/** \file common.c Various functions, mostly string utilities, that are used by most parts of fish. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef HOST_NAME_MAX #define HOST_NAME_MAX 255 #endif #if HAVE_NCURSES_H #include #else #include #endif #if HAVE_TERMIO_H #include #endif #include #include "config.h" #include "util.h" #include "wutil.h" #include "common.h" #include "expand.h" #include "proc.h" #include "wildcard.h" #include "parser.h" /** Error message to show on string convertion error */ #define STR2WCS_MSG "fish: Invalid multibyte sequence \'" /** The maximum number of minor errors to report. Further errors will be omitted. */ #define ERROR_MAX_COUNT 1 /** The number of milliseconds to wait between polls when attempting to acquire a lockfile */ #define LOCKPOLLINTERVAL 10 struct termios shell_modes; /** Number of error encountered. This is reset after each command, and used to limit the number of error messages on commands with many string convertion problems. */ static int error_count=0; int error_max=1; wchar_t ellipsis_char; static int c1=0, c2=0, c3=0, c4=0, c5; char *profile=0; wchar_t *program_name; int debug_level=1; static int block_count=0; void common_destroy() { debug( 3, L"Calls: wcsdupcat %d, wcsdupcat2 %d, wcsndup %d, str2wcs %d, wcs2str %d", c1, c2, c3, c4, c5 ); } wchar_t **list_to_char_arr( array_list_t *l ) { wchar_t ** res = malloc( sizeof(wchar_t *)*(al_get_count( l )+1) ); int i; if( res == 0 ) { die_mem(); } for( i=0; i= *len ) { int new_len = maxi( 128, (*len)*2); buff = realloc( buff, sizeof(wchar_t)*new_len ); if( buff == 0 ) { die_mem(); } else { *len = new_len; *b = buff; } } errno=0; c = getwc( f ); if( errno == EILSEQ ) { getc( f ); continue; } //fwprintf( stderr, L"b\n" ); switch( c ) { /* End of line */ case WEOF: case L'\n': case L'\0': buff[i]=L'\0'; unblock(); return i; /* Ignore carriage returns */ case L'\r': break; default: buff[i++]=c; break; } } unblock(); } /** Wrapper for wcsfilecmp */ static int completion_cmp( const void *a, const void *b ) { wchar_t *c= *((wchar_t **)a); wchar_t *d= *((wchar_t **)b); return wcsfilecmp( c, d ); } void sort_list( array_list_t *comp ) { qsort( comp->arr, al_get_count( comp ), sizeof( void*), &completion_cmp ); } wchar_t *str2wcs( const char *in ) { c4++; wchar_t *res; res = malloc( sizeof(wchar_t)*(strlen(in)+1) ); if( !res ) { die_mem(); } if( (size_t)-1 == mbstowcs( res, in, sizeof(wchar_t)*(strlen(in)) +1) ) { error_count++; if( error_count <=error_max ) { fflush( stderr ); write( 2, STR2WCS_MSG, strlen(STR2WCS_MSG) ); write( 2, in, strlen(in )); write( 2, "\'\n", 2 ); } free(res); return 0; } return res; } void error_reset() { error_count=0; } char *wcs2str( const wchar_t *in ) { c5++; char *res = malloc( MAX_UTF8_BYTES*wcslen(in)+1 ); if( res == 0 ) { die_mem(); } wcstombs( res, in, MAX_UTF8_BYTES*wcslen(in)+1 ); res = realloc( res, strlen( res )+1 ); return res; } char **wcsv2strv( const wchar_t **in ) { int count =0; int i; while( in[count] != 0 ) count++; char **res = malloc( sizeof( char *)*(count+1)); if( res == 0 ) { die_mem(); } for( i=0; i= L'0') ) { res = d - L'0'; } else if( (d <= L'z') && (d >= L'a') ) { res = d + 10 - L'a'; } else if( (d <= L'Z') && (d >= L'A') ) { res = d + 10 - L'A'; } if( res >= base ) { res = -1; } return res; } long wcstol(const wchar_t *nptr, wchar_t **endptr, int base) { long long res=0; int is_set=0; if( base > 36 ) { errno = EINVAL; return 0; } while( 1 ) { long nxt = convert_digit( *nptr, base ); if( endptr != 0 ) *endptr = (wchar_t *)nptr; if( nxt < 0 ) { if( !is_set ) { errno = EINVAL; } return res; } res = (res*base)+nxt; is_set = 1; if( res > LONG_MAX ) { errno = ERANGE; return LONG_MAX; } if( res < LONG_MIN ) { errno = ERANGE; return LONG_MIN; } nptr++; } } /*$OpenBSD: strlcat.c,v 1.11 2003/06/17 21:56:24 millert Exp $*/ /* * Copyright (c) 1998 Todd C. Miller * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /** Appends src to string dst of size siz (unlike wcsncat, siz is the full size of dst, not space left). At most siz-1 characters will be copied. Always NUL terminates (unless siz <= wcslen(dst)). Returns wcslen(src) + MIN(siz, wcslen(initial dst)). If retval >= siz, truncation occurred. This is the OpenBSD strlcat function, modified for wide characters, and renamed to reflect this change. */ size_t wcslcat(wchar_t *dst, const wchar_t *src, size_t siz) { register wchar_t *d = dst; register const wchar_t *s = src; register size_t n = siz; size_t dlen; /* Find the end of dst and adjust bytes left but don't go past end */ while (n-- != 0 && *d != '\0') d++; dlen = d - dst; n = siz - dlen; if (n == 0) return(dlen + wcslen(s)); while (*s != '\0') { if (n != 1) { *d++ = *s; n--; } s++; } *d = '\0'; return(dlen + (s - src)); /* count does not include NUL */ } /*$OpenBSD: strlcpy.c,v 1.8 2003/06/17 21:56:24 millert Exp $*/ /* * Copyright (c) 1998 Todd C. Miller * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /** Copy src to string dst of size siz. At most siz-1 characters will be copied. Always NUL terminates (unless siz == 0). Returns wcslen(src); if retval >= siz, truncation occurred. This is the OpenBSD strlcpy function, modified for wide characters, and renamed to reflect this change. */ size_t wcslcpy(wchar_t *dst, const wchar_t *src, size_t siz) { register wchar_t *d = dst; register const wchar_t *s = src; register size_t n = siz; /* Copy as many bytes as will fit */ if (n != 0 && --n != 0) { do { if ((*d++ = *s++) == 0) break; } while (--n != 0); } /* Not enough room in dst, add NUL and traverse rest of src */ if (n == 0) { if (siz != 0) *d = '\0'; /* NUL-terminate dst */ while (*s++) ; } return(s - src - 1); /* count does not include NUL */ } wchar_t *wcsdup( const wchar_t *in ) { wchar_t *out = malloc( sizeof( wchar_t)*(wcslen(in)+1)); if( out == 0 ) { die_mem(); } wcscpy( out, in ); return out; } int wcscasecmp( const wchar_t *a, const wchar_t *b ) { if( *a == 0 ) { return (*b==0)?0:-1; } else if( *b == 0 ) { return 1; } int diff = towlower(*a)-towlower(*b); if( diff != 0 ) return diff; else return wcscasecmp( a+1,b+1); } int wcsncasecmp( const wchar_t *a, const wchar_t *b, int count ) { if( count == 0 ) return 0; if( *a == 0 ) { return (*b==0)?0:-1; } else if( *b == 0 ) { return 1; } int diff = towlower(*a)-towlower(*b); if( diff != 0 ) return diff; else return wcsncasecmp( a+1,b+1, count-1); } int wcsvarname( wchar_t *str ) { while( *str ) { if( (!iswalnum(*str)) && (*str != L'_' ) ) { return 0; } str++; } return 1; } #if !HAVE_WCWIDTH /** Return the number of columns used by a character. In locales without a native wcwidth, Unicode is probably so broken that it isn't worth trying to implement a real wcwidth. This wcwidth assumes any printing character takes up one column. */ int wcwidth( wchar_t c ) { if( c < 32 ) return 0; if ( c == 127 ) return 0; return 1; } #endif /** The glibc version of wcswidth seems to hang on some strings. fish uses this replacement. */ int my_wcswidth( const wchar_t *c ) { int res=0; while( *c ) { int w = wcwidth( *c++ ); if( w < 0 ) w = 1; if( w > 2 ) w=1; res += w; } return res; } wchar_t *quote_end( const wchar_t *in ) { int level=1; int offset = (*in != L'\"'); in++; while(1) { /* fwprintf( stderr, L"Check %c\n", *tok->buff );*/ switch( *in ) { case L'\\': in++; if( *in == L'\0' ) { return 0; } break; case L'\"': case L'\'': if( (((level+offset) % 2)?L'\"':L'\'') == *in ) { level--; } else { level++; } break; } if( (*in == L'\0') ||(level==0)) break; in++; } return level?0:(wchar_t *)in; } void fish_setlocale(int category, const wchar_t *locale) { char *lang = wcs2str( locale ); setlocale(category,lang); free( lang ); /* Use ellipsis if on known unicode system, otherwise use $ */ if( wcslen( locale ) ) { ellipsis_char = wcsstr( locale, L".UTF")?L'\u2026':L'$'; } else { char *lang = getenv( "LANG" ); if( lang ) ellipsis_char = strstr( lang, ".UTF")?L'\u2026':L'$'; else ellipsis_char = L'$'; } } int contains_str( const wchar_t *a, ... ) { wchar_t *arg; va_list va; int res = 0; va_start( va, a ); while( (arg=va_arg(va, wchar_t *) )!= 0 ) { if( wcscmp( a,arg) == 0 ) { res=1; break; } } va_end( va ); return res; } int read_blocked(int fd, void *buf, size_t count) { int res; sigset_t chldset, oldset; sigemptyset( &chldset ); sigaddset( &chldset, SIGCHLD ); sigprocmask(SIG_BLOCK, &chldset, &oldset); res = read( fd, buf, count ); sigprocmask( SIG_SETMASK, &oldset, 0 ); return res; } int writeb( tputs_arg_t b ) { write( 1, &b, 1 ); // putc( b, stdout ); return 0; } void die_mem() { /* int *foo=0; *foo = 6; */ debug( 0, L"Out of memory, shutting down fish." ); exit(1); } void debug( int level, wchar_t *msg, ... ) { va_list va; if( level > debug_level ) return; va_start( va, msg ); fwprintf( stderr, L"%ls: ", program_name ); vfwprintf( stderr, msg, va ); va_end( va ); fwprintf( stderr, L"\n" ); } wchar_t *escape( wchar_t *in, int escape_all ) { wchar_t *killme=in; wchar_t *out = malloc( sizeof(wchar_t)*(wcslen(in)*4 + 1)); if( out != 0 ) { wchar_t *pos=out; while( *in != 0 ) { switch( *in ) { case L'\t': *(pos++) = L'\\'; *(pos++) = L't'; break; case L'\n': *(pos++) = L'\\'; *(pos++) = L'n'; break; case L'\b': *(pos++) = L'\\'; *(pos++) = L'b'; break; case L'\r': *(pos++) = L'\\'; *(pos++) = L'r'; break; case L'\e': *(pos++) = L'\\'; *(pos++) = L'e'; break; case L'\\': case L'&': case L'$': case L' ': case L'#': case L'^': case L'<': case L'>': case L'@': case L'(': case L')': case L'{': case L'}': case L'?': case L'*': case L'|': case L';': case L':': case L'\'': case L'\"': if( escape_all ) *pos++ = L'\\'; *pos++ = *in; break; default: if( *in < 32 ) { int tmp = (*in)%16; *pos++ = L'\\'; *pos++ = L'x'; *pos++ = ((*in>15)? L'1' : L'0'); *pos++ = tmp > 9? L'a'+(tmp-10):L'0'+tmp; } else *pos++ = *in; break; } in++; } *pos = 0; free(killme); } return out; } wchar_t *unescape( wchar_t * in, int escape_special ) { int in_pos, out_pos, len = wcslen( in ); int c; int bracket_count=0; wchar_t prev=0; for( in_pos=0, out_pos=0; in_pos 0 && in[out_pos-1]==ANY_STRING ) { out_pos--; in[out_pos] = ANY_STRING_RECURSIVE; } else in[out_pos]=ANY_STRING; } else in[out_pos]=in[in_pos]; break; case L'?': if( escape_special ) in[out_pos]=ANY_CHAR; else in[out_pos]=in[in_pos]; break; case L'$': if( escape_special ) in[out_pos]=VARIABLE_EXPAND; else in[out_pos]=in[in_pos]; break; case L'{': if( escape_special ) { bracket_count++; in[out_pos]=BRACKET_BEGIN; } else in[out_pos]=in[in_pos]; break; case L'}': if( escape_special ) { bracket_count--; in[out_pos]=BRACKET_END; } else in[out_pos]=in[in_pos]; break; case L',': if( escape_special && bracket_count && prev!=BRACKET_SEP) { in[out_pos]=BRACKET_SEP; } else in[out_pos]=in[in_pos]; break; case L'\'': case L'\"': { wchar_t *end = quote_end( &in[in_pos] ); int len; if( end == 0 ) { error( SYNTAX_ERROR, L"Unexpected end of string", -1 ); return in; } len = end- &in[in_pos]-1; if( escape_special) in[out_pos++]=INTERNAL_SEPARATOR; memmove( &in[out_pos], &in[in_pos+1], sizeof(wchar_t)*(len) ); in_pos += len+1; out_pos += len-1; if( escape_special) in[++out_pos]=INTERNAL_SEPARATOR; break; } default: in[out_pos] = in[in_pos]; break; } } } in[out_pos]=L'\0'; return in; } /** Writes a pid_t in decimal representation to str. str must contain sufficient space. The conservatively approximate maximum number of characters a pid_t will represent is given by: (int)(0.31 * sizeof(pid_t) + CHAR_BIT + 1) Returns the length of the string */ static int sprint_pid_t( pid_t pid, char *str ) { int len, i = 0; int dig; /* Store digits in reverse order into string */ while( pid != 0 ) { dig = pid % 10; str[i] = '0' + dig; pid = ( pid - dig ) / 10; i++; } len = i; /* Reverse digits */ i /= 2; while( i ) { i--; dig = str[i]; str[i] = str[len - 1 - i]; str[len - 1 - i] = dig; } return len; } /** Writes a pseudo-random number (between one and maxlen) of pseudo-random digits into str. str must point to an allocated buffer of size of at least maxlen chars. Returns the number of digits written. Since the randomness in part depends on machine time it has _some_ extra strength but still not enough for use in concurrent locking schemes on a single machine because gettimeofday may not return a different value on consecutive calls when: a) the OS does not support fine enough resolution b) the OS is running on an SMP machine. Additionally, gettimeofday errors are ignored. Excludes chars other than digits since ANSI C only guarantees that digits are consecutive. */ static int sprint_rand_digits( char *str, int maxlen ) { int i, max; struct timeval tv; /* Seed the pseudo-random generator based on time - this assumes that consecutive calls to gettimeofday will return different values and ignores errors returned by gettimeofday. Cast to unsigned so that wrapping occurs on overflow as per ANSI C. */ (void)gettimeofday( &tv, NULL ); srand( (unsigned int)tv.tv_sec + (unsigned int)tv.tv_usec * 1000000UL ); max = 1 + (maxlen - 1) * (rand() / (RAND_MAX + 1.0)); for( i = 0; i < max; i++ ) { str[i] = '0' + 10 * (rand() / (RAND_MAX + 1.0)); } return i; } /** Generate a filename unique in an NFS namespace by creating a copy of str and appending .. to it. If gethostname() fails then a pseudo- random string is substituted for - the randomness of the string should be strong enough across different machines. The main assumption though is that gethostname will not fail and this is just a "safe enough" fallback. The memory returned should be freed using free(). */ static char *gen_unique_nfs_filename( const char *filename ) { int pidlen, hnlen, orglen = strlen( filename ); char hostname[HOST_NAME_MAX + 1]; char *newname; if ( gethostname( hostname, HOST_NAME_MAX + 1 ) == 0 ) { hnlen = strlen( hostname ); } else { hnlen = sprint_rand_digits( hostname, HOST_NAME_MAX ); hostname[hnlen] = '\0'; } newname = malloc( orglen + 1 /* period */ + hnlen + 1 /* period */ + /* max possible pid size: 0.31 ~= log(10)2 */ (int)(0.31 * sizeof(pid_t) * CHAR_BIT + 1) + 1 /* '\0' */ ); if ( newname == NULL ) { debug( 1, L"gen_unique_nfs_filename: %s", strerror( errno ) ); return newname; } memcpy( newname, filename, orglen ); newname[orglen] = '.'; memcpy( newname + orglen + 1, hostname, hnlen ); newname[orglen + 1 + hnlen] = '.'; pidlen = sprint_pid_t( getpid(), newname + orglen + 1 + hnlen + 1 ); newname[orglen + 1 + hnlen + 1 + pidlen] = '\0'; /* debug( 1, L"gen_unique_nfs_filename returning with: newname = \"%s\"; " L"HOST_NAME_MAX = %d; hnlen = %d; orglen = %d; " L"sizeof(pid_t) = %d; maxpiddigits = %d; malloc'd size: %d", newname, (int)HOST_NAME_MAX, hnlen, orglen, (int)sizeof(pid_t), (int)(0.31 * sizeof(pid_t) * CHAR_BIT + 1), (int)(orglen + 1 + hnlen + 1 + (int)(0.31 * sizeof(pid_t) * CHAR_BIT + 1) + 1) ); */ return newname; } /** Attempt to acquire a lock based on a lockfile, waiting LOCKPOLLINTERVAL milliseconds between polls and timing out after timeout seconds, thereafter forcibly attempting to obtain the lock if force is non-zero. Returns 1 on success, 0 on failure. To release the lock the lockfile must be unlinked. A unique temporary file named by appending characters to the lockfile name is used; any pre-existing file of the same name is subject to deletion. */ int acquire_lock_file( const char *lockfile, const int timeout, int force ) { int fd, timed_out = 0; int ret = 0; /* early exit returns failure */ struct timespec pollint; struct timeval start, end; double elapsed; struct stat statbuf; /* (Re)create a unique file and check that it has one only link. */ char *linkfile = gen_unique_nfs_filename( lockfile ); if( linkfile == NULL ) { goto done; } (void)unlink( linkfile ); if( ( fd = open( linkfile, O_CREAT|O_RDONLY ) ) == -1 ) { debug( 1, L"acquire_lock_file: open: %s", strerror( errno ) ); goto done; } close( fd ); if( stat( linkfile, &statbuf ) != 0 ) { debug( 1, L"acquire_lock_file: stat: %s", strerror( errno ) ); goto done; } if ( statbuf.st_nlink != 1 ) { debug( 1, L"acquire_lock_file: number of hardlinks on unique " L"tmpfile is %d instead of 1.", (int)statbuf.st_nlink ); goto done; } if( gettimeofday( &start, NULL ) != 0 ) { debug( 1, L"acquire_lock_file: gettimeofday: %s", strerror( errno ) ); goto done; } end = start; pollint.tv_sec = 0; pollint.tv_nsec = LOCKPOLLINTERVAL * 1000000; do { /* Try to create a hard link to the unique file from the lockfile. This will only succeed if the lockfile does not already exist. It is guaranteed to provide race-free semantics over NFS which the alternative of calling open(O_EXCL|O_CREAT) on the lockfile is not. The lock succeeds if the call to link returns 0 or the link count on the unique file increases to 2. */ if( link( linkfile, lockfile ) == 0 || ( stat( linkfile, &statbuf ) == 0 && statbuf.st_nlink == 2 ) ) { /* Successful lock */ ret = 1; break; } elapsed = end.tv_sec + end.tv_usec/1000000.0 - ( start.tv_sec + start.tv_usec/1000000.0 ); /* The check for elapsed < 0 is to deal with the unlikely event that after the loop is entered the system time is set forward past the loop's end time. This would otherwise result in a (practically) infinite loop. */ if( timed_out || elapsed >= timeout || elapsed < 0 ) { if ( timed_out == 0 && force ) { /* Timed out and force was specified - attempt to remove stale lock and try a final time */ (void)unlink( lockfile ); timed_out = 1; continue; } else { /* Timed out and final try was unsuccessful or force was not specified */ debug( 1, L"acquire_lock_file: timed out " L"trying to obtain lockfile %s using " L"linkfile %s", lockfile, linkfile ); break; } } nanosleep( &pollint, NULL ); } while( gettimeofday( &end, NULL ) == 0 ); done: /* The linkfile is not needed once the lockfile has been created */ (void)unlink( linkfile ); free( linkfile ); return ret; }