--- parser3/src/main/pa_common.C 2005/11/22 11:31:58 1.211 +++ parser3/src/main/pa_common.C 2005/12/09 07:18:07 1.216 @@ -26,46 +26,14 @@ * */ -static const char * const IDENT_COMMON_C="$Date: 2005/11/22 11:31:58 $"; +static const char * const IDENT_COMMON_C="$Date: 2005/12/09 07:18:07 $"; #include "pa_common.h" #include "pa_exception.h" #include "pa_hash.h" #include "pa_globals.h" -#include "pa_request_charsets.h" #include "pa_charsets.h" - -#define PA_HTTP - -#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 -#include -// SOCKET -typedef u_int SOCKET; -int PASCAL closesocket(SOCKET); -#else -# if defined(WIN32) -# include -# else -# define closesocket close -# endif -#endif - -#else - -# if defined(WIN32) -# include -# endif - -#endif +#include "pa_http.h" // some maybe-undefined constants @@ -86,10 +54,6 @@ int PASCAL closesocket(SOCKET); # endif #endif -# ifndef INADDR_NONE -# define INADDR_NONE ((ulong) -1) -# endif - // defines for globals #define FILE_STATUS_NAME "status" @@ -98,23 +62,6 @@ 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) { @@ -153,7 +100,7 @@ char* file_read_text(Request_charsets& c } /// these options were handled but not checked elsewhere, now check them -static int get_valid_file_options_count(HashStringValue& options) +int pa_get_valid_file_options_count(HashStringValue& options) { int result=0; if(options.get(PA_SQL_LIMIT_NAME)) @@ -167,640 +114,6 @@ static int get_valid_file_options_count( return result; } -//http request stuff -#ifdef PA_HTTP - -#undef CRLF -#define CRLF "\r\n" - -static bool set_addr(struct sockaddr_in *addr, const char* host, const short port){ - memset(addr, 0, sizeof(*addr)); - addr->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; -}; -#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); - } - - 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; - } - } - 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); - - head << "content-length: " << format(strlen(body_cstr), "%u") << CRLF; - } - - const char* head_cstr=head.cstr(String::L_UNSPECIFIED); - - // 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(); - - 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.headers->put(HEADER_NAME, new VString(header_value)); - } - } - - // 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 - #ifndef DOXYGEN struct File_read_action_info { char **data; size_t *data_size; @@ -843,7 +156,6 @@ File_read_result file_read(Request_chars 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", @@ -851,15 +163,14 @@ File_read_result file_read(Request_chars "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(charsets, file_spec, as_text, params); 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); + int valid_options=pa_get_valid_file_options_count(*params); if(valid_options!=params->count()) throw Exception("parser.runtime", 0, @@ -871,9 +182,7 @@ File_read_result file_read(Request_chars 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 @@ -1121,17 +430,17 @@ static bool entry_readable(const String& } 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 << "/"; 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; @@ -1244,7 +553,7 @@ char* unescape_chars(const char* cp, int EscapeFirst, EscapeSecond } escapeState=EscapeRest; - uchar escapedValue=0; + uint escapedValue=0; int srcPos=0; int dstPos=0; while(srcPos < len) { @@ -1260,12 +569,12 @@ char* unescape_chars(const char* cp, int } break; case EscapeFirst: - escapedValue=(uchar)(hex_value[ch] << 4); + escapedValue=hex_value[ch] << 4; escapeState=EscapeSecond; break; case EscapeSecond: escapedValue +=hex_value[ch]; - s[dstPos++]=escapedValue; + s[dstPos++]=(char)escapedValue; escapeState=EscapeRest; break; } @@ -1500,7 +809,7 @@ 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", @@ -1593,10 +902,10 @@ static unsigned char gmime_base64_rank[2 size_t g_mime_utils_base64_decode_step (const unsigned char *in, size_t inlen, unsigned char *out, int *state, int *save) { - 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; @@ -1613,9 +922,9 @@ g_mime_utils_base64_decode_step (const u saved = (saved << 6) | c; i++; if (i == 4) { - *outptr++ = saved >> 16; - *outptr++ = saved >> 8; - *outptr++ = saved; + *outptr++ = (unsigned char)saved >> 16; + *outptr++ = (unsigned char)saved >> 8; + *outptr++ = (unsigned char)saved; i = 0; } }