatInvalid: + throw Exception(PARSER_RUNTIME, 0, "Incorrect format string '%s' was specified.", fmt); + } + } else + size=snprintf(local_buf, sizeof(local_buf), "%d", (int)value); + + if(size < 0 || size >= MAX_NUMBER-1){ // on win32 we manually reduce max size while printing + throw Exception(PARSER_RUNTIME, 0, "Error occure white executing snprintf with format string '%s'.", fmt); + } + + return pa_strdup(local_buf, (size_t)size); } size_t stdout_write(const void *buf, size_t size) { @@ -1237,44 +736,100 @@ size_t stdout_write(const void *buf, siz #endif } -char* unescape_chars(const char* cp, int len) { - char* s=new(PointerFreeGC) char[len + 1]; - enum EscapeState { - EscapeRest, - EscapeFirst, - EscapeSecond - } escapeState=EscapeRest; - uchar escapedValue=0; +enum EscapeState { + EscapeRest, + EscapeFirst, + EscapeSecond, + EscapeUnicode +}; + +// @todo prescan for reduce required size (unescaped sting in 1 byte charset requires less memory usually) +char* unescape_chars(const char* cp, int len, Charset* charset, bool js){ + char* s=new(PointerFreeGC) char[len+1]; // must be enough (%uXXXX==6 bytes, max utf-8 char length==6 bytes) + char* dst=s; + EscapeState escapeState=EscapeRest; + uint escapedValue=0; int srcPos=0; - int dstPos=0; - while(srcPos < len) { - uchar ch=(uchar)cp[srcPos]; - switch(escapeState) { - case EscapeRest: - if(ch=='%') { - escapeState=EscapeFirst; - } else if(ch=='+') { - s[dstPos++]=' '; - } else { - s[dstPos++]=ch; + short int jsCnt=0; + while(srcPosstore_Char((XMLByte*&)dst, (XMLCh)escapedValue, '?'); + escapeState=EscapeRest; + } + } else { + // not full unicode value + escapeState=EscapeRest; + } + break; } - break; - case EscapeFirst: - escapedValue=(uchar)(hex_value[ch] << 4); - escapeState=EscapeSecond; - break; - case EscapeSecond: - escapedValue +=hex_value[ch]; - s[dstPos++]=escapedValue; - escapeState=EscapeRest; - break; } - srcPos++; + + srcPos++; } - s[dstPos]=0; + + *dst=0; // zero-termination return s; } +char *search_stop(char*& current, char cstop_at) { + // sanity check + if(!current) + return 0; + + // skip leading WS + while(*current==' ' || *current=='\t') + current++; + if(!*current) + return current=0; + + char *result=current; + if(char *pstop_at=strchr(current, cstop_at)) { + *pstop_at=0; + current=pstop_at+1; + } else + current=0; + return result; +} + #ifdef WIN32 void back_slashes_to_slashes(char* s) { if(s) @@ -1282,121 +837,57 @@ void back_slashes_to_slashes(char* s) { if(*s=='\\') *s='/'; } -/* -void slashes_to_back_slashes(char* s) { - if(s) - for(; *s; s++) - if(*s=='/') - *s='\\'; -} -*/ #endif -bool StrEqNc(const char* s1, const char* s2, bool strict) { - while(true) { - if(!(*s1)) { - if(!(*s2)) - return true; - else - return !strict; - } else if(!(*s2)) - return !strict; - if(isalpha((unsigned char)*s1)) { - if(tolower((unsigned char)*s1) !=tolower((unsigned char)*s2)) - return false; - } else if((*s1) !=(*s2)) - return false; - s1++; - s2++; - } -} - -static bool isLeap(int year) { - return !( - (year % 4) || ((year % 400) && !(year % 100)) - ); -} - -int getMonthDays(int year, int month) { - int monthDays[]={ - 31, - isLeap(year) ? 29 : 28, - 31, - 30, - 31, - 30, - 31, - 31, - 30, - 31, - 30, - 31 - }; - return monthDays[month]; -} - -void remove_crlf(char* start, char* end) { - for(char* p=start; phh*/+1/*for zero-teminator*/]; + unsigned char *src=bytes; + unsigned char *end=bytes+size; + char *dest=bytes_hex; - int r; - // 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; + const char *hex=upcase? hex_digits : "0123456789abcdef"; -#if _MSC_VER - /* - win32: - mk:@MSITStore:C:\Program%20Files\Microsoft%20Visual%20Studio\MSDN\2001APR\1033\vccore.chm::/html/_crt__vsnprintf.2c_._vsnwprintf.htm - - if the number of bytes to write exceeds buffer, then count bytes are written and Ö1 is returned - */ - r=_vsnprintf(b, s, f, l); - if(r<0) - r=s; -#else - r=vsnprintf(b, s, f, l); - /* - solaris: - man vsnprintf - - The snprintf() function returns the number of characters - formatted, that is, the number of characters that would have - been written to the buffer if it were large enough. If the - value of n is 0 on a call to snprintf(), an unspecified - value less than 1 is returned. - */ - - if(r<0) - r=0; - else if((size_t)r>s) - r=s; -#endif - b[r]=0; - return r; -} + for(; src> 6)]; *outptr++ = base64_alphabet [c3 & 0x3f]; /* this is a bit ugly ... */ - if ((++already) >= 19) { + if ((++already) >= BASE64_GROUPS_IN_LINE) { *outptr++ = '\n'; already = 0; } @@ -1500,7 +994,7 @@ g_mime_utils_base64_encode_step (const u case 2: *saveout++ = *inptr++; case 1: *saveout++ = *inptr++; } - ((char *)save)[0] += inlen; + *(char *)save = *(char *)save+(char)inlen; } /*d(printf ("mode = %d\nc1 = %c\nc2 = %c\n", @@ -1560,9 +1054,9 @@ g_mime_utils_base64_encode_close (const } static unsigned char gmime_base64_rank[256] = { + 255,255,255,255,255,255,255,255,255,254,254,255,255,254,255,255, 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, - 255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255, - 255,255,255,255,255,255,255,255,255,255,255, 62,255,255,255, 63, + 254,255,255,255,255,255,255,255,255,255,255, 62,255,255,255, 63, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61,255,255,255, 0,255,255, 255, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25,255,255,255,255,255, @@ -1585,18 +1079,19 @@ static unsigned char gmime_base64_rank[2 * @out: output stream * @state: holds the number of bits that are stored in @save * @save: leftover bits that have not yet been decoded + * @strict: only base64 and whitespace chars are allowed * * Decodes a chunk of base64 encoded data. * * Returns the number of bytes decoded (which have been dumped in @out). **/ size_t -g_mime_utils_base64_decode_step (const unsigned char *in, size_t inlen, unsigned char *out, int *state, int *save) +g_mime_utils_base64_decode_step(const unsigned char *in, size_t inlen, unsigned char *out, int *state, int *save, bool strict=false) { - const register unsigned char *inptr; - register unsigned char *outptr; + const unsigned char *inptr; + unsigned char *outptr; const unsigned char *inend; - register guint32 saved; + int saved; unsigned char c; int i; @@ -1609,15 +1104,21 @@ g_mime_utils_base64_decode_step (const u inptr = in; while (inptr < inend) { c = gmime_base64_rank[*inptr++]; - if (c != 0xff) { - saved = (saved << 6) | c; - i++; - if (i == 4) { - *outptr++ = saved >> 16; - *outptr++ = saved >> 8; - *outptr++ = saved; - i = 0; - } + switch(c) { + case 0xff: // non-base64 and non-whitespace chars. not allowed in strict mode + if(strict) + throw Exception(BASE64_FORMAT, 0, "Invalid base64 char on position %d is detected", inptr-in-1); + case 0xfe: // whitespace chars 0x09, 0x0A, 0x0D, 0x20 are allowed in any mode + break; + default: + saved = (saved << 6) | c; + i++; + if (i == 4) { + *outptr++ = (unsigned char)(saved >> 16); + *outptr++ = (unsigned char)(saved >> 8); + *outptr++ = (unsigned char)(saved); + i = 0; + } } } @@ -1629,7 +1130,7 @@ g_mime_utils_base64_decode_step (const u i = 2; while (inptr > in && i) { inptr--; - if (gmime_base64_rank[*inptr] != 0xff) { + if (gmime_base64_rank[*inptr] <= 0xfe) { if (*inptr == '=' && outptr > out) outptr--; i--; @@ -1641,31 +1142,279 @@ g_mime_utils_base64_decode_step (const u } -char* pa_base64_encode(const char *in, size_t in_size) -{ - /* wont go to more than 2x size (overly conservative) */ - char* result=new(PointerFreeGC) char[in_size * 2 + 6]; +char* pa_base64_encode(const char *in, size_t in_size){ + size_t new_size = ((in_size / 3 + 1) * 4); + new_size += new_size / (BASE64_GROUPS_IN_LINE * 4)/*new lines*/ + 1/*zero terminator*/; + char* result = new(PointerFreeGC) char[new_size]; int state=0; int save=0; #ifndef NDEBUG size_t filled= #endif - g_mime_utils_base64_encode_close ((const unsigned char*)in, in_size, - (unsigned char*)result, &state, &save); - assert(filled <= in_size * 2 + 6); + g_mime_utils_base64_encode_close ((const unsigned char*)in, in_size, (unsigned char*)result, &state, &save); + + //throw Exception(PARSER_RUNTIME, 0, "%d %d %d", in_size, new_size, filled); + assert(filled <= new_size); return result; } -void pa_base64_decode(const char *in, size_t in_size, void*& result, size_t& result_size) -{ - /* wont go to more than had (overly conservative) */ - result=static_cast(new(PointerFreeGC) char[in_size+1/*terminator*/]); +struct File_base64_action_info { + unsigned char** base64; +}; + +static void file_base64_file_action(struct stat& finfo, int f, const String&, void *context) { + + if(finfo.st_size) { + File_base64_action_info& info=*static_cast(context); + *info.base64=new(PointerFreeGC) unsigned char[finfo.st_size * 2 + 6]; + unsigned char* base64 = *info.base64; + int state=0; + int save=0; + int nCount; + do { + unsigned char buffer[FILE_BUFFER_SIZE]; + nCount = file_block_read(f, buffer, sizeof(buffer)); + if( nCount ){ + size_t filled=g_mime_utils_base64_encode_step ((const unsigned char*)buffer, nCount, base64, &state, &save); + base64+=filled; + } + } while(nCount > 0); + g_mime_utils_base64_encode_close (0, 0, base64, &state, &save); + } +} + +char* pa_base64_encode(const String& file_spec){ + unsigned char* base64=0; + File_base64_action_info info={&base64}; + + file_read_action_under_lock(file_spec, "pa_base64_encode", file_base64_file_action, &info); + + return (char*)base64; +} + +void pa_base64_decode(const char *in, size_t in_size, char*& result, size_t& result_size, bool strict) { + // every 4 base64 bytes are converted into 3 normal bytes + // not full set (tail) of 4-bytes set is ignored + size_t new_size=in_size/4*3; + result=new(PointerFreeGC) char[new_size+1/*terminator*/]; + int state=0; int save=0; - result_size= - g_mime_utils_base64_decode_step ((const unsigned char*)in, in_size, - (unsigned char*)result, &state, &save); - assert(result_size <= in_size); - static_cast(result)[result_size]=0; // for text files + result_size=g_mime_utils_base64_decode_step ((const unsigned char*)in, in_size, (unsigned char*)result, &state, &save, strict); + assert(result_size <= new_size); + result[result_size]=0; // for text files + + if(strict && state!=0) + throw Exception(BASE64_FORMAT, 0, "Unexpected end of chars"); } + + +int file_block_read(const int f, unsigned char* buffer, const size_t size){ + int nCount = read(f, buffer, size); + if (nCount < 0) + throw Exception("file.read", 0, "read failed: %s (%d)", strerror(errno), errno); + return nCount; +} + +static unsigned long crc32Table[256]; +static void InitCrc32Table() +{ + if(crc32Table[1] == 0){ + // This is the official polynomial used by CRC32 in PKZip. + // Often times the polynomial shown reversed as 0x04C11DB7. + static const unsigned long dwPolynomial = 0xEDB88320; + + for(int i = 0; i < 256; i++) + { + unsigned long dwCrc = i; + for(int j = 8; j > 0; j--) + { + if(dwCrc & 1) + dwCrc = (dwCrc >> 1) ^ dwPolynomial; + else + dwCrc >>= 1; + } + crc32Table[i] = dwCrc; + } + } +} + +inline void CalcCrc32(const unsigned char byte, unsigned long &crc32) +{ + crc32 = ((crc32) >> 8) ^ crc32Table[(byte) ^ ((crc32) & 0x000000FF)]; +} + + +unsigned long pa_crc32(const char *in, size_t in_size){ + unsigned long crc32=0xFFFFFFFF; + + InitCrc32Table(); + for(size_t i = 0; i(context); + if(finfo.st_size) { + InitCrc32Table(); + int nCount=0; + do { + unsigned char buffer[FILE_BUFFER_SIZE]; + nCount = file_block_read(f, buffer, sizeof(buffer)); + for(int i = 0; i < nCount; i++) CalcCrc32(buffer[i], crc32); + } while(nCount > 0); + } +} + +unsigned long pa_crc32(const String& file_spec){ + unsigned long crc32=0xFFFFFFFF; + file_read_action_under_lock(file_spec, "crc32", file_crc32_file_action, &crc32); + return ~crc32; +} + +// content-type: xxx; charset=WE-NEED-THIS +// content-type: xxx; charset="WE-NEED-THIS" +// content-type: xxx; charset="WE-NEED-THIS"; +Charset* detect_charset(const char* content_type){ + if(content_type){ + char* CONTENT_TYPE=str_upper(content_type); + + if(const char* begin=strstr(CONTENT_TYPE, "CHARSET=")){ + begin+=8; // skip "CHARSET=" + char* end=0; + if(*begin && (*begin=='"' || *begin =='\'')){ + char quote=*begin; + begin++; + end=(char*)strchr(begin, quote); + } + if(!end) + end=(char*)strchr(begin, ';'); + + if(end) + *end=0; // terminator + + return *begin ? &pa_charsets.get_direct(begin) : 0; + } + } + return 0; +} + + +static bool is_latin(const char *in){ + for(; *in; in++){ + if ((unsigned char)(*in) > 0x7F) + return false; + } + return true; +} + +#define MAX_IDNA_LENGTH 256 + +const char *pa_idna_encode(const char *in, Charset &source_charset){ + if(!in || is_latin(in)) + return in; + + uint32_t utf32[MAX_IDNA_LENGTH]; + uint32_t *utf32_end=utf32; + + String::C sIn(in,strlen(in)); + + if(!source_charset.isUTF8()) + sIn=Charset::transcode(sIn, source_charset, pa_UTF8_charset); + + int status=pa_convertUTF8toUTF32((const UTF8**)&sIn.str, (const UTF8*)(sIn.str+sIn.length), &utf32_end, utf32+MAX_IDNA_LENGTH-1, strictConversion); + if(status != conversionOK) + throw Exception("idna encode", new String(in), "utf conversion failed (%d)", status); + + *utf32_end=0; + + char *result = (char *)pa_malloc(MAX_IDNA_LENGTH); + status=pa_idna_to_ascii_4z(utf32, result, MAX_IDNA_LENGTH, 0); + if(status != IDNA_SUCCESS) + throw Exception("idna encode", new String(in), "encode failed: %s", pa_idna_strerror(status)); + + return result; +} + +const char *pa_idna_decode(const char *in, Charset &asked_charset){ + if(!in || !(*in)) + return in; + + uint32_t utf32[MAX_IDNA_LENGTH]; + const uint32_t *utf32_start=utf32; + uint32_t *utf32_end; + + int status=pa_idna_to_unicode_4z(in, utf32, MAX_IDNA_LENGTH, 0); + if(status != IDNA_SUCCESS) + throw Exception("idna decode", new String(in), "decode failed: %s", pa_idna_strerror(status)); + + for(utf32_end=utf32; *utf32_end; utf32_end++); + + char *result = (char *)pa_malloc(MAX_IDNA_LENGTH); + char *result_end = result; + + status=pa_convertUTF32toUTF8(&utf32_start, utf32_end, (UTF8**)&result_end, (UTF8*)(result+MAX_IDNA_LENGTH-1), strictConversion); + + if(status != conversionOK) + throw Exception("idna decode", new String(in), "utf conversion failed (%d)", status); + + *result_end='\0'; + + if(!asked_charset.isUTF8()) + result = (char *)Charset::transcode(result, pa_UTF8_charset, asked_charset).cstr(); + + return result; +} +/// must be last in this file +#undef vsnprintf +int pa_vsnprintf(char* b, size_t s, const char* f, va_list l) { + if(!s) + return 0; + + int r; + // note: on win32 & maybe somewhere else + // vsnprintf do not writes terminating 0 in 'buffer full' case, reducing + // http://stackoverflow.com/questions/2915672/snprintf-and-visual-studio-2010 + --s; + + // clients do not check for negative 's', feature: ignore such prints + if((ssize_t)s<0) + return 0; + +#ifdef _MSC_VER + // win32: if the number of bytes to write exceeds buffer, then count bytes are written and -1 is returned + r=_vsnprintf(b, s, f, l); + if(r<0) + r=s; +#else + r=vsnprintf(b, s, f, l); + /* + solaris: man vsnprintf + + The snprintf() function returns the number of characters + formatted, that is, the number of characters that would have + been written to the buffer if it were large enough. If the + value of n is 0 on a call to snprintf(), an unspecified + value less than 1 is returned. + */ + + if(r<0) + r=0; + else if((size_t)r>s) + r=s; +#endif + b[r]=0; + return r; +} + +int pa_snprintf(char* b, size_t s, const char* f, ...) { + va_list l; + va_start(l, f); + int r=pa_vsnprintf(b, s, f, l); + va_end(l); + return r; +} +