--- parser3/src/main/pa_common.C 2002/11/25 15:03:41 1.128 +++ parser3/src/main/pa_common.C 2003/04/15 06:44:34 1.140.2.3 @@ -5,13 +5,15 @@ Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char* IDENT_COMMON_C="$Date: 2002/11/25 15:03:41 $"; +static const char* IDENT_COMMON_C="$Date: 2003/04/15 06:44:34 $"; #include "pa_common.h" #include "pa_exception.h" #include "pa_globals.h" #include "pa_hash.h" #include "pa_vstring.h" +#include "pa_vdate.h" +#include "pa_request.h" #ifdef WIN32 # include @@ -28,6 +30,16 @@ static const char* IDENT_COMMON_C="$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 +88,17 @@ 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); @@ -117,9 +118,10 @@ void fix_line_breaks(char* buf, size_t& char* file_read_text(Pool& pool, const String& file_spec, bool fail_on_read_problem, - Hash *params, Hash** out_fields) { + Hash* options, Request* r, + Hash** out_fields) { void *result; size_t size; - return file_read(pool, file_spec, result, size, true, params, out_fields, fail_on_read_problem)?(char *)result:0; + return file_read(pool, file_spec, result, size, true, options, r, out_fields, fail_on_read_problem)?(char *)result:0; } //http request stuff @@ -140,22 +142,41 @@ static bool set_addr(struct sockaddr_in } static void http_read_response(String& response, int sock){ + bool check_status=true; + ssize_t EOLat=0; while(true) { char *buf=(char *)response.pool().malloc(MAX_STRING); ssize_t size=recv(sock, buf, MAX_STRING, 0); if(size<=0) break; - response.APPEND_TAINTED(buf, size, "http response", 0); + 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; + + 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); + + if(!status_code || *status_code!="200") + throw Exception("http.status", + status_code, + "invalid HTTP response status"); + } } + if(check_status) + throw Exception("http.response", + 0, + "bad response from host - no status found (size=%lu)", 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); @@ -166,56 +187,60 @@ static void http_request(String& respons const String *origin_string, const char* host, int port, const char* request, - int timeout){ + int timeout) { if(!host) throw Exception("http.host", origin_string, "zero hostname"); //never -#ifdef WE_CAN_USE_ALARM +#ifdef PA_USE_ALARM signal(SIGALRM, timeout_handler); #endif int sock=-1; +#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"); + } else + alarm(timeout); +#endif try { -#ifdef WE_CAN_USE_ALARM - if(sigsetjmp(timeout_env, 1)) - throw Exception("http.timeout", + struct sockaddr_in dest; + + if(!set_addr(&dest, host, port)) + throw Exception("http.host", origin_string, - "timeout occured while retrieving document"); - else { - alarm(timeout); -#endif - struct sockaddr_in dest; + "can not resolve hostname \"%s\"", host); - if(!set_addr(&dest, host, port)) - throw Exception("http.host", - origin_string, - "can not resolve hostname \"%s\"", host); - - if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) - throw Exception("http.connect", - origin_string, - "can not make socket: %s (%d)", strerror(errno), errno); - if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) - throw Exception("http.connect", - origin_string, - "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, - "error sending request: %s (%d)", strerror(errno), errno); - - http_read_response(response, sock); - if(sock>=0) - closesocket(sock); -#ifdef WE_CAN_USE_ALARM - } + if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) + throw Exception("http.connect", + origin_string, + "can not make socket: %s (%d)", strerror(errno), errno); + if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) + throw Exception("http.connect", + origin_string, + "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, + "error sending request: %s (%d)", strerror(errno), errno); + + http_read_response(response, sock); + closesocket(sock); +#ifdef PA_USE_ALARM + alarm(0); #endif } catch(...) { if(sock>=0) closesocket(sock); -#ifdef WE_CAN_USE_ALARM +#ifdef PA_USE_ALARM alarm(0); #endif /*re*/throw; @@ -233,8 +258,10 @@ 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; } @@ -242,22 +269,26 @@ static void file_read_http(Pool& pool, c void*& data, size_t& data_size, Hash *options=0, Hash** out_fields=0) { char host[MAX_STRING]; - char* uri; + const char* uri; int port; const char* method="GET"; int timeout=2; Value *vheaders=0; - char* connect_string=file_spec.cstr(String::UL_UNSPECIFIED); - if(strncmp(connect_string, "http://", 7)!=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, - &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; @@ -279,20 +310,20 @@ static void file_read_http(Pool& pool, c 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 @@ -302,13 +333,13 @@ 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); + &connect_string, host, port, request.cstr(String::UL_UNSPECIFIED), timeout); //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); @@ -323,21 +354,16 @@ static void file_read_http(Pool& pool, c const String *status_line=aheaders.get_string(0); if(!status_line){ throw Exception("http.response", - &file_spec, + &connect_string, "bad response from host - no status line "); } //processing headers - Array astatus(pool); - pos_after_ref=0; - status_line->split(astatus, &pos_after_ref, " ", 1); - const String * status_code=astatus.get_string(1); - 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( @@ -345,21 +371,20 @@ 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) { - headers.put(*file_status_name, new(pool) VString(*status_code)); + if(out_fields) *out_fields=&headers; - } } #ifndef DOXYGEN struct File_read_action_info { void **data; size_t *data_size; + Hash* options; Request* r; }; #endif static void file_read_action(Pool& pool, @@ -368,16 +393,27 @@ static void file_read_action(Pool& pool, 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_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", - *info.data_size, to_read_size); - } else { // empty file + int offset=0; + int limit=-1; + if(info.options) { + int valid_options=0; + if(Value *voffset=(Value *)info.options->get(*sql_offset_name)) { + valid_options++; + offset=info.r->process_to_value(*voffset).as_int(); + } + if(Value *vlimit=(Value *)info.options->get(*sql_limit_name)) { + valid_options++; + limit=info.r->process_to_value(*vlimit).as_int(); + } + if(valid_options!=info.options->size()) + throw Exception("parser.runtime", + &file_spec, + "load called with invalid option"); + } + + int m=finfo.st_size; + if(offset<0 || offset>=m || limit==0) { // bad options/empty file: returning empty file if(as_text) { *info.data=pool.malloc(1); *(char*)(*info.data)=0; @@ -386,24 +422,53 @@ static void file_read_action(Pool& pool, *info.data_size=0; return; } + if(limit<0 || limit>m) + limit=m; + + if(offset) + if(lseek(f, offset, SEEK_SET)<0) + throw Exception("file.read", + &file_spec, + "seek to %d position failed: %s (%d), actual filename '%s'", + offset, strerror(errno), errno, fname); + + *info.data=pool.malloc(limit+(as_text?1:0), 3); + *info.data_size=(size_t)read(f, *info.data, limit); + + if((ssize_t)*info.data_size<0) + throw Exception(0, + &file_spec, + "failed to read %d bytes from %d offset of %d size (stat)", // may be in text mode... + limit, offset, m); } bool file_read(Pool& pool, const String& file_spec, void*& data, size_t& data_size, - bool as_text, Hash *params, Hash** out_fields, + bool as_text, + Hash* options, Request* r, + Hash** out_fields, bool fail_on_read_problem) { bool result; if(file_spec.starts_with("http://", 7)) { // fail on read problem - file_read_http(pool, file_spec, data, data_size, params, out_fields); + file_read_http(pool, file_spec, data, data_size, options, out_fields); result=true; } else { - File_read_action_info info={&data, &data_size}; + File_read_action_info info={&data, &data_size, options, r}; result=file_read_action_under_lock(pool, file_spec, "read", file_read_action, &info, as_text, fail_on_read_problem); } if(result && 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 + } + } + fix_line_breaks((char *)(data), data_size); // note: after fixing ((char*&)(data))[data_size]=0; @@ -414,8 +479,7 @@ bool file_read(Pool& pool, const String& 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, - bool fail_on_read_problem) { + bool as_text, bool fail_on_read_problem) { const char* fname=file_spec.cstr(String::UL_FILE_SPEC); int f; @@ -443,11 +507,12 @@ 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()) { +#ifdef PA_SAFE_MODE + if(finfo.st_uid/*foreign?*/!=geteuid() + && 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'", + "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); #endif @@ -495,7 +560,7 @@ 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, @@ -510,14 +575,18 @@ 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 @@ -914,3 +983,76 @@ int pa_sleep(unsigned long secs, unsigne 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; +}