diff options
author | Werner Koch <[email protected]> | 2010-03-10 12:24:58 +0000 |
---|---|---|
committer | Werner Koch <[email protected]> | 2010-03-10 12:24:58 +0000 |
commit | d8b1099d01ebc1d305d47ec6dcb326980ad56396 (patch) | |
tree | cb92563d8b116872c34ad26dcdacfd02ed3da04a /common/stringhelp.c | |
parent | Fix for latest libgpg-error. (diff) | |
download | gnupg-d8b1099d01ebc1d305d47ec6dcb326980ad56396.tar.gz gnupg-d8b1099d01ebc1d305d47ec6dcb326980ad56396.zip |
Merged jnlib into common.
Diffstat (limited to 'common/stringhelp.c')
-rw-r--r-- | common/stringhelp.c | 1151 |
1 files changed, 1151 insertions, 0 deletions
diff --git a/common/stringhelp.c b/common/stringhelp.c new file mode 100644 index 000000000..36f96b8ff --- /dev/null +++ b/common/stringhelp.c @@ -0,0 +1,1151 @@ +/* stringhelp.c - standard string helper functions + * Copyright (C) 1998, 1999, 2000, 2001, 2003, 2004, 2005, + * 2006, 2007, 2008, 2009 Free Software Foundation, Inc. + * + * This file is part of JNLIB. + * + * JNLIB is free software; you can redistribute it and/or modify it + * under the terms of the GNU Lesser General Public License as + * published by the Free Software Foundation; either version 3 of + * the License, or (at your option) any later version. + * + * JNLIB is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this program; if not, see <http://www.gnu.org/licenses/>. + */ + +#include <config.h> +#include <stdlib.h> +#include <string.h> +#include <stdarg.h> +#include <ctype.h> +#include <errno.h> +#ifdef HAVE_PWD_H +# include <pwd.h> +#endif +#include <unistd.h> +#include <sys/types.h> +#ifdef HAVE_W32_SYSTEM +# include <windows.h> +#endif + +#include "libjnlib-config.h" +#include "utf8conv.h" +#include "stringhelp.h" + + +#define tohex_lower(n) ((n) < 10 ? ((n) + '0') : (((n) - 10) + 'a')) + +/* Sometimes we want to avoid mixing slashes and backslashes on W32 + and prefer backslashes. There is usual no problem with mixing + them, however a very few W32 API calls can't grok plain slashes. + Printing filenames with mixed slashes also looks a bit strange. + This function has no effext on POSIX. */ +static inline char * +change_slashes (char *name) +{ +#ifdef HAVE_DRIVE_LETTERS + char *p; + + if (strchr (name, '\\')) + { + for (p=name; *p; p++) + if (*p == '/') + *p = '\\'; + } +#endif /*HAVE_DRIVE_LETTERS*/ + return name; +} + + +/* + * Look for the substring SUB in buffer and return a pointer to that + * substring in BUFFER or NULL if not found. + * Comparison is case-insensitive. + */ +const char * +memistr (const void *buffer, size_t buflen, const char *sub) +{ + const unsigned char *buf = buffer; + const unsigned char *t = (const unsigned char *)buffer; + const unsigned char *s = (const unsigned char *)sub; + size_t n = buflen; + + for ( ; n ; t++, n-- ) + { + if ( toupper (*t) == toupper (*s) ) + { + for ( buf=t++, buflen = n--, s++; + n && toupper (*t) == toupper (*s); t++, s++, n-- ) + ; + if (!*s) + return (const char*)buf; + t = buf; + s = (const unsigned char *)sub ; + n = buflen; + } + } + return NULL; +} + +const char * +ascii_memistr ( const void *buffer, size_t buflen, const char *sub ) +{ + const unsigned char *buf = buffer; + const unsigned char *t = (const unsigned char *)buf; + const unsigned char *s = (const unsigned char *)sub; + size_t n = buflen; + + for ( ; n ; t++, n-- ) + { + if (ascii_toupper (*t) == ascii_toupper (*s) ) + { + for ( buf=t++, buflen = n--, s++; + n && ascii_toupper (*t) == ascii_toupper (*s); t++, s++, n-- ) + ; + if (!*s) + return (const char*)buf; + t = (const unsigned char *)buf; + s = (const unsigned char *)sub ; + n = buflen; + } + } + return NULL; +} + +/* This function is similar to strncpy(). However it won't copy more + than N - 1 characters and makes sure that a '\0' is appended. With + N given as 0, nothing will happen. With DEST given as NULL, memory + will be allocated using jnlib_xmalloc (i.e. if it runs out of core + the function terminates). Returns DES or a pointer to the + allocated memory. + */ +char * +mem2str( char *dest , const void *src , size_t n ) +{ + char *d; + const char *s; + + if( n ) { + if( !dest ) + dest = jnlib_xmalloc( n ) ; + d = dest; + s = src ; + for(n--; n && *s; n-- ) + *d++ = *s++; + *d = '\0' ; + } + + return dest ; +} + + +/**************** + * remove leading and trailing white spaces + */ +char * +trim_spaces( char *str ) +{ + char *string, *p, *mark; + + string = str; + /* find first non space character */ + for( p=string; *p && isspace( *(byte*)p ) ; p++ ) + ; + /* move characters */ + for( (mark = NULL); (*string = *p); string++, p++ ) + if( isspace( *(byte*)p ) ) { + if( !mark ) + mark = string ; + } + else + mark = NULL ; + if( mark ) + *mark = '\0' ; /* remove trailing spaces */ + + return str ; +} + +/**************** + * remove trailing white spaces + */ +char * +trim_trailing_spaces( char *string ) +{ + char *p, *mark; + + for( mark = NULL, p = string; *p; p++ ) { + if( isspace( *(byte*)p ) ) { + if( !mark ) + mark = p; + } + else + mark = NULL; + } + if( mark ) + *mark = '\0' ; + + return string ; +} + + +unsigned +trim_trailing_chars( byte *line, unsigned len, const char *trimchars ) +{ + byte *p, *mark; + unsigned n; + + for(mark=NULL, p=line, n=0; n < len; n++, p++ ) { + if( strchr(trimchars, *p ) ) { + if( !mark ) + mark = p; + } + else + mark = NULL; + } + + if( mark ) { + *mark = 0; + return mark - line; + } + return len; +} + +/**************** + * remove trailing white spaces and return the length of the buffer + */ +unsigned +trim_trailing_ws( byte *line, unsigned len ) +{ + return trim_trailing_chars( line, len, " \t\r\n" ); +} + +size_t +length_sans_trailing_chars (const unsigned char *line, size_t len, + const char *trimchars ) +{ + const unsigned char *p, *mark; + size_t n; + + for( mark=NULL, p=line, n=0; n < len; n++, p++ ) + { + if (strchr (trimchars, *p )) + { + if( !mark ) + mark = p; + } + else + mark = NULL; + } + + if (mark) + return mark - line; + return len; +} + +/* + * Return the length of line ignoring trailing white-space. + */ +size_t +length_sans_trailing_ws (const unsigned char *line, size_t len) +{ + return length_sans_trailing_chars (line, len, " \t\r\n"); +} + + + +/* + * Extract from a given path the filename component. This function + * terminates the process on memory shortage. + */ +char * +make_basename(const char *filepath, const char *inputpath) +{ +#ifdef __riscos__ + return riscos_make_basename(filepath, inputpath); +#else + char *p; + + (void)inputpath; /* Only required for riscos. */ + + if ( !(p=strrchr(filepath, '/')) ) +#ifdef HAVE_DRIVE_LETTERS + if ( !(p=strrchr(filepath, '\\')) ) + if ( !(p=strrchr(filepath, ':')) ) +#endif + { + return jnlib_xstrdup(filepath); + } + + return jnlib_xstrdup(p+1); +#endif +} + + + +/* + * Extract from a given filename the path prepended to it. If there + * isn't a path prepended to the filename, a dot is returned ('.'). + * This function terminates the process on memory shortage. + */ +char * +make_dirname(const char *filepath) +{ + char *dirname; + int dirname_length; + char *p; + + if ( !(p=strrchr(filepath, '/')) ) +#ifdef HAVE_DRIVE_LETTERS + if ( !(p=strrchr(filepath, '\\')) ) + if ( !(p=strrchr(filepath, ':')) ) +#endif + { + return jnlib_xstrdup("."); + } + + dirname_length = p-filepath; + dirname = jnlib_xmalloc(dirname_length+1); + strncpy(dirname, filepath, dirname_length); + dirname[dirname_length] = 0; + + return dirname; +} + + + +static char * +get_pwdir (int xmode, const char *name) +{ + char *result = NULL; +#ifdef HAVE_PWD_H + struct passwd *pwd = NULL; + + if (name) + { +#ifdef HAVE_GETPWNAM + /* Fixme: We should use getpwnam_r if available. */ + pwd = getpwnam (name); +#endif + } + else + { +#ifdef HAVE_GETPWUID + /* Fixme: We should use getpwuid_r if available. */ + pwd = getpwuid (getuid()); +#endif + } + if (pwd) + { + if (xmode) + result = jnlib_xstrdup (pwd->pw_dir); + else + result = jnlib_strdup (pwd->pw_dir); + } +#else /*!HAVE_PWD_H*/ + /* No support at all. */ + (void)xmode; + (void)name; +#endif /*HAVE_PWD_H*/ + return result; +} + +static char * +do_make_filename (int xmode, const char *first_part, va_list arg_ptr) +{ + const char *argv[32]; + int argc; + size_t n; + int skip = 1; + char *home_buffer = NULL; + char *name, *home, *p; + + n = strlen (first_part) + 1; + argc = 0; + while ( (argv[argc] = va_arg (arg_ptr, const char *)) ) + { + n += strlen (argv[argc]) + 1; + if (argc >= DIM (argv)-1) + { + if (xmode) + BUG (); + jnlib_set_errno (EINVAL); + return NULL; + } + argc++; + } + n++; + + home = NULL; + if (*first_part == '~') + { + if (first_part[1] == '/' || !first_part[1]) + { + /* This is the "~/" or "~" case. */ + home = getenv("HOME"); + if (!home) + home = home_buffer = get_pwdir (xmode, NULL); + if (home && *home) + n += strlen (home); + } + else + { + /* This is the "~username/" or "~username" case. */ + char *user; + + if (xmode) + user = jnlib_xstrdup (first_part+1); + else + { + user = jnlib_strdup (first_part+1); + if (!user) + return NULL; + } + p = strchr (user, '/'); + if (p) + *p = 0; + skip = 1 + strlen (user); + + home = home_buffer = get_pwdir (xmode, user); + jnlib_free (user); + if (home) + n += strlen (home); + else + skip = 1; + } + } + + if (xmode) + name = jnlib_xmalloc (n); + else + { + name = jnlib_malloc (n); + if (!name) + { + jnlib_free (home_buffer); + return NULL; + } + } + + if (home) + p = stpcpy (stpcpy (name, home), first_part + skip); + else + p = stpcpy (name, first_part); + + jnlib_free (home_buffer); + + for (argc=0; argv[argc]; argc++) + p = stpcpy (stpcpy (p, "/"), argv[argc]); + + return change_slashes (name); +} + +/* Construct a filename from the NULL terminated list of parts. Tilde + expansion is done for the first argument. This function terminates + the process on memory shortage. */ +char * +make_filename (const char *first_part, ... ) +{ + va_list arg_ptr; + char *result; + + va_start (arg_ptr, first_part); + result = do_make_filename (1, first_part, arg_ptr); + va_end (arg_ptr); + return result; +} + +/* Construct a filename from the NULL terminated list of parts. Tilde + expansion is done for the first argument. This function may return + NULL on error. */ +char * +make_filename_try (const char *first_part, ... ) +{ + va_list arg_ptr; + char *result; + + va_start (arg_ptr, first_part); + result = do_make_filename (0, first_part, arg_ptr); + va_end (arg_ptr); + return result; +} + + + +/* Compare whether the filenames are identical. This is a + special version of strcmp() taking the semantics of filenames in + account. Note that this function works only on the supplied names + without considereing any context like the current directory. See + also same_file_p(). */ +int +compare_filenames (const char *a, const char *b) +{ +#ifdef HAVE_DRIVE_LETTERS + for ( ; *a && *b; a++, b++ ) + { + if (*a != *b + && (toupper (*(const unsigned char*)a) + != toupper (*(const unsigned char*)b) ) + && !((*a == '/' && *b == '\\') || (*a == '\\' && *b == '/'))) + break; + } + if ((*a == '/' && *b == '\\') || (*a == '\\' && *b == '/')) + return 0; + else + return (toupper (*(const unsigned char*)a) + - toupper (*(const unsigned char*)b)); +#else + return strcmp(a,b); +#endif +} + + +/* Convert 2 hex characters at S to a byte value. Return this value + or -1 if there is an error. */ +int +hextobyte (const char *s) +{ + int c; + + if ( *s >= '0' && *s <= '9' ) + c = 16 * (*s - '0'); + else if ( *s >= 'A' && *s <= 'F' ) + c = 16 * (10 + *s - 'A'); + else if ( *s >= 'a' && *s <= 'f' ) + c = 16 * (10 + *s - 'a'); + else + return -1; + s++; + if ( *s >= '0' && *s <= '9' ) + c += *s - '0'; + else if ( *s >= 'A' && *s <= 'F' ) + c += 10 + *s - 'A'; + else if ( *s >= 'a' && *s <= 'f' ) + c += 10 + *s - 'a'; + else + return -1; + return c; +} + + +/* Print a BUFFER to stream FP while replacing all control characters + and the characters DELIM and DELIM2 with standard C escape + sequences. Returns the number of characters printed. */ +size_t +print_sanitized_buffer2 (FILE *fp, const void *buffer, size_t length, + int delim, int delim2) +{ + const unsigned char *p = buffer; + size_t count = 0; + + for (; length; length--, p++, count++) + { + if (*p < 0x20 + || *p == 0x7f + || *p == delim + || *p == delim2 + || ((delim || delim2) && *p=='\\')) + { + putc ('\\', fp); + count++; + if (*p == '\n') + { + putc ('n', fp); + count++; + } + else if (*p == '\r') + { + putc ('r', fp); + count++; + } + else if (*p == '\f') + { + putc ('f', fp); + count++; + } + else if (*p == '\v') + { + putc ('v', fp); + count++; + } + else if (*p == '\b') + { + putc ('b', fp); + count++; + } + else if (!*p) + { + putc('0', fp); + count++; + } + else + { + fprintf (fp, "x%02x", *p); + count += 3; + } + } + else + { + putc (*p, fp); + count++; + } + } + + return count; +} + +/* Same as print_sanitized_buffer2 but with just one delimiter. */ +size_t +print_sanitized_buffer (FILE *fp, const void *buffer, size_t length, + int delim) +{ + return print_sanitized_buffer2 (fp, buffer, length, delim, 0); +} + + +size_t +print_sanitized_utf8_buffer (FILE *fp, const void *buffer, + size_t length, int delim) +{ + const char *p = buffer; + size_t i; + + /* We can handle plain ascii simpler, so check for it first. */ + for (i=0; i < length; i++ ) + { + if ( (p[i] & 0x80) ) + break; + } + if (i < length) + { + char *buf = utf8_to_native (p, length, delim); + /*(utf8 conversion already does the control character quoting)*/ + i = strlen (buf); + fputs (buf, fp); + jnlib_free (buf); + return i; + } + else + return print_sanitized_buffer (fp, p, length, delim); +} + + +size_t +print_sanitized_string2 (FILE *fp, const char *string, int delim, int delim2) +{ + return string? print_sanitized_buffer2 (fp, string, strlen (string), + delim, delim2):0; +} + +size_t +print_sanitized_string (FILE *fp, const char *string, int delim) +{ + return string? print_sanitized_buffer (fp, string, strlen (string), delim):0; +} + +size_t +print_sanitized_utf8_string (FILE *fp, const char *string, int delim) +{ + return string? print_sanitized_utf8_buffer (fp, + string, strlen (string), + delim) : 0; +} + +/* Create a string from the buffer P_ARG of length N which is suitable + for printing. Caller must release the created string using xfree. + This function terminates the process on memory shortage. */ +char * +sanitize_buffer (const void *p_arg, size_t n, int delim) +{ + const unsigned char *p = p_arg; + size_t save_n, buflen; + const unsigned char *save_p; + char *buffer, *d; + + /* First count length. */ + for (save_n = n, save_p = p, buflen=1 ; n; n--, p++ ) + { + if ( *p < 0x20 || *p == 0x7f || *p == delim || (delim && *p=='\\')) + { + if ( *p=='\n' || *p=='\r' || *p=='\f' + || *p=='\v' || *p=='\b' || !*p ) + buflen += 2; + else + buflen += 5; + } + else + buflen++; + } + p = save_p; + n = save_n; + /* And now make the string */ + d = buffer = jnlib_xmalloc( buflen ); + for ( ; n; n--, p++ ) + { + if (*p < 0x20 || *p == 0x7f || *p == delim || (delim && *p=='\\')) { + *d++ = '\\'; + if( *p == '\n' ) + *d++ = 'n'; + else if( *p == '\r' ) + *d++ = 'r'; + else if( *p == '\f' ) + *d++ = 'f'; + else if( *p == '\v' ) + *d++ = 'v'; + else if( *p == '\b' ) + *d++ = 'b'; + else if( !*p ) + *d++ = '0'; + else { + sprintf(d, "x%02x", *p ); + d += 3; + } + } + else + *d++ = *p; + } + *d = 0; + return buffer; +} + + +/* Given a string containing an UTF-8 encoded text, return the number + of characters in this string. It differs from strlen in that it + only counts complete UTF-8 characters. Note, that this function + does not take combined characters into account. */ +size_t +utf8_charcount (const char *s) +{ + size_t n; + + for (n=0; *s; s++) + if ( (*s&0xc0) != 0x80 ) /* Exclude continuation bytes: 10xxxxxx */ + n++; + + return n; +} + + +/**************************************************** + ********** W32 specific functions **************** + ****************************************************/ + +#ifdef HAVE_W32_SYSTEM +const char * +w32_strerror (int ec) +{ + static char strerr[256]; + + if (ec == -1) + ec = (int)GetLastError (); +#ifdef HAVE_W32CE_SYSTEM + /* There is only a wchar_t FormatMessage. It does not make much + sense to play the conversion game; we print only the code. */ + snprintf (strerr, sizeof strerr, "ec=%d", (int)GetLastError ()); +#else + FormatMessage (FORMAT_MESSAGE_FROM_SYSTEM, NULL, ec, + MAKELANGID (LANG_NEUTRAL, SUBLANG_DEFAULT), + strerr, DIM (strerr)-1, NULL); +#endif + return strerr; +} +#endif /*HAVE_W32_SYSTEM*/ + + +/**************************************************** + ******** Locale insensitive ctype functions ******** + ****************************************************/ +/* FIXME: replace them by a table lookup and macros */ +int +ascii_isupper (int c) +{ + return c >= 'A' && c <= 'Z'; +} + +int +ascii_islower (int c) +{ + return c >= 'a' && c <= 'z'; +} + +int +ascii_toupper (int c) +{ + if (c >= 'a' && c <= 'z') + c &= ~0x20; + return c; +} + +int +ascii_tolower (int c) +{ + if (c >= 'A' && c <= 'Z') + c |= 0x20; + return c; +} + + +int +ascii_strcasecmp( const char *a, const char *b ) +{ + if (a == b) + return 0; + + for (; *a && *b; a++, b++) { + if (*a != *b && ascii_toupper(*a) != ascii_toupper(*b)) + break; + } + return *a == *b? 0 : (ascii_toupper (*a) - ascii_toupper (*b)); +} + +int +ascii_strncasecmp (const char *a, const char *b, size_t n) +{ + const unsigned char *p1 = (const unsigned char *)a; + const unsigned char *p2 = (const unsigned char *)b; + unsigned char c1, c2; + + if (p1 == p2 || !n ) + return 0; + + do + { + c1 = ascii_tolower (*p1); + c2 = ascii_tolower (*p2); + + if ( !--n || c1 == '\0') + break; + + ++p1; + ++p2; + } + while (c1 == c2); + + return c1 - c2; +} + + +int +ascii_memcasecmp (const void *a_arg, const void *b_arg, size_t n ) +{ + const char *a = a_arg; + const char *b = b_arg; + + if (a == b) + return 0; + for ( ; n; n--, a++, b++ ) + { + if( *a != *b && ascii_toupper (*a) != ascii_toupper (*b) ) + return *a == *b? 0 : (ascii_toupper (*a) - ascii_toupper (*b)); + } + return 0; +} + +int +ascii_strcmp( const char *a, const char *b ) +{ + if (a == b) + return 0; + + for (; *a && *b; a++, b++) { + if (*a != *b ) + break; + } + return *a == *b? 0 : (*(signed char *)a - *(signed char *)b); +} + + +void * +ascii_memcasemem (const void *haystack, size_t nhaystack, + const void *needle, size_t nneedle) +{ + + if (!nneedle) + return (void*)haystack; /* finding an empty needle is really easy */ + if (nneedle <= nhaystack) + { + const char *a = haystack; + const char *b = a + nhaystack - nneedle; + + for (; a <= b; a++) + { + if ( !ascii_memcasecmp (a, needle, nneedle) ) + return (void *)a; + } + } + return NULL; +} + +/********************************************* + ********** missing string functions ********* + *********************************************/ + +#ifndef HAVE_STPCPY +char * +stpcpy(char *a,const char *b) +{ + while( *b ) + *a++ = *b++; + *a = 0; + + return (char*)a; +} +#endif + +#ifndef HAVE_STRSEP +/* Code taken from glibc-2.2.1/sysdeps/generic/strsep.c. */ +char * +strsep (char **stringp, const char *delim) +{ + char *begin, *end; + + begin = *stringp; + if (begin == NULL) + return NULL; + + /* A frequent case is when the delimiter string contains only one + character. Here we don't need to call the expensive `strpbrk' + function and instead work using `strchr'. */ + if (delim[0] == '\0' || delim[1] == '\0') + { + char ch = delim[0]; + + if (ch == '\0') + end = NULL; + else + { + if (*begin == ch) + end = begin; + else if (*begin == '\0') + end = NULL; + else + end = strchr (begin + 1, ch); + } + } + else + /* Find the end of the token. */ + end = strpbrk (begin, delim); + + if (end) + { + /* Terminate the token and set *STRINGP past NUL character. */ + *end++ = '\0'; + *stringp = end; + } + else + /* No more delimiters; this is the last token. */ + *stringp = NULL; + + return begin; +} +#endif /*HAVE_STRSEP*/ + + +#ifndef HAVE_STRLWR +char * +strlwr(char *s) +{ + char *p; + for(p=s; *p; p++ ) + *p = tolower(*p); + return s; +} +#endif + + +#ifndef HAVE_STRCASECMP +int +strcasecmp( const char *a, const char *b ) +{ + for( ; *a && *b; a++, b++ ) { + if( *a != *b && toupper(*a) != toupper(*b) ) + break; + } + return *(const byte*)a - *(const byte*)b; +} +#endif + + +/**************** + * mingw32/cpd has a memicmp() + */ +#ifndef HAVE_MEMICMP +int +memicmp( const char *a, const char *b, size_t n ) +{ + for( ; n; n--, a++, b++ ) + if( *a != *b && toupper(*(const byte*)a) != toupper(*(const byte*)b) ) + return *(const byte *)a - *(const byte*)b; + return 0; +} +#endif + + +#ifndef HAVE_MEMRCHR +void * +memrchr (const void *buffer, int c, size_t n) +{ + const unsigned char *p = buffer; + + for (p += n; n ; n--) + if (*--p == c) + return (void *)p; + return NULL; +} +#endif /*HAVE_MEMRCHR*/ + + +/* Percent-escape the string STR by replacing colons with '%3a'. If + EXTRA is not NULL all characters in EXTRA are also escaped. */ +static char * +do_percent_escape (const char *str, const char *extra, int die) +{ + int i, j; + char *ptr; + + if (!str) + return NULL; + + for (i=j=0; str[i]; i++) + if (str[i] == ':' || str[i] == '%' || (extra && strchr (extra, str[i]))) + j++; + if (die) + ptr = jnlib_xmalloc (i + 2 * j + 1); + else + { + ptr = jnlib_malloc (i + 2 * j + 1); + if (!ptr) + return NULL; + } + i = 0; + while (*str) + { + if (*str == ':') + { + ptr[i++] = '%'; + ptr[i++] = '3'; + ptr[i++] = 'a'; + } + else if (*str == '%') + { + ptr[i++] = '%'; + ptr[i++] = '2'; + ptr[i++] = '5'; + } + else if (extra && strchr (extra, *str)) + { + ptr[i++] = '%'; + ptr[i++] = tohex_lower ((*str>>4)&15); + ptr[i++] = tohex_lower (*str&15); + } + else + ptr[i++] = *str; + str++; + } + ptr[i] = '\0'; + + return ptr; +} + +/* Percent-escape the string STR by replacing colons with '%3a'. If + EXTRA is not NULL all characters in EXTRA are also escaped. This + function terminates the process on memory shortage. */ +char * +percent_escape (const char *str, const char *extra) +{ + return do_percent_escape (str, extra, 1); +} + +/* Same as percent_escape but return NULL instead of exiting on memory + error. */ +char * +try_percent_escape (const char *str, const char *extra) +{ + return do_percent_escape (str, extra, 0); +} + + + +static char * +do_strconcat (const char *s1, va_list arg_ptr) +{ + const char *argv[48]; + size_t argc; + size_t needed; + char *buffer, *p; + + argc = 0; + argv[argc++] = s1; + needed = strlen (s1); + while (((argv[argc] = va_arg (arg_ptr, const char *)))) + { + needed += strlen (argv[argc]); + if (argc >= DIM (argv)-1) + { + jnlib_set_errno (EINVAL); + return NULL; + } + argc++; + } + needed++; + buffer = jnlib_malloc (needed); + if (buffer) + { + for (p = buffer, argc=0; argv[argc]; argc++) + p = stpcpy (p, argv[argc]); + } + return buffer; +} + + +/* Concatenate the string S1 with all the following strings up to a + NULL. Returns a malloced buffer with the new string or NULL on a + malloc error or if too many arguments are given. */ +char * +strconcat (const char *s1, ...) +{ + va_list arg_ptr; + char *result; + + if (!s1) + result = jnlib_strdup (""); + else + { + va_start (arg_ptr, s1); + result = do_strconcat (s1, arg_ptr); + va_end (arg_ptr); + } + return result; +} + +/* Same as strconcat but terminate the process with an error message + if something goes wrong. */ +char * +xstrconcat (const char *s1, ...) +{ + va_list arg_ptr; + char *result; + + if (!s1) + result = jnlib_xstrdup (""); + else + { + va_start (arg_ptr, s1); + result = do_strconcat (s1, arg_ptr); + va_end (arg_ptr); + } + if (!result) + { + if (errno == EINVAL) + fputs ("\nfatal: too many args for xstrconcat\n", stderr); + else + fputs ("\nfatal: out of memory\n", stderr); + exit (2); + } + return result; +} + |