--- parser3/src/main/pa_common.C 2003/01/16 09:07:20 1.142 +++ parser3/src/main/pa_common.C 2004/02/11 15:33:15 1.174 @@ -1,23 +1,48 @@ /** @file Parser: commonly functions. - Copyright(c) 2001, 2002 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: 2003/01/16 09:07:20 $"; +static const char * const IDENT_COMMON_C="$Date: 2004/02/11 15:33:15 $"; #include "pa_common.h" #include "pa_exception.h" -#include "pa_globals.h" #include "pa_hash.h" +#include "pa_globals.h" +#include "pa_request_charsets.h" +#include "pa_charsets.h" + +#define PA_HTTP + +#ifdef PA_HTTP #include "pa_vstring.h" -#include "pa_vdate.h" +#include "pa_vint.h" +#include "pa_vhash.h" +#include "pa_vtable.h" + +#ifdef CYGWIN +#define _GNU_H_WINDOWS32_SOCKETS +// for PASCAL +#include +// SOCKET +typedef u_int SOCKET; +int PASCAL closesocket(SOCKET); +#else +# if defined(WIN32) +# include +# else +# define closesocket close +# endif +#endif -#ifdef WIN32 -# include #else -# define closesocket close + +# if defined(WIN32) +# include +# endif + #endif // some maybe-undefined constants @@ -39,64 +64,45 @@ static const char* IDENT_COMMON_C="$Date # 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 +// defines for globals -#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); } +#define FILE_STATUS_NAME "status" -#else -#ifdef HAVE_FCNTL +// globals -#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); } +const String file_status_name(FILE_STATUS_NAME); -#else -#ifdef HAVE_LOCKF +// defines for statics -#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); } +#define HTTP_METHOD_NAME "method" +#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" -#else +// statics -#error unable to find file locking func +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); +static const String http_charset_name(HTTP_CHARSET_NAME); +static const String http_tables_name(HTTP_TABLES_NAME); -#endif -#endif -#endif -#endif +// defines #define DEFAULT_USER_AGENT "parser3" +// 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) @@ -106,24 +112,30 @@ void fix_line_breaks(char* buf, size_t& if(&eol[1]=0) { // checking status in first response - const String& status_line=response.mid(0, (size_t)EOLat); - Array astatus(response.pool()); - size_t pos_after_ref=0; status_line.split(astatus, &pos_after_ref, " ", 1); - status_code=astatus.get_string(1); + 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, CRLF))) { // checking status in first response + const String status_line(pa_strdup(response, EOLat-response)); + ArrayString astatus; + 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 && *status_code!="200") + if(fail_on_status_ne_200 && result!=200) throw Exception("http.status", - status_code, + &status_code, "invalid HTTP response status"); } } - if(status_code) - return status_code->as_int(); + if(result) + return result; 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){ 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, int port, + const char* request, + int +#ifdef PA_USE_ALARM + timeout +#endif + , + 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); + 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 - /*re*/throw; + 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; }; #endif -static void http_pass_header(const Hash::Key& key, Hash::Val *value, void *info) -{ - Http_pass_header_info& i=*static_cast(info); - Pool& pool=i.request->pool(); - - *(i.request)<(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 <charsets->source(), String::CC_UPPER)=="USER-AGENT") + info->user_agent_specified=true; } -static void file_read_http(Pool& pool, const String& file_spec, - void*& data, size_t& data_size, - Hash *options=0, Hash** out_fields=0) { + + +static Charset* detect_charset(Charset& source_charset, const String& content_type_value) { + const String::Body CONTENT_TYPE_VALUE= + content_type_value.change_case(source_charset, String::CC_UPPER); + // content-type: xxx/xxx; source_charset=WE-NEED-THIS + // content-type: xxx/xxx; source_charset="WE-NEED-THIS" + // content-type: xxx/xxx; source_charset="WE-NEED-THIS"; + size_t before_charseteq_pos=CONTENT_TYPE_VALUE.pos("CHARSET="); + if(before_charseteq_pos!=STRING_NOT_FOUND) { + size_t charset_begin=before_charseteq_pos+8/*CHARSET="*/; + size_t open_quote_pos=CONTENT_TYPE_VALUE.pos('"', charset_begin); + bool quoted=open_quote_pos==charset_begin; + if(quoted) + charset_begin++; // skip opening '"' + size_t charset_end=CONTENT_TYPE_VALUE.length(); + if(quoted) { + size_t close_quote_pos=CONTENT_TYPE_VALUE.pos('"', charset_begin); + if(close_quote_pos!=STRING_NOT_FOUND) + charset_end=close_quote_pos; + } else { + size_t delim_pos=CONTENT_TYPE_VALUE.pos(';', charset_begin); + if(delim_pos!=STRING_NOT_FOUND) + charset_end=delim_pos; + } + const String::Body CHARSET_NAME_BODY= + CONTENT_TYPE_VALUE.mid(charset_begin, charset_end); + + return &charsets.get(CHARSET_NAME_BODY); + } + + return 0; +} + + +#ifndef DOXYGEN +struct File_read_http_result { + char *str; size_t length; + HashStringValue* headers; +}; +#endif +/// @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="GET"; + const char* method=0; int timeout=2; bool fail_on_status_ne_200=true; - Value *vheaders=0; - - String connect_string(pool); - // not in ^sql{... UL_SQL ...} spirit, but closer to ^file::load one - connect_string.append(file_spec, String::UL_URI); // tainted pieces -> URI pieces - - char* connect_string_cstr=connect_string.cstr(String::UL_UNSPECIFIED); - if(strncmp(connect_string_cstr, "http://", 7)!=0) - throw Exception(0, - &connect_string, - "does not start with http://"); //never - connect_string_cstr+=7; - - strncpy(host, connect_string_cstr, sizeof(host)-1); host[sizeof(host)-1]=0; - char* host_uri=lsplit(host, '/'); - uri=host_uri?connect_string_cstr+(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; - if(Value *vmethod=static_cast(options->get(*http_method_name))) { + if(Value* vmethod=options->get(http_method_name)) { valid_options++; - method=vmethod->as_string().cstr(); + method=vmethod->as_string().cstr(); } - if(Value *vtimeout=static_cast(options->get(*http_timeout_name))) { + if(Value* vtimeout=options->get(http_timeout_name)) { valid_options++; timeout=vtimeout->as_int(); } - if(vheaders=static_cast(options->get(*http_headers_name))) { + if((vheaders=options->get(http_headers_name))) { valid_options++; } - if(Value *vany_status=static_cast(options->get(*http_any_status_name))) { + 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(valid_options!=options->size()) + 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(pool); - request<< method <<" "<< uri <<" HTTP/1.0\nHost: "<< host<<"\n"; - bool user_agent_specified=false; - if(vheaders && !vheaders->is_string()) { // allow empty - if(Hash *headers=vheaders->get_hash(&connect_string)) { - Http_pass_header_info info={&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, 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 + 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(pool); - int status_code=http_request(response, - &connect_string, host, port, request.cstr(String::UL_UNSPECIFIED), + char* response; + size_t response_size; + int status_code=http_request(response, response_size, + host, port, request_cstr, timeout, fail_on_status_ne_200); //processing results - int pos=response.pos("\r\n\r\n", 4); - if(pos<1){ + char* raw_body; size_t raw_body_size; + char* headers_end_at=strstr(response, CRLF CRLF /*change '4' below along!*/); + if(headers_end_at) { + raw_body=headers_end_at+4; + raw_body_size=response_size-(raw_body-response); + } else throw Exception("http.response", &connect_string, "bad response from host - no headers found"); - } - String header_block=response.mid(0, pos); - String body=response.mid(pos+4, response.size()); - Array aheaders(pool); - Hash& headers=*new(pool) Hash(pool); - size_t pos_after_ref=0; - header_block.split(aheaders, &pos_after_ref, "\r\n", 2); + *headers_end_at=0; + const String header_block(response, headers_end_at-response, true); + + ArrayString aheaders; + result.headers=new HashStringValue; + VHash* vtables=new VHash; + result.headers->put(http_tables_name, vtables); + HashStringValue& tables=vtables->hash(); + + size_t pos_after=0; + header_block.split(aheaders, pos_after, CRLF); //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()); - - headers.put( - line->mid(0, pos).change_case(pool, String::CC_UPPER), - new(pool) VString(line->mid(pos+2, line->size()))); - } else + size_t aheaders_count=aheaders.count(); + Charset* real_remote_charset=0; // undetected, yet + 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)); } + // defaulting to used-asked charset [it's never empty!] + if(!real_remote_charset) + real_remote_charset=asked_remote_charset; // output response - data=body.cstr(); data_size=body.size(); - if(out_fields) { - headers.put(*file_status_name, new(pool) VInt(pool, status_code)); - *out_fields=&headers; - } + 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 + 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 + #ifndef DOXYGEN struct File_read_action_info { - void **data; size_t *data_size; + char **data; size_t *data_size; }; #endif -static void file_read_action(Pool& pool, - struct stat& finfo, - int f, - const String& file_spec, const char* fname, bool as_text, - void *context) { +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); if(size_t to_read_size=(size_t)finfo.st_size) { - *info.data=pool.malloc(to_read_size+(as_text?1:0), 3); + *info.data=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) { - *info.data=pool.malloc(1); + *info.data=new(PointerFreeGC) char[1]; *(char*)(*info.data)=0; } else *info.data=0; @@ -410,45 +556,70 @@ static void file_read_action(Pool& pool, return; } } -bool file_read(Pool& pool, const String& file_spec, - void*& data, size_t& data_size, - bool as_text, Hash *params, Hash** out_fields, +File_read_result file_read(Request_charsets& charsets, const String& file_spec, + bool as_text, HashStringValue *params, bool fail_on_read_problem) { - bool result; - if(file_spec.starts_with("http://", 7)) { + File_read_result result={false, 0, 0, 0}; +#ifdef PA_HTTP + if(file_spec.starts_with("http://")) { // fail on read problem - file_read_http(pool, file_spec, data, data_size, params, out_fields); - result=true; + File_read_http_result http=file_read_http(charsets, file_spec, as_text, params); + result.success=true; + result.str=http.str; + result.length=http.length; + result.headers=http.headers; } else { - File_read_action_info info={&data, &data_size}; - result=file_read_action_under_lock(pool, file_spec, +#endif + if(params && params->count()) + throw Exception("parser.runtime", + 0, + "invalid option passed"); + + File_read_action_info info={&result.str, &result.length}; + 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 && as_text) { + if(result.success && as_text) { // UTF-8 signature: EF BB BF - if(data_size>=3) { - char *in=(char *)data; - if((in[0] == '\xEF') && (in[1] == '\xBB') && - (in[2] == '\xBF')) { - data=in+3; data_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 *)(data), data_size); - // note: after fixing - ((char*&)(data))[data_size]=0; + fix_line_breaks((char *)(result.str), result.length); } return result; } -bool file_read_action_under_lock(Pool& pool, const String& file_spec, +#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: @@ -462,53 +633,48 @@ bool file_read_action_under_lock(Pool& p // 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'", - strerror(errno), errno, fname); + strerror(errno), errno, fname); struct stat finfo; if(stat(fname, &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); + 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(pool, finfo, f, file_spec, fname, as_text, context); + action(finfo, f, file_spec, fname, as_text, context); } catch(...) { - unlock(f);close(f); + pa_unlock(f);close(f); if(fail_on_read_problem) - /*re*/throw; + rethrow; return false; } - unlock(f);close(f); + pa_unlock(f);close(f); return true; } 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); + action_name, strerror(errno), errno, fname); return false; } } static void create_dir_for_file(const String& file_spec) { size_t pos_after=1; - int pos_before; - while((pos_before=file_spec.pos("/", 1, 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; } } @@ -520,7 +686,7 @@ 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); @@ -529,11 +695,11 @@ 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'", - strerror(errno), errno, fname); + strerror(errno), errno, fname); close(f); if(fail_on_lock_problem) throw e; @@ -547,33 +713,33 @@ 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); - /*re*/throw; + pa_unlock(f);close(f); + rethrow; } #ifdef HAVE_FTRUNCATE 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, &file_spec, "%s failed: %s (%d), actual filename '%s'", - action_name, strerror(errno), errno, fname); + action_name, strerror(errno), errno, fname); // here should be nothing, see rethrow above } #ifndef DOXYGEN struct File_write_action_info { - const void *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, @@ -582,33 +748,33 @@ static void file_write_action(int f, voi } void file_write( const String& file_spec, - const void *data, size_t size, + const char* data, size_t size, bool as_text, bool do_append) { File_write_action_info info={data, size}; file_write_action_under_lock( - file_spec, - "write", file_write_action, &info, - as_text, - do_append); + file_spec, + "write", file_write_action, &info, + as_text, + do_append); } // 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("/", 1, 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'", - strerror(errno), errno, fname); + strerror(errno), errno, fname); else return false; @@ -616,8 +782,8 @@ 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); @@ -625,7 +791,7 @@ void file_move(const String& old_spec, c 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); + strerror(errno), errno, old_spec_cstr, new_spec_cstr); rmdir(old_spec, 1); } @@ -640,12 +806,12 @@ 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.cstr(String::UL_FILE_SPEC); + char* fname=file_spec.cstrm(String::L_FILE_SPEC); if(need_dir) { size_t size=strlen(fname); while(size) { @@ -669,14 +835,14 @@ bool file_readable(const String& file_sp bool dir_readable(const String& file_spec) { return entry_readable(file_spec, true); } -String *file_readable(const String& path, const String& name) { - String *result=new(path.pool()) String(path); - *result << "/"; - *result << name; - return file_readable(*result)?result:0; +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, @@ -685,15 +851,14 @@ bool file_stat(const String& file_spec, time_t& rmtime, time_t& rctime, bool fail_on_read_problem) { - Pool& pool=file_spec.pool(); - 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", &file_spec, "getting file size failed: %s (%d), real filename '%s'", - strerror(errno), errno, fname); + strerror(errno), errno, fname); else return false; rsize=finfo.st_size; @@ -745,7 +910,7 @@ char* rsplit(char* string, char delim) { } /// @todo less stupid type detection -char* format(Pool& pool, double value, char* fmt) { +const char* format(double value, char* fmt) { char local_buf[MAX_NUMBER]; size_t size; @@ -760,15 +925,13 @@ char* format(Pool& pool, double value, c else size=snprintf(local_buf, sizeof(local_buf), "%d", (int)value); - char* pool_buf=(char *)pool.malloc(size+1, 4); - memcpy(pool_buf, local_buf, size+1); - return pool_buf; + return pa_strdup(local_buf, size); } size_t stdout_write(const void *buf, size_t size) { #ifdef WIN32 do{ - int chunk_written=fwrite(buf, 1, min(8*0x400, size), stdout); + int chunk_written=fwrite(buf, 1, min((size_t)8*0x400, size), stdout); if(chunk_written<=0) break; size-=chunk_written; @@ -781,8 +944,8 @@ size_t stdout_write(const void *buf, siz #endif } -char* unescape_chars(Pool& pool, const char* cp, int len) { - char* s=(char *)pool.malloc(len + 1, 5); +char* unescape_chars(const char* cp, int len) { + char* s=new(PointerFreeGC) char[len + 1]; enum EscapeState { EscapeRest, EscapeFirst, @@ -898,12 +1061,17 @@ 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: 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 –1 is returned + 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) @@ -923,7 +1091,7 @@ 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; @@ -937,90 +1105,3 @@ int __snprintf(char* b, size_t s, const va_end(l); return r; } - -int pa_sleep(unsigned long secs, unsigned long usecs) { - for (; usecs >= 1000000; ++secs, usecs -= 1000000); - -#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 -} - - -// attributed meaning - -static size_t date_attribute(const VDate& vdate, char *buf, size_t buf_size) { - const char month_names[12][4]={ - "Jan","Feb","Mar","Apr","May","Jun","Jul","Aug","Sep","Oct","Nov","Dec"}; - const char days[7][4]={ - "Sun","Mon","Tue","Wed","Thu","Fri","Sat"}; - - time_t when=vdate.get_time(); - struct tm *tms=gmtime(&when); - if(!tms) - throw Exception(0, - 0, - "bad time in attribute value (seconds from epoch=%ld)", when); - return snprintf(buf, MAX_STRING, "%s, %.2d-%s-%.4d %.2d:%.2d:%.2d GMT", - days[tms->tm_wday], - tms->tm_mday,month_names[tms->tm_mon],tms->tm_year+1900, - tms->tm_hour,tms->tm_min,tms->tm_sec); -} -static void append_attribute_meaning(String& result, - Value& value, String::Untaint_lang lang, bool forced) { - if(const String *string=value.get_string()) - result.append(string->join_chains(result.pool(), 0), lang, forced); - else - if(Value *vdate=value.as(VDATE_TYPE, false)) { - char *buf=(char *)result.malloc(MAX_STRING); - size_t size=date_attribute(*static_cast(vdate), - buf, MAX_STRING); - - result.APPEND_CLEAN(buf, size, "converted from date", 0); - } else - throw Exception("parser.runtime", - &result, - "trying to append here neither string nor date (%s)", - value.type()); -} -#ifndef DOXYGEN -struct Attributed_meaning_info { - String *header; // header line being constructed - String::Untaint_lang lang; // language in which to append to that line - bool forced; // do they force that lang? -}; -#endif -static void append_attribute_subattribute(const Hash::Key& akey, Hash::Val *avalue, - void *info) { - if(akey==VALUE_NAME) - return; - - Attributed_meaning_info& ami=*static_cast(info); - - // ...; charset=windows1251 - *ami.header << "; "; - ami.header->append(akey, ami.lang, ami.forced); - *ami.header << "="; - append_attribute_meaning(*ami.header, *static_cast(avalue), ami.lang, ami.forced); -} -const String& attributed_meaning_to_string(Value& meaning, - String::Untaint_lang lang, bool forced) { - String &result=*new(meaning.pool()) String(meaning.pool()); - if(Hash *hash=meaning.get_hash(0)) { - // $value(value) $subattribute(subattribute value) - if(Value *value=static_cast(hash->get(*value_name))) - append_attribute_meaning(result, *value, lang, forced); - - Attributed_meaning_info attributed_meaning_info={&result, lang}; - hash->for_each(append_attribute_subattribute, &attributed_meaning_info); - } else // result value - append_attribute_meaning(result, meaning, lang, forced); - - return result; -}