--- parser3/src/main/pa_common.C 2001/08/27 13:29:07 1.62 +++ parser3/src/main/pa_common.C 2002/11/21 12:47:23 1.124 @@ -1,62 +1,168 @@ /** @file Parser: commonly functions. - Copyright(c) 2001 ArtLebedev Group(http://www.artlebedev.com) - - Author: Alexander Petrosyan (http://design.ru/paf) + Copyright(c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Author: Alexandr Petrosian (http://paf.design.ru) */ -static const char *RCSId="$Id: pa_common.C,v 1.62 2001/08/27 13:29:07 parser Exp $"; + +static const char* IDENT_COMMON_C="$Date: 2002/11/21 12:47:23 $"; #include "pa_common.h" -#include "pa_types.h" #include "pa_exception.h" -#include "pa_pool.h" #include "pa_globals.h" -#include "pa_value.h" -#include "pa_hash.h" -#include "pa_string.h" - -#ifndef WIN32 -# ifndef _O_TEXT -# define _O_TEXT 0 -# endif -# ifndef _O_BINARY -# define _O_BINARY 0 -# endif + +#ifdef WIN32 +# include #endif -#if _MSC_VER +// some maybe-undefined constants -int __vsnprintf(char *b, size_t s, const char *f, va_list l) { - int r=_vsnprintf(b, --s, f, l); - b[s]=0; - return r; +#ifndef _O_TEXT +# define _O_TEXT 0 +#endif +#ifndef _O_BINARY +# define _O_BINARY 0 +#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 + +#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 + +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; } -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; + +void fix_line_breaks(char *buf, size_t& size) { + //_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')) { + size_t len=eol-bol; + if(dest!=bol) + memcpy(dest, bol, len); + dest+=len; + *dest++='\n'; + + if(&eol[1](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 + if(as_text) { + *info.data=pool.malloc(1); + *(char*)(*info.data)=0; + } else + *info.data=0; + *info.data_size=0; + return; + } -char *file_read_text(Pool& pool, const String& file_spec, bool fail_on_read_problem) { - void *result; - size_t size; - return file_read(pool, file_spec, result, size, true, - fail_on_read_problem)?(char *)result:0; + if(as_text) { + fix_line_breaks((char *)(*info.data), *info.data_size); + // note: after fixing + ((char*&)(*info.data))[*info.data_size]=0; + } } bool file_read(Pool& pool, const String& file_spec, - void*& data, size_t& read_size, bool as_text, - bool fail_on_read_problem, - size_t offset, size_t limit) { - const char *fname=file_spec.cstr(String::UL_FILE_NAME); -//printf("file_read(%s)\n", fname); + void*& data, size_t& data_size, bool as_text, + bool fail_on_read_problem) { + File_read_action_info info={&data, &data_size}; + return file_read_action_under_lock(pool, file_spec, + "read", file_read_action, &info, + as_text, fail_on_read_problem); +} + +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) { + const char *fname=file_spec.cstr(String::UL_FILE_SPEC); int f; - struct stat finfo; // first open, next stat: // directory update of NTFS hard links performed on open. @@ -66,123 +172,201 @@ bool file_read(Pool& pool, const String& // 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. - if( - (f=open(fname, O_RDONLY|(as_text?_O_TEXT:_O_BINARY)))>=0 && - stat(fname, &finfo)==0) { - /*if(exclusive) - flock(f, LOCK_EX);*/ - size_t max_size=limit?min(offset+limit, finfo.st_size)-offset:finfo.st_size; - if(!max_size) { // eof - data=0; - read_size=0; - } else { - data=pool.malloc(max_size+(as_text?1:0)); - if(offset) - lseek(f, offset, SEEK_SET); - read_size=read(f, data, max_size); - } - /*if(exclusive) - flock(f, LOCK_UN);*/ - close(f); - if(!max_size) // eof - return true; - - if(read_size>=0 && read_size<=max_size) { - if(as_text) - ((char*&)data)[read_size]=0; - } else - PTHROW(0, 0, + // 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 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); +#endif + + action(pool, finfo, f, file_spec, fname, as_text, context); + } catch(...) { + unlock(f);close(f); + if(fail_on_read_problem) + /*re*/throw; + 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, - "read failed: actually read %d bytes count not in [0..%lu] valid range", - read_size, (unsigned long)max_size); //never - - return true;//prepare_config(result, remove_empty_lines); - } - if(fail_on_read_problem) - PTHROW(0, 0, - &file_spec, - "read failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); - return false; + "%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; + 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); + pos_after=pos_before+1; + } } -void file_write(Pool& pool, +bool file_write_action_under_lock( const String& file_spec, - const void *data, size_t size, - bool as_text/*, - bool exclusive*/) { - const char *fname=file_spec.cstr(String::UL_FILE_NAME); + 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::UL_FILE_SPEC); int f; - if(access(fname, W_OK)!=0) {/*no*/ - 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_NAME), 0775); - pos_after=pos_before+1; - } + if(access(fname, W_OK)!=0) // no + create_dir_for_file(file_spec); - if((f=open(fname, O_WRONLY|O_CREAT|_O_BINARY, 0666))>0) - close(f); - } - if(access(fname, R_OK|W_OK)==0) { - int mode=O_RDWR|(as_text?_O_TEXT:_O_BINARY) -#ifdef WIN32 - |O_TRUNC -#endif - ; - if((f=open(fname, mode, 0666))>=0) { - /*if(exclusive) - flock(f, LOCK_EX);*/ - - if(size) write(f, data, size); -#ifndef WIN32 - ftruncate(f, size); -#endif - /*if(exclusive) - flock(f, LOCK_UN);*/ + if((f=open(fname, + O_CREAT|O_RDWR + |(as_text?_O_TEXT:_O_BINARY) + |(do_append?O_APPEND:0), 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); - return; + if(fail_on_lock_problem) + throw e; + return false; + } + + try { + action(f, context); + } catch(...) { + if(!do_append) + ftruncate(f, tell(f)); // one can not use O_TRUNC, read lower + unlock(f);close(f); + /*re*/throw; } + + if(!do_append) + ftruncate(f, tell(f)); // O_TRUNC truncates even exclusevely write-locked file [thanks to Igor Milyakov for discovering] + 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 void *data; size_t size; +}; +#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(written<0) + throw Exception(0, + 0, + "write failed: %s (%d)", strerror(errno), errno); } - PTHROW(0, 0, - &file_spec, - "write failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); +} +void file_write( + const String& file_spec, + const void *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) { int pos_before; if((pos_before=file_spec.pos("/", 1, pos_after))>=0) rmdir(file_spec, pos_before+1); - rmdir(file_spec.mid(0, pos_after-1/* / */).cstr(String::UL_FILE_NAME)); + rmdir(file_spec.mid(0, pos_after-1/* / */).cstr(String::UL_FILE_SPEC)); } -void file_delete(Pool& pool, const String& file_spec) { - const char *fname=file_spec.cstr(String::UL_FILE_NAME); +bool file_delete(const String& file_spec, bool fail_on_read_problem) { + const char *fname=file_spec.cstr(String::UL_FILE_SPEC); if(unlink(fname)!=0) - PTHROW(0, 0, - &file_spec, - "unlink failed: %s (%d), actual filename '%s'", - strerror(errno), errno, fname); + 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(Pool& pool, const String& old_spec, const String& new_spec) { - const char *old_spec_cstr=old_spec.cstr(String::UL_FILE_NAME); - const char *new_spec_cstr=new_spec.cstr(String::UL_FILE_NAME); +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); + + create_dir_for_file(new_spec); + if(rename(old_spec_cstr, new_spec_cstr)!=0) - PTHROW(0, 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::UL_FILE_SPEC); + return entry_exists(fname, 0); +} + static bool entry_readable(const String& file_spec, bool need_dir) { - const char *fname=file_spec.cstr(String::UL_FILE_NAME); + char *fname=file_spec.cstr(String::UL_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 && stat(fname, &finfo)==0) { - bool is_dir=(bool)(finfo.st_mode&S_IFDIR); + 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; @@ -193,27 +377,38 @@ 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; +} bool file_executable(const String& file_spec) { - return access(file_spec.cstr(String::UL_FILE_NAME), X_OK)==0; + return access(file_spec.cstr(String::UL_FILE_SPEC), X_OK)==0; } -void file_stat(const String& file_spec, +bool file_stat(const String& file_spec, size_t& rsize, time_t& ratime, time_t& rmtime, - time_t& rctime) { + time_t& rctime, + bool fail_on_read_problem) { Pool& pool=file_spec.pool(); - const char *fname=file_spec.cstr(String::UL_FILE_NAME); + const char *fname=file_spec.cstr(String::UL_FILE_SPEC); struct stat finfo; if(stat(fname, &finfo)!=0) - PTHROW(0, 0, - &file_spec, - "getting file size failed: %s (%d), real filename '%s'", - strerror(errno), errno, fname); + 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) { @@ -259,19 +454,23 @@ char *rsplit(char *string, char delim) { /// @todo less stupid type detection char *format(Pool& pool, double value, char *fmt) { - char *result=(char *)pool.malloc(MAX_NUMBER); + char local_buf[MAX_NUMBER]; + size_t size; + if(fmt) if(strpbrk(fmt, "diouxX")) if(strpbrk(fmt, "ouxX")) - snprintf(result, MAX_NUMBER, fmt, (uint)value ); + size=snprintf(local_buf, sizeof(local_buf), fmt, (uint)value); else - snprintf(result, MAX_NUMBER, fmt, (int)value ); + size=snprintf(local_buf, sizeof(local_buf), fmt, (int)value); else - snprintf(result, MAX_NUMBER, fmt, value); + size=snprintf(local_buf, sizeof(local_buf), fmt, value); else - snprintf(result, MAX_NUMBER, "%d", (int)value); + size=snprintf(local_buf, sizeof(local_buf), "%d", (int)value); - return result; + char *pool_buf=(char *)pool.malloc(size+1, 4); + memcpy(pool_buf, local_buf, size+1); + return pool_buf; } size_t stdout_write(const void *buf, size_t size) { @@ -290,8 +489,8 @@ size_t stdout_write(const void *buf, siz #endif } -const char *unescape_chars(Pool& pool, const char *cp, int len) { - char *s=(char *)pool.malloc(len + 1); +char *unescape_chars(Pool& pool, const char *cp, int len) { + char *s=(char *)pool.malloc(len + 1, 5); enum EscapeState { EscapeRest, EscapeFirst, @@ -328,43 +527,6 @@ const char *unescape_chars(Pool& pool, c return s; } -/// used by attributed_meaning_to_string / append_attribute_subattribute -struct Attributed_meaning_info { - String *header; // header line being constructed - String::Untaint_lang lang; // language in which to append to that line -}; -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, true); - *ami.header << "="; - ami.header->append(static_cast(avalue)->as_string(), ami.lang, true); -} -const String& attributed_meaning_to_string(Value& meaning, - String::Untaint_lang lang) { - String &result=*new(meaning.pool()) String(meaning.pool()); - if(Hash *hash=meaning.get_hash()) { - // $value(value) $subattribute(subattribute value) - if(Value *value=static_cast(hash->get(*value_name))) - result.append(value->as_string(), lang, true); - - Attributed_meaning_info attributed_meaning_info={ - &result, - lang - }; - hash->for_each(append_attribute_subattribute, &attributed_meaning_info); - } else // result value - result.append(meaning.as_string(), lang, true); - - return result; -} - #ifdef WIN32 void back_slashes_to_slashes(char *s) { if(s) @@ -401,44 +563,7 @@ bool StrEqNc(const char *s1, const char } } -char *unquote(char*& current, char stop_at) { - char *result=current; - char *dest=current; - // skip leading WS - while(*current==' ' || *current=='\t') - current++; - if(!*current) - return current=0; - - bool quoted=*current=='"'; - if(quoted) - current++; - - for(; *current; ) { - if(quoted) - switch(*current) { - case '\\': // "...\c - if(current[1]) - current++; - break; - case '"': - current++; - goto break2; - } - else - if(*current==stop_at) - break; - - *dest++=*current++; - } -break2: - if(*current) - current++; // skip 'stop_at' - *dest=0; - return result; -} - -static int isLeap(int year) { +static bool isLeap(int year) { return !( (year % 4) || ((year % 400) && !(year % 100)) ); @@ -461,3 +586,76 @@ int getMonthDays(int year, int month) { }; return monthDays[month]; } + +void remove_crlf(char *start, char *end) { + for(char *p=start; ps) + 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 +}