--- parser3/src/main/pa_common.C 2002/11/26 12:39:44 1.131 +++ parser3/src/main/pa_common.C 2003/05/30 10:45:37 1.151 @@ -1,17 +1,18 @@ /** @file Parser: commonly functions. - Copyright(c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Copyright(c) 2001, 2003 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_COMMON_C="$Date: 2002/11/26 12:39:44 $"; +static const char* IDENT_COMMON_C="$Date: 2003/05/30 10:45:37 $"; #include "pa_common.h" #include "pa_exception.h" #include "pa_globals.h" #include "pa_hash.h" #include "pa_vstring.h" +#include "pa_vdate.h" #ifdef WIN32 # include @@ -28,6 +29,16 @@ static const char* IDENT_COMMON_C="$Date # define _O_BINARY 0 #endif +#ifdef HAVE_FTRUNCATE +# define PA_O_TRUNC 0 +#else +# ifdef _O_TRUNC +# define PA_O_TRUNC _O_TRUNC +# else +# error you must have either ftruncate function or _O_TRUNC bit declared +# endif +#endif + // locking constants #ifdef HAVE_FLOCK @@ -76,28 +87,17 @@ static int unlock(int fd) { FLOCK(F_TLOC #define DEFAULT_USER_AGENT "parser3" - -static char* strnchr(char* buf, size_t size, char c) { - // sanity check - if(!buf) - return 0; - - for(; size-->0; buf++) { - if(*buf==c) - return buf; - } - - return 0; -} - void fix_line_breaks(char* buf, size_t& size) { + if(size==0) + return; + //_asm int 3; const char* const eob=buf+size; char* dest=buf; // fix DOS: \r\n -> \n // fix Macintosh: \r -> \n char* bol=buf; - while(char* eol=strnchr(bol, eob -bol, '\r')) { + while(char* eol=(char*)memchr(bol, '\r', eob -bol)) { size_t len=eol-bol; if(dest!=bol) memcpy(dest, bol, len); @@ -139,8 +139,8 @@ static bool set_addr(struct sockaddr_in return true; } -static void http_read_response(String& response, int sock){ - bool check_status=true; +static int http_read_response(String& response, int sock, bool fail_on_status_ne_200){ + const String* status_code=0; ssize_t EOLat=0; while(true) { char *buf=(char *)response.pool().malloc(MAX_STRING); @@ -148,21 +148,21 @@ static void http_read_response(String& r if(size<=0) break; response.APPEND_TAINTED(buf, size, "remote HTTP server response", 0); - if(check_status && (EOLat=response.pos("\r\n", 2))>=0) { // checking status in first response - check_status=false; - + if(!status_code && (EOLat=response.pos("\r\n", 2))>=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); - const String* status_code=astatus.get_string(1); + status_code=astatus.get_string(1); - if(!status_code || *status_code!="200") + if(fail_on_status_ne_200 && *status_code!="200") throw Exception("http.status", status_code, "invalid HTTP response status"); } - } - if(check_status) + } + if(status_code) + return status_code->as_int(); + else throw Exception("http.response", 0, "bad response from host - no status found (size=%lu)", response.size()); @@ -171,42 +171,50 @@ static void http_read_response(String& r /* ********************** 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 void http_request(String& response, +static int http_request(String& response, const String *origin_string, const char* host, int port, const char* request, - int timeout){ + int timeout, + bool fail_on_status_ne_200) { if(!host) throw Exception("http.host", origin_string, "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 { -#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, "can not resolve hostname \"%s\"", host); @@ -225,20 +233,23 @@ static void http_request(String& respons origin_string, "error sending request: %s (%d)", strerror(errno), errno); - http_read_response(response, sock); + result=http_read_response(response, 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); -#ifdef WE_CAN_USE_ALARM + /*re*/throw; } -#endif - } catch(...) { - if(sock>=0) - closesocket(sock); -#ifdef WE_CAN_USE_ALARM - alarm(0); -#endif - /*re*/throw; +#ifdef PA_USE_ALARM } +#endif } #ifndef DOXYGEN @@ -252,8 +263,10 @@ static void http_pass_header(const Hash: Http_pass_header_info& i=*static_cast(info); Pool& pool=i.request->pool(); - *(i.request)<(value))->get_string()<<"\n"; - + *(i.request)<(value), String::UL_HTTP_HEADER, false) + <<"\n"; + if(key.change_case(pool, String::CC_UPPER)=="USER-AGENT") i.user_agent_specified=true; } @@ -265,18 +278,23 @@ static void file_read_http(Pool& pool, c int port; const char* method="GET"; int timeout=2; + bool fail_on_status_ne_200=true; Value *vheaders=0; - char* connect_string=file_spec.cstr(String::UL_UNSPECIFIED); - if(strncmp(connect_string, "http://", 7)!=0) + String& connect_string=*new(pool) String(pool); // must not be local [exception may be reported outside] + // 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, - &file_spec, + &connect_string, "does not start with http://"); //never - connect_string+=7; + connect_string_cstr+=7; - strncpy(host, connect_string, sizeof(host)-1); host[sizeof(host)-1]=0; + strncpy(host, connect_string_cstr, sizeof(host)-1); host[sizeof(host)-1]=0; char* host_uri=lsplit(host, '/'); - uri=host_uri?connect_string+(host_uri-1-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; @@ -294,24 +312,29 @@ static void file_read_http(Pool& pool, c if(vheaders=static_cast(options->get(*http_headers_name))) { valid_options++; } + if(Value *vany_status=static_cast(options->get(*http_any_status_name))) { + valid_options++; + fail_on_status_ne_200=!vany_status->as_bool(); + } + if(valid_options!=options->size()) throw Exception("parser.runtime", 0, "invalid option passed"); - } + } //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(&file_spec)) { + 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", - &file_spec, + &connect_string, "headers param must be hash"); }; if(!user_agent_specified) // defaulting @@ -320,14 +343,15 @@ static void file_read_http(Pool& pool, c //sending request String response(pool); - http_request(response, - &file_spec, host, port, request.cstr(String::UL_UNSPECIFIED), timeout); + int status_code=http_request(response, + &connect_string, host, port, request.cstr(String::UL_UNSPECIFIED), + timeout, fail_on_status_ne_200); //processing results int pos=response.pos("\r\n\r\n", 4); if(pos<1){ throw Exception("http.response", - &file_spec, + &connect_string, "bad response from host - no headers found"); } String header_block=response.mid(0, pos); @@ -338,20 +362,13 @@ static void file_read_http(Pool& pool, c size_t pos_after_ref=0; header_block.split(aheaders, &pos_after_ref, "\r\n", 2); - //processing status code - const String *status_line=aheaders.get_string(0); - if(!status_line){ - throw Exception("http.response", - &file_spec, - "bad response from host - no status line "); - } //processing headers for(int i=1;ipos(": ", 2); if(pos<1) throw Exception("http.response", - &file_spec, + &connect_string, "bad response from host - bad header \"%s\"", line->cstr()); headers.put( @@ -359,14 +376,16 @@ static void file_read_http(Pool& pool, c new(pool) VString(line->mid(pos+2, line->size()))); } else throw Exception("http.response", - &file_spec, + &connect_string, "bad response from host - bad headers \"%s\"", header_block.cstr()); } // output response data=body.cstr(); data_size=body.size(); - if(out_fields) + if(out_fields) { + headers.put(*file_status_name, new(pool) VInt(pool, status_code)); *out_fields=&headers; + } } #ifndef DOXYGEN @@ -433,6 +452,23 @@ bool file_read(Pool& pool, const String& 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(Pool& pool, const String& file_spec, const char* action_name, File_read_action action, void *context, bool as_text, @@ -464,12 +500,8 @@ bool file_read_action_under_lock(Pool& p "stat failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname); -#ifdef NO_FOREIGN_GROUP_FILES - if(finfo.st_gid/*foreign?*/!=getegid()) { - throw Exception("parser.runtime", - &file_spec, - "parser reading files of foreign group disabled [recompile parser without --disable-foreign-group-files configure option], actual filename '%s'", - fname); +#ifdef PA_SAFE_MODE + check_safe_mode(finfo, file_spec, fname); #endif action(pool, finfo, f, file_spec, fname, as_text, context); @@ -516,7 +548,7 @@ bool file_write_action_under_lock( if((f=open(fname, O_CREAT|O_RDWR |(as_text?_O_TEXT:_O_BINARY) - |(do_append?O_APPEND:0), 0664))>=0) { + |(do_append?O_APPEND:PA_O_TRUNC), 0664))>=0) { if((do_block?lock_exclusive_blocking(f):lock_exclusive_nonblocking(f))!=0) { Exception e("file.lock", &file_spec, @@ -531,14 +563,18 @@ bool file_write_action_under_lock( try { action(f, context); } catch(...) { +#ifdef HAVE_FTRUNCATE 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; } +#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); return true; } else @@ -935,3 +971,77 @@ int pa_sleep(unsigned long secs, unsigne return (select(0, NULL, NULL, NULL, &t) == -1 ? errno : 0); #endif } + + +// attributed meaning + +/// http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.3 +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; +}