File:  [parser3project] / parser3 / src / main / pa_common.C
Revision 1.143.2.21.2.19: download - view: text, annotated - select for diffs - revision graph
Tue Apr 1 16:16:23 2003 UTC (23 years, 3 months ago) by paf
no message

/** @file
	Parser: commonly functions.

	Copyright(c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
*/

static const char* IDENT_COMMON_C="$Date: 2003/04/01 16:16:23 $"; 

#include "pa_common.h"
#include "pa_exception.h"
#include "pa_hash.h"
#include "pa_globals.h"

#define PA_HTTP

#ifdef PA_HTTP
#include "pa_vstring.h"
#include "pa_vint.h"

#ifdef CYGWIN
#define _GNU_H_WINDOWS32_SOCKETS
// for PASCAL
#include <windows.h>
// SOCKET
typedef u_int	SOCKET;
int PASCAL closesocket(SOCKET);
#else
#	if defined(WIN32)
#		include <windows.h>
#	else
#		define closesocket close
#	endif
#endif

#else

#	if defined(WIN32)
#		include <windows.h>
#	endif

#endif

// some maybe-undefined constants

#ifndef _O_TEXT
#	define _O_TEXT 0
#endif
#ifndef _O_BINARY
#	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
//#define PA_DEBUG_NO_LOCKING

#ifdef PA_DEBUG_NO_LOCKING

#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

#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); }

#else
#ifdef HAVE_FCNTL

#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); }

#else
#ifdef HAVE_LOCKF

#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); }

#else

#error unable to find file locking func

#endif
#endif
#endif
#endif

#else
static int lock_shared_blocking(int fd) { return 0; }
static int lock_exclusive_blocking(int fd) { return 0; }
static int lock_exclusive_nonblocking(int fd) { return 0; }
static int unlock(int fd) { return 0; }

#endif

// defines for globals

#define FILE_STATUS_NAME  "status"

// globals

const String file_status_name(FILE_STATUS_NAME);

// defines for statics

#define HTTP_METHOD_NAME  "method"
#define HTTP_TIMEOUT_NAME    "timeout"
#define HTTP_HEADERS_NAME "headers"
#define HTTP_ANY_STATUS_NAME "any-status"

// statics

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);

// defines

#define DEFAULT_USER_AGENT "parser3"

// functions

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=(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]<eob && eol[1]=='\n') { // \r, \n = DOS
			bol=eol+2;
			size--; 
		} else // \r, not \n = Macintosh
			bol=eol+1;
	}
	// last piece without \r, including terminating 0
	if(dest!=bol)
		memcpy(dest, bol, eob-bol); 
}

char* file_read_text(Charset& charset, 
					 const String& file_spec, 
					 bool fail_on_read_problem,
					 HashStringValue* params/*, HashStringValue* * out_fields*/) {
	File_read_result file=
		file_read(charset, file_spec, true, params, fail_on_read_problem);
	return file.success?file.str:0;
}

#ifdef PA_HTTP
//http request stuff

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) {
		if(struct hostent *hostIP=gethostbyname(host)) 
			memcpy(&addr->sin_addr, hostIP->h_addr, hostIP->h_length); 
		else
			return false;
    } else 
		addr->sin_addr.s_addr=INADDR_ANY;
    return true;
}

static int http_read_response(String& response, int sock, bool fail_on_status_ne_200){
	int result=0;
	size_t EOLat=0;
	while(true) {
		char *buf=new(PointerFreeGC) char[MAX_STRING]; 
		ssize_t size=recv(sock, buf, MAX_STRING, 0); 
		if(size<=0)
			break;
		response.append_strdup(buf, size, String::L_TAINTED); 
		if(!result && (EOLat=response.pos("\r\n", 2))!=STRING_NOT_FOUND) { // checking status in first response
			const String& status_line=response.mid(0, (size_t)EOLat);
			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 && result!=200)
				throw Exception("http.status",
					&status_code,
					"invalid HTTP response status");
		}
	}
	if(result)
		return result;
	else
		throw Exception("http.response",
			0,
			"bad response from host - no status found (size=%lu)", response.length()); 
}

/* ********************** request *************************** */

#if defined(SIGALRM) && defined(HAVE_SIGSETJMP) && defined(HAVE_SIGLONGJMP)
#	define WE_CAN_USE_ALARM
#endif

#ifdef WE_CAN_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 char* host, int port, 
			const char* request, 
			int timeout,
			bool fail_on_status_ne_200) {
	if(!host)
		throw Exception("http.host", 
			0, 
			"zero hostname");  //never

#ifdef WE_CAN_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", 
				0, 
				"timeout occured while retrieving document"); 
		else {
			alarm(timeout); 
#endif
			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)
				throw Exception("http.connect", 
					0, 
					"can not make socket: %s (%d)", strerror(errno), errno); 
			if(connect(sock, (struct sockaddr *)&dest, sizeof(dest)))
				throw Exception("http.connect", 
					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", 
					0, 
					"error sending request: %s (%d)", strerror(errno), errno); 

			result=http_read_response(response, sock, fail_on_status_ne_200); 
			closesocket(sock); 
#ifdef WE_CAN_USE_ALARM
			alarm(0); 
		}
#endif
		return result;
	} catch(...) {
		if(sock>=0) 
			closesocket(sock); 
#ifdef WE_CAN_USE_ALARM
		alarm(0); 
#endif
		rethrow;
	}
}

#ifndef DOXYGEN
struct Http_pass_header_info {
	Charset* charset;
	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 <<key<<": "
		<< attributed_meaning_to_string(*value, String::L_HTTP_HEADER, false)
		<< "\n"; 
	
    if(String(key, String::L_TAINTED).change_case(*info->charset, String::CC_UPPER)=="USER-AGENT")
		info->user_agent_specified=true;
}
#ifndef DOXYGEN
struct File_read_http_result {
	char *str; size_t length;
	HashStringValue* headers;
}; 
#endif
static File_read_http_result file_read_http(Charset& charset, const String& file_spec, 
					    HashStringValue *options=0) {
	File_read_http_result result;
	char host[MAX_STRING]; 
	const char* uri; 
	int port;
	const char* method=0;
	int timeout=2;
	bool fail_on_status_ne_200=true;
	Value* vheaders=0;

	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* 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;

	if(options) {
		int valid_options=0;
		if(Value* vmethod=options->get(http_method_name)) {
			valid_options++;
			method=vmethod->as_string().cstr();
		}
		if(Value* vtimeout=options->get(http_timeout_name)) {
			valid_options++;
			timeout=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(valid_options!=options->count())
			throw Exception("parser.runtime",
				0,
				"invalid option passed");
	} 

	//making request
	String request;
	if(method)
		request<<method;
	else
		request<<"GET";
	request<< " "<< uri <<" HTTP/1.0\nHost: "<< host<<"\n"; 
	bool user_agent_specified=false;
	if(vheaders && !vheaders->is_string()) { // allow empty
		if(HashStringValue *headers=vheaders->get_hash()) {
			Http_pass_header_info info={&charset, &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"; 
	
	//sending request
	String response;
	const char* request_cstr=request.cstr(String::L_UNSPECIFIED);
	int status_code=http_request(response,
		host, port, request_cstr, 
		timeout, fail_on_status_ne_200); 
	
	//processing results	
	size_t pos=response.pos("\r\n\r\n", 4); 
	if(pos==STRING_NOT_FOUND || pos<1){
		throw Exception("http.response", 
			&connect_string,
			"bad response from host - no headers found"); 
	}
	const String& header_block=response.mid(0, pos); 
	const String& body=response.mid(pos+4, response.length()); 
	
	ArrayString aheaders;
	result.headers=new HashStringValue;
	size_t pos_after=0;
	header_block.split(aheaders, pos_after, "\r\n"); 
	
	//processing headers
	size_t aheaders_count=aheaders.count();
	for(size_t i=1; i<aheaders_count; i++) {
		const String& line=*aheaders.get(i);
		pos=line.pos(": ", 2); 
		if(pos==STRING_NOT_FOUND || pos<1)
			throw Exception("http.response", 
				&connect_string,
				"bad response from host - bad header \"%s\"", line.cstr());
			
		result.headers->put(
			line.mid(0, pos).change_case(charset, String::CC_UPPER), 
			new VString(line.mid(pos+2, line.length())));
	}

	// output response
	result.str=body.cstrm();
	result.length=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;
}; 
#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<File_read_action_info *>(context); 
	if(size_t to_read_size=(size_t)finfo.st_size) { 
		*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", 
					*info.data_size, to_read_size); 
	} else { // empty file
		if(as_text) {
			*info.data=new(PointerFreeGC) char[1]; 
			*(char*)(*info.data)=0;
		} else 
			*info.data=0;
		*info.data_size=0;
		return;
	}
}
File_read_result file_read(Charset& charset, const String& file_spec, 
			   bool as_text, HashStringValue *params,
			   bool fail_on_read_problem) {
	File_read_result result={false};
#ifdef PA_HTTP
	if(file_spec.starts_with("http://")) {
		// fail on read problem
		File_read_http_result http;//=file_read_http(charset, file_spec, params);
		result.success=true;
		result.str="HELLO";//http.str;
		result.length=5;//http.length;
		result.headers=new HashStringValue;//http.headers; 
	} else {
#endif
		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.success && as_text) {
		// UTF-8 signature: EF BB BF
		if(result.length>=3) {
			char *in=(char *)result.str;
			if((in[0] == '\xEF') && (in[1] == '\xBB') &&
				(in[2] == '\xBF')) {
				result.str=in+3; result.length-=3;// skip prefix
			}
		}

		fix_line_breaks((char *)(result.str), result.length); 
		// note: after fixing
		((char*&)(result.str))[result.length]=0;
	}

	return result;
}

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); 
	int f;

	// first open, next stat:
	// directory update of NTFS hard links performed on open.
	// ex: 
	//   a.html:^test[] and b.html hardlink to a.html
	//   user inserts ! before ^test in a.html
	//   directory entry of b.html in NTFS not updated at once, 
	//   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) {
		try {
			if(lock_shared_blocking(f)!=0)
				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)
				throw Exception("file.missing", // hardly possible: we just opened it OK
					&file_spec, 
					"stat failed: %s (%d), actual filename '%s'", 
						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);
#endif

			action(finfo, f, file_spec, fname, as_text, context); 
		} catch(...) {
			unlock(f);close(f); 
			if(fail_on_read_problem)
				rethrow;
			return false;			
		} 

		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);
		return false;
	}
}

static void create_dir_for_file(const String& file_spec) {
	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); 
		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, 
				bool as_text, 
				bool do_append, 
				bool do_block, 
				bool fail_on_lock_problem) {
	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); 

	if((f=open(fname, 
		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) {
			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;
			return false;
		}

		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); 
			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 <virtan@rotabanner.com> for discovering]
#endif
		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);
	// here should be nothing, see rethrow above
}

#ifndef DOXYGEN
struct File_write_action_info {
	const char* str; size_t length;
}; 
#endif
static void file_write_action(int f, void *context) {
	File_write_action_info& info=*static_cast<File_write_action_info *>(context); 
	if(info.length) {
		int written=write(f, info.str, info.length); 
		if(written<0)
			throw Exception(0, 
				0, 
				"write failed: %s (%d)",  strerror(errno), errno); 
	}
}
void file_write(
				const String& file_spec, 
				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); 
}

// 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)); 
}
bool file_delete(const String& file_spec, bool fail_on_read_problem) {
	const char* fname=file_spec.cstr(String::L_FILE_SPEC); 
	if(unlink(fname)!=0)
		if(fail_on_read_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);
		else
			return false;

	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); 
	
	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);

	rmdir(old_spec, 1); 
}


bool entry_exists(const char* fname, struct stat *afinfo) {
	struct stat lfinfo;
	bool result=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); 
	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) {
	return entry_readable(file_spec, false); 
}
bool dir_readable(const String& file_spec) {
	return entry_readable(file_spec, true); 
}
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::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); 
	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);
		else
			return false;
	rsize=finfo.st_size;
	ratime=finfo.st_atime;
	rmtime=finfo.st_mtime;
	rctime=finfo.st_ctime;
	return true;
}

char* getrow(char* *row_ref, char delim) {
    char* result=*row_ref;
    if(result) {
		*row_ref=strchr(result, delim); 
		if(*row_ref) 
			*((*row_ref)++)=0; 
		else if(!*result) 
			return 0;
    }
    return result;
}

char* lsplit(char* string, char delim) {
    if(string) {
		char* v=strchr(string, delim); 
		if(v) {
			*v=0;
			return v+1;
		}
    }
    return 0;
}

char* lsplit(char* *string_ref, char delim) {
    char* result=*string_ref;
	char* next=lsplit(*string_ref, delim); 
    *string_ref=next;
    return result;
}

char* rsplit(char* string, char delim) {
    if(string) {
		char* v=strrchr(string, delim); 
		if(v) {
			*v=0;
			return v+1;
		}
    }
    return NULL;	
}

/// @todo less stupid type detection
const char* format(double value, char* fmt) {
	char local_buf[MAX_NUMBER]; 
	size_t size;
	
	if(fmt)
		if(strpbrk(fmt, "diouxX"))
			if(strpbrk(fmt, "ouxX"))
				size=snprintf(local_buf, sizeof(local_buf), fmt, (uint)value); 
			else
				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);
}

size_t stdout_write(const void *buf, size_t size) {
#ifdef WIN32
	do{
		int chunk_written=fwrite(buf, 1, min((size_t)8*0x400, size), stdout); 
		if(chunk_written<=0)
			break;
		size-=chunk_written;
		buf=((const char*)buf)+chunk_written;
	} while(size>0); 

	return size;
#else
	return fwrite(buf, 1, size, stdout); 
#endif
}

char* unescape_chars(const char* cp, int len) {
	char* s=new(PointerFreeGC) char[len + 1]; 
	enum EscapeState {
		EscapeRest, 
		EscapeFirst, 
		EscapeSecond
	} escapeState=EscapeRest;
	int escapedValue=0;
	int srcPos=0;
	int dstPos=0;
	while(srcPos < len) {
		int ch=cp[srcPos]; 
		switch(escapeState) {
			case EscapeRest:
			if(ch=='%') {
				escapeState=EscapeFirst;
			} else if(ch=='+') {
				s[dstPos++]=' '; 
			} else {
				s[dstPos++]=ch;	
			}
			break;
			case EscapeFirst:
			escapedValue=hex_value[ch] << 4;	
			escapeState=EscapeSecond;
			break;
			case EscapeSecond:
			escapedValue +=hex_value[ch]; 
			s[dstPos++]=escapedValue;
			escapeState=EscapeRest;
			break;
		}
		srcPos++; 
	}
	s[dstPos]=0;
	return s;
}

#ifdef WIN32
void back_slashes_to_slashes(char* s) {
	if(s)
		for(; *s; 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(*s1)) {
			if(tolower(*s1) !=tolower(*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; p<end; p++)
		switch(*p) {
			case '\n': *p='|';  break;
			case '\r': *p=' ';  break;
		}
}


/// must be last in this file
#undef vsnprintf
int __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
	--s;
#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(r>s)
		r=s;
#endif
	b[r]=0;
	return r;
}

int __snprintf(char* b, size_t s, const char* f, ...) {
	va_list l;
    va_start(l, f); 
    int r=__vsnprintf(b, s, f, l); 
    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
}



E-mail: