1
0
mirror of git://git.gnupg.org/gnupg.git synced 2024-11-10 21:38:50 +01:00
gnupg/util/miscutil.c
Werner Koch 3a22b622c8 Allow building with the 32 bit mingw-w64 toolchain.
* scripts/autogen.sh <--build-w32>: Support i686-w64-mingw32 and use
it by default if installed.
* keyserver/gpgkeys_ldap.c (my_ldap_start_tls_s): Define macro
depending on compiler version.
(main): Use new macro.
* util/miscutil.c [!HAVE_TIMEGM]: Add prototype for the timegm
autoconf replacement function.
--

It seems that the LDAP keyserver helper build with the old mingw32
toolchain never worked correctly for LDAPS.  The prototype there for
ldap_start_tls_s is plainly wrong for Windows.  Anyway I included
special support so not to break building with the old compiler.
2012-01-12 18:10:30 +01:00

525 lines
12 KiB
C

/* miscutil.c - miscellaneous utilities
* Copyright (C) 1998, 1999, 2000, 2001, 2003,
* 2004, 2005 Free Software Foundation, Inc.
*
* This file is part of GnuPG.
*
* GnuPG is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* GnuPG 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include <config.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <time.h>
#include <ctype.h>
#ifdef HAVE_LANGINFO_H
#include <langinfo.h>
#endif
#include "types.h"
#include "util.h"
#include "i18n.h"
#ifndef HAVE_TIMEGM
time_t timegm (struct tm *tm);
#endif
#ifdef HAVE_UNSIGNED_TIME_T
# define INVALID_TIME_CHECK(a) ((a) == (time_t)(-1))
#else
/* Error or 32 bit time_t and value after 2038-01-19. */
# define INVALID_TIME_CHECK(a) ((a) < 0)
#endif
/****************
* I know that the OpenPGP protocol has a Y2106 problem ;-)
*/
u32
make_timestamp()
{
return time(NULL);
}
/****************
* Scan a date string and return a timestamp.
* The only supported format is "yyyy-mm-dd"
* Returns 0 for an invalid date.
*/
u32
scan_isodatestr( const char *string )
{
int year, month, day;
struct tm tmbuf;
time_t stamp;
int i;
if( strlen(string) != 10 || string[4] != '-' || string[7] != '-' )
return 0;
for( i=0; i < 4; i++ )
if( !digitp(string+i) )
return 0;
if( !digitp(string+5) || !digitp(string+6) )
return 0;
if( !digitp(string+8) || !digitp(string+9) )
return 0;
year = atoi(string);
month = atoi(string+5);
day = atoi(string+8);
/* some basic checks */
if( year < 1970 || month < 1 || month > 12 || day < 1 || day > 31 )
return 0;
memset( &tmbuf, 0, sizeof tmbuf );
tmbuf.tm_mday = day;
tmbuf.tm_mon = month-1;
tmbuf.tm_year = year - 1900;
tmbuf.tm_isdst = -1;
stamp = mktime( &tmbuf );
if( stamp == (time_t)-1 )
return 0;
return stamp;
}
/****************
* Return a string with a time value in the form: x Y, n D, n H
*/
const char *
strtimevalue( u32 value )
{
static char buffer[30];
unsigned int years, days, hours, minutes;
value /= 60;
minutes = value % 60;
value /= 60;
hours = value % 24;
value /= 24;
days = value % 365;
value /= 365;
years = value;
sprintf(buffer,"%uy%ud%uh%um", years, days, hours, minutes );
if( years )
return buffer;
if( days )
return strchr( buffer, 'y' ) + 1;
return strchr( buffer, 'd' ) + 1;
}
/****************
* Note: this function returns GMT
*/
const char *
strtimestamp( u32 stamp )
{
static char buffer[11+5];
struct tm *tp;
time_t atime = stamp;
if (INVALID_TIME_CHECK (atime)) {
strcpy (buffer, "????" "-??" "-??");
}
else {
tp = gmtime( &atime );
sprintf(buffer,"%04d-%02d-%02d",
1900+tp->tm_year, tp->tm_mon+1, tp->tm_mday );
}
return buffer;
}
/****************
* Note: this function returns GMT
*/
const char *
isotimestamp (u32 stamp)
{
static char buffer[25+5];
struct tm *tp;
time_t atime = stamp;
if (INVALID_TIME_CHECK (atime)) {
strcpy (buffer, "????" "-??" "-??" " " "??" ":" "??" ":" "??");
}
else {
tp = gmtime( &atime );
sprintf(buffer,"%04d-%02d-%02d %02d:%02d:%02d",
1900+tp->tm_year, tp->tm_mon+1, tp->tm_mday,
tp->tm_hour, tp->tm_min, tp->tm_sec);
}
return buffer;
}
/* Scan am ISO timestamp and return an Epoch based timestamp. The only
supported format is "yyyymmddThhmmss" delimited by white space, nul, a
colon or a comma. Returns 0 for an invalid string. */
u32
isotime2seconds (const char *string)
{
const char *s;
int year, month, day, hour, minu, sec;
struct tm tmbuf;
int i;
time_t result;
if (!*string)
return 0;
for (s=string, i=0; i < 8; i++, s++)
if (!digitp (s))
return 0;
if (*s != 'T')
return 0;
for (s++, i=9; i < 15; i++, s++)
if (!digitp (s))
return 0;
if ( !(!*s || (isascii (*s) && isspace(*s)) || *s == ':' || *s == ','))
return 0; /* Wrong delimiter. */
year = atoi_4 (string);
month = atoi_2 (string + 4);
day = atoi_2 (string + 6);
hour = atoi_2 (string + 9);
minu = atoi_2 (string + 11);
sec = atoi_2 (string + 13);
/* Basic checks. */
if (year < 1970 || month < 1 || month > 12 || day < 1 || day > 31
|| hour > 23 || minu > 59 || sec > 61 )
return 0;
memset (&tmbuf, 0, sizeof tmbuf);
tmbuf.tm_sec = sec;
tmbuf.tm_min = minu;
tmbuf.tm_hour = hour;
tmbuf.tm_mday = day;
tmbuf.tm_mon = month-1;
tmbuf.tm_year = year - 1900;
tmbuf.tm_isdst = -1;
result = timegm (&tmbuf);
return (result == (time_t)(-1))? 0 : (u32)result;
}
/****************
* Note: this function returns local time
*/
const char *
asctimestamp( u32 stamp )
{
static char buffer[50];
#if defined (HAVE_STRFTIME) && defined (HAVE_NL_LANGINFO)
static char fmt[50];
#endif
struct tm *tp;
time_t atime = stamp;
if (INVALID_TIME_CHECK (atime))
{
strcpy (buffer, "????" "-??" "-??");
return buffer;
}
tp = localtime( &atime );
#ifdef HAVE_STRFTIME
#if defined(HAVE_NL_LANGINFO)
# if GNUPG_GCC_VERSION >= 40600
# pragma GCC diagnostic push
# pragma GCC diagnostic ignored "-Wformat-nonliteral"
# endif
mem2str( fmt, nl_langinfo(D_T_FMT), DIM(fmt)-3 );
if( strstr( fmt, "%Z" ) == NULL )
strcat( fmt, " %Z");
strftime( buffer, DIM(buffer)-1, fmt, tp );
# if GNUPG_GCC_VERSION >= 40600
# pragma GCC diagnostic pop
# endif
#else
/* fixme: we should check whether the locale appends a " %Z"
* These locales from glibc don't put the " %Z":
* fi_FI hr_HR ja_JP lt_LT lv_LV POSIX ru_RU ru_SU sv_FI sv_SE zh_CN
*/
strftime( buffer, DIM(buffer)-1,
#ifdef HAVE_W32_SYSTEM
"%c"
#else
"%c %Z"
#endif
, tp );
#endif
buffer[DIM(buffer)-1] = 0;
#else
mem2str( buffer, asctime(tp), DIM(buffer) );
#endif
return buffer;
}
/****************
* Print a string to FP, but filter all control characters out.
*/
void
print_string2( FILE *fp, const byte *p, size_t n, int delim, int delim2 )
{
for( ; n; n--, p++ )
if (*p < 0x20
|| *p == 0x7f
|| *p == delim || *p == delim2
|| ((delim || delim2) && *p=='\\'))
{
putc('\\', fp);
if( *p == '\n' )
putc('n', fp);
else if( *p == '\r' )
putc('r', fp);
else if( *p == '\f' )
putc('f', fp);
else if( *p == '\v' )
putc('v', fp);
else if( *p == '\b' )
putc('b', fp);
else if( !*p )
putc('0', fp);
else
fprintf(fp, "x%02x", *p );
}
else
putc(*p, fp);
}
void
print_string( FILE *fp, const byte *p, size_t n, int delim )
{
print_string2(fp,p,n,delim,0);
}
/****************
* Print an UTF8 string to FP and filter all control characters out.
*/
void
print_utf8_string2 ( FILE *fp, const byte *p, size_t n, int delim )
{
size_t i;
char *buf;
/* we can handle plain ascii simpler, so check for it first */
for(i=0; i < n; i++ ) {
if( p[i] & 0x80 )
break;
}
if( i < n ) {
buf = utf8_to_native ( p, n, delim );
/*(utf8 conversion already does the control character quoting)*/
fputs( buf, fp );
xfree( buf );
}
else
print_string( fp, p, n, delim );
}
void
print_utf8_string( FILE *fp, const byte *p, size_t n )
{
print_utf8_string2 (fp, p, n, 0);
}
/****************
* This function returns a string which is suitable for printing
* Caller must release it with xfree()
*/
char *
make_printable_string( const byte *p, size_t n, int delim )
{
size_t save_n, buflen;
const byte *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 < 0xa0) || *p == delim ||
(delim && *p=='\\')) {
if( *p=='\n' || *p=='\r' || *p=='\f'
|| *p=='\v' || *p=='\b' || !*p )
buflen += 2;
else
buflen += 4;
}
else
buflen++;
}
p = save_p;
n = save_n;
/* and now make the string */
d = buffer = xmalloc( buflen );
for( ; n; n--, p++ ) {
if( *p < 0x20 || (*p >= 0x7f && *p < 0xa0) || *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;
}
int
answer_is_yes_no_default( const char *s, int def_answer )
{
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_yes = _("yes");
const char *short_yes = _("yY");
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_no = _("no");
const char *short_no = _("nN");
/* Note: we have to use the local dependent strcasecmp here */
if( match_multistr(long_yes,s) )
return 1;
if( *s && strchr( short_yes, *s ) && !s[1] )
return 1;
/* test for no strings to catch ambiguities for the next test */
if( match_multistr(long_no,s) )
return 0;
if( *s && strchr( short_no, *s ) && !s[1] )
return 0;
/* test for the english version (for those who are used to type yes) */
if( !ascii_strcasecmp(s, "yes" ) )
return 1;
if( *s && strchr( "yY", *s ) && !s[1] )
return 1;
return def_answer;
}
int
answer_is_yes( const char *s )
{
return answer_is_yes_no_default(s,0);
}
/****************
* Return 1 for yes, -1 for quit, or 0 for no
*/
int
answer_is_yes_no_quit( const char *s )
{
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_yes = _("yes");
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_no = _("no");
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_quit = _("quit");
const char *short_yes = _("yY");
const char *short_no = _("nN");
const char *short_quit = _("qQ");
if( match_multistr(long_no,s) )
return 0;
if( match_multistr(long_yes,s) )
return 1;
if( match_multistr(long_quit,s) )
return -1;
if( *s && strchr( short_no, *s ) && !s[1] )
return 0;
if( *s && strchr( short_yes, *s ) && !s[1] )
return 1;
if( *s && strchr( short_quit, *s ) && !s[1] )
return -1;
/* but not here */
if( !ascii_strcasecmp(s, "yes" ) )
return 1;
if( !ascii_strcasecmp(s, "quit" ) )
return -1;
if( *s && strchr( "yY", *s ) && !s[1] )
return 1;
if( *s && strchr( "qQ", *s ) && !s[1] )
return -1;
return 0;
}
/*
Return 1 for okay, 0 for for cancel or DEF_ANSWER for default.
*/
int
answer_is_okay_cancel (const char *s, int def_answer)
{
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_okay = _("okay|okay");
/* TRANSLATORS: See doc/TRANSLATE about this string. */
const char *long_cancel = _("cancel|cancel");
const char *short_okay = _("oO");
const char *short_cancel = _("cC");
/* Note: We have to use the locale dependent strcasecmp */
if ( match_multistr(long_okay,s) )
return 1;
if ( match_multistr(long_cancel,s) )
return 0;
if ( *s && strchr( short_okay, *s ) && !s[1] )
return 1;
if ( *s && strchr( short_cancel, *s ) && !s[1] )
return 0;
/* Always test for the English values (not locale here) */
if ( !ascii_strcasecmp(s, "okay" ) )
return 1;
if ( !ascii_strcasecmp(s, "ok" ) )
return 1;
if ( !ascii_strcasecmp(s, "cancel" ) )
return 0;
if ( *s && strchr( "oO", *s ) && !s[1] )
return 1;
if ( *s && strchr( "cC", *s ) && !s[1] )
return 0;
return def_answer;
}
/* Try match against each substring of multistr, delimited by | */
int
match_multistr(const char *multistr,const char *match)
{
do
{
size_t seglen=strcspn(multistr,"|");
if(!seglen)
break;
/* Using the localized strncasecmp */
if(strncasecmp(multistr,match,seglen)==0)
return 1;
multistr+=seglen;
if(*multistr=='|')
multistr++;
}
while(*multistr);
return 0;
}