1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) University of Cambridge 1995 - 2018 */
6 /* Copyright (c) The Exim Maintainers 2020 */
7 /* See the file NOTICE for conditions of use and distribution. */
9 /* Miscellaneous string-handling functions. Some are not required for
10 utilities and tests, and are cut out by the COMPILE_UTILITY macro. */
17 #ifndef COMPILE_UTILITY
18 /*************************************************
19 * Test for IP address *
20 *************************************************/
22 /* This used just to be a regular expression, but with IPv6 things are a bit
23 more complicated. If the address contains a colon, it is assumed to be a v6
24 address (assuming HAVE_IPV6 is set). If a mask is permitted and one is present,
25 and maskptr is not NULL, its offset is placed there.
29 maskptr NULL if no mask is permitted to follow
30 otherwise, points to an int where the offset of '/' is placed
31 if there is no / followed by trailing digits, *maskptr is set 0
33 Returns: 0 if the string is not a textual representation of an IP address
34 4 if it is an IPv4 address
35 6 if it is an IPv6 address
39 string_is_ip_address(const uschar *s, int *maskptr)
43 /* If an optional mask is permitted, check for it. If found, pass back the
48 const uschar *ss = s + Ustrlen(s);
50 if (s != ss && isdigit(*(--ss)))
52 while (ss > s && isdigit(ss[-1])) ss--;
53 if (ss > s && *(--ss) == '/') *maskptr = ss - s;
57 /* A colon anywhere in the string => IPv6 address */
59 if (Ustrchr(s, ':') != NULL)
61 BOOL had_double_colon = FALSE;
66 /* An IPv6 address must start with hex digit or double colon. A single
69 if (*s == ':' && *(++s) != ':') return 0;
71 /* Now read up to 8 components consisting of up to 4 hex digits each. There
72 may be one and only one appearance of double colon, which implies any number
73 of binary zero bits. The number of preceding components is held in count. */
75 for (int count = 0; count < 8; count++)
77 /* If the end of the string is reached before reading 8 components, the
78 address is valid provided a double colon has been read. This also applies
79 if we hit the / that introduces a mask or the % that introduces the
80 interface specifier (scope id) of a link-local address. */
82 if (*s == 0 || *s == '%' || *s == '/') return had_double_colon ? yield : 0;
84 /* If a component starts with an additional colon, we have hit a double
85 colon. This is permitted to appear once only, and counts as at least
86 one component. The final component may be of this form. */
90 if (had_double_colon) return 0;
91 had_double_colon = TRUE;
96 /* If the remainder of the string contains a dot but no colons, we
97 can expect a trailing IPv4 address. This is valid if either there has
98 been no double-colon and this is the 7th component (with the IPv4 address
99 being the 7th & 8th components), OR if there has been a double-colon
100 and fewer than 6 components. */
102 if (Ustrchr(s, ':') == NULL && Ustrchr(s, '.') != NULL)
104 if ((!had_double_colon && count != 6) ||
105 (had_double_colon && count > 6)) return 0;
111 /* Check for at least one and not more than 4 hex digits for this
114 if (!isxdigit(*s++)) return 0;
115 if (isxdigit(*s) && isxdigit(*(++s)) && isxdigit(*(++s))) s++;
117 /* If the component is terminated by colon and there is more to
118 follow, skip over the colon. If there is no more to follow the address is
121 if (*s == ':' && *(++s) == 0) return 0;
124 /* If about to handle a trailing IPv4 address, drop through. Otherwise
125 all is well if we are at the end of the string or at the mask or at a percent
126 sign, which introduces the interface specifier (scope id) of a link local
130 return (*s == 0 || *s == '%' ||
131 (*s == '/' && maskptr != NULL && *maskptr != 0))? yield : 0;
134 /* Test for IPv4 address, which may be the tail-end of an IPv6 address. */
136 for (int i = 0; i < 4; i++)
141 if (i != 0 && *s++ != '.') return 0;
142 n = strtol(CCS s, CSS &end, 10);
143 if (n > 255 || n < 0 || end <= s || end > s+3) return 0;
147 return !*s || (*s == '/' && maskptr && *maskptr != 0) ? yield : 0;
149 #endif /* COMPILE_UTILITY */
152 /*************************************************
153 * Format message size *
154 *************************************************/
156 /* Convert a message size in bytes to printing form, rounding
157 according to the magnitude of the number. A value of zero causes
158 a string of spaces to be returned.
161 size the message size in bytes
162 buffer where to put the answer
164 Returns: pointer to the buffer
165 a string of exactly 5 characters is normally returned
169 string_format_size(int size, uschar *buffer)
171 if (size == 0) Ustrcpy(buffer, US" ");
172 else if (size < 1024) sprintf(CS buffer, "%5d", size);
173 else if (size < 10*1024)
174 sprintf(CS buffer, "%4.1fK", (double)size / 1024.0);
175 else if (size < 1024*1024)
176 sprintf(CS buffer, "%4dK", (size + 512)/1024);
177 else if (size < 10*1024*1024)
178 sprintf(CS buffer, "%4.1fM", (double)size / (1024.0 * 1024.0));
180 sprintf(CS buffer, "%4dM", (size + 512 * 1024)/(1024*1024));
186 #ifndef COMPILE_UTILITY
187 /*************************************************
188 * Convert a number to base 62 format *
189 *************************************************/
191 /* Convert a long integer into an ASCII base 62 string. For Cygwin the value of
192 BASE_62 is actually 36. Always return exactly 6 characters plus zero, in a
195 Argument: a long integer
196 Returns: pointer to base 62 string
200 string_base62(unsigned long int value)
202 static uschar yield[7];
203 uschar *p = yield + sizeof(yield) - 1;
207 *(--p) = base62_chars[value % BASE_62];
212 #endif /* COMPILE_UTILITY */
216 /*************************************************
217 * Interpret escape sequence *
218 *************************************************/
220 /* This function is called from several places where escape sequences are to be
221 interpreted in strings.
224 pp points a pointer to the initiating "\" in the string;
225 the pointer gets updated to point to the final character
226 If the backslash is the last character in the string, it
228 Returns: the value of the character escape
232 string_interpret_escape(const uschar **pp)
234 #ifdef COMPILE_UTILITY
235 const uschar *hex_digits= CUS"0123456789abcdef";
238 const uschar *p = *pp;
240 if (ch == '\0') return **pp;
241 if (isdigit(ch) && ch != '8' && ch != '9')
244 if (isdigit(p[1]) && p[1] != '8' && p[1] != '9')
246 ch = ch * 8 + *(++p) - '0';
247 if (isdigit(p[1]) && p[1] != '8' && p[1] != '9')
248 ch = ch * 8 + *(++p) - '0';
253 case 'b': ch = '\b'; break;
254 case 'f': ch = '\f'; break;
255 case 'n': ch = '\n'; break;
256 case 'r': ch = '\r'; break;
257 case 't': ch = '\t'; break;
258 case 'v': ch = '\v'; break;
264 Ustrchr(hex_digits, tolower(*(++p))) - hex_digits;
265 if (isxdigit(p[1])) ch = ch * 16 +
266 Ustrchr(hex_digits, tolower(*(++p))) - hex_digits;
276 #ifndef COMPILE_UTILITY
277 /*************************************************
278 * Ensure string is printable *
279 *************************************************/
281 /* This function is called for critical strings. It checks for any
282 non-printing characters, and if any are found, it makes a new copy
283 of the string with suitable escape sequences. It is most often called by the
284 macro string_printing(), which sets flags to 0.
288 flags Bit 0: convert tabs. Bit 1: convert spaces.
290 Returns: string with non-printers encoded as printing sequences
294 string_printing2(const uschar *s, int flags)
296 int nonprintcount = 0;
305 || flags & SP_TAB && c == '\t'
306 || flags & SP_SPACE && c == ' '
311 if (nonprintcount == 0) return s;
313 /* Get a new block of store guaranteed big enough to hold the
316 tt = ss = store_get(length + nonprintcount * 3 + 1, is_tainted(s));
318 /* Copy everything, escaping non printers. */
324 && (!(flags & SP_TAB) || c != '\t')
325 && (!(flags & SP_SPACE) || c != ' ')
333 case '\n': *tt++ = 'n'; break;
334 case '\r': *tt++ = 'r'; break;
335 case '\b': *tt++ = 'b'; break;
336 case '\v': *tt++ = 'v'; break;
337 case '\f': *tt++ = 'f'; break;
338 case '\t': *tt++ = 't'; break;
339 default: sprintf(CS tt, "%03o", *t); tt += 3; break;
347 #endif /* COMPILE_UTILITY */
349 /*************************************************
350 * Undo printing escapes in string *
351 *************************************************/
353 /* This function is the reverse of string_printing2. It searches for
354 backslash characters and if any are found, it makes a new copy of the
355 string with escape sequences parsed. Otherwise it returns the original
361 Returns: string with printing escapes parsed back
365 string_unprinting(uschar *s)
367 uschar *p, *q, *r, *ss;
370 p = Ustrchr(s, '\\');
373 len = Ustrlen(s) + 1;
374 ss = store_get(len, is_tainted(s));
388 *q++ = string_interpret_escape((const uschar **)&p);
393 r = Ustrchr(p, '\\');
419 #if (defined(HAVE_LOCAL_SCAN) || defined(EXPAND_DLFUNC)) \
420 && !defined(MACRO_PREDEF) && !defined(COMPILE_UTILITY)
421 /*************************************************
422 * Copy and save string *
423 *************************************************/
426 Argument: string to copy
427 Returns: copy of string in new store with the same taint status
431 string_copy_function(const uschar *s)
433 return string_copy_taint(s, is_tainted(s));
436 /* This function assumes that memcpy() is faster than strcpy().
437 As above, but explicitly specifying the result taint status
441 string_copy_taint_function(const uschar * s, BOOL tainted)
443 int len = Ustrlen(s) + 1;
444 uschar *ss = store_get(len, tainted);
451 /*************************************************
452 * Copy and save string, given length *
453 *************************************************/
455 /* It is assumed the data contains no zeros. A zero is added
460 n number of characters
462 Returns: copy of string in new store
466 string_copyn_function(const uschar *s, int n)
468 uschar *ss = store_get(n + 1, is_tainted(s));
476 /*************************************************
477 * Copy and save string in malloc'd store *
478 *************************************************/
480 /* This function assumes that memcpy() is faster than strcpy().
482 Argument: string to copy
483 Returns: copy of string in new store
487 string_copy_malloc(const uschar *s)
489 int len = Ustrlen(s) + 1;
490 uschar *ss = store_malloc(len);
497 /*************************************************
498 * Copy string if long, inserting newlines *
499 *************************************************/
501 /* If the given string is longer than 75 characters, it is copied, and within
502 the copy, certain space characters are converted into newlines.
504 Argument: pointer to the string
505 Returns: pointer to the possibly altered string
509 string_split_message(uschar *msg)
513 if (msg == NULL || Ustrlen(msg) <= 75) return msg;
514 s = ss = msg = string_copy(msg);
519 while (i < 75 && *ss != 0 && *ss != '\n') ss++, i++;
531 if (t[-1] == ':') { tt = t; break; }
532 if (tt == NULL) tt = t;
536 if (tt == NULL) /* Can't split behind - try ahead */
541 if (*t == ' ' || *t == '\n')
547 if (tt == NULL) break; /* Can't find anywhere to split */
558 /*************************************************
559 * Copy returned DNS domain name, de-escaping *
560 *************************************************/
562 /* If a domain name contains top-bit characters, some resolvers return
563 the fully qualified name with those characters turned into escapes. The
564 convention is a backslash followed by _decimal_ digits. We convert these
565 back into the original binary values. This will be relevant when
566 allow_utf8_domains is set true and UTF-8 characters are used in domain
567 names. Backslash can also be used to escape other characters, though we
568 shouldn't come across them in domain names.
570 Argument: the domain name string
571 Returns: copy of string in new store, de-escaped
575 string_copy_dnsdomain(uschar *s)
578 uschar *ss = yield = store_get(Ustrlen(s) + 1, is_tainted(s));
584 else if (isdigit(s[1]))
586 *ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0';
589 else if (*(++s) != 0)
598 #ifndef COMPILE_UTILITY
599 /*************************************************
600 * Copy space-terminated or quoted string *
601 *************************************************/
603 /* This function copies from a string until its end, or until whitespace is
604 encountered, unless the string begins with a double quote, in which case the
605 terminating quote is sought, and escaping within the string is done. The length
606 of a de-quoted string can be no longer than the original, since escaping always
607 turns n characters into 1 character.
609 Argument: pointer to the pointer to the first character, which gets updated
610 Returns: the new string
614 string_dequote(const uschar **sptr)
616 const uschar *s = *sptr;
619 /* First find the end of the string */
622 while (*s != 0 && !isspace(*s)) s++;
626 while (*s && *s != '\"')
628 if (*s == '\\') (void)string_interpret_escape(&s);
634 /* Get enough store to copy into */
636 t = yield = store_get(s - *sptr + 1, is_tainted(*sptr));
642 while (*s != 0 && !isspace(*s)) *t++ = *s++;
646 while (*s != 0 && *s != '\"')
648 *t++ = *s == '\\' ? string_interpret_escape(&s) : *s;
654 /* Update the pointer and return the terminated copy */
660 #endif /* COMPILE_UTILITY */
664 /*************************************************
665 * Format a string and save it *
666 *************************************************/
668 /* The formatting is done by string_vformat, which checks the length of
669 everything. Taint is taken from the worst of the arguments.
672 format a printf() format - deliberately char * rather than uschar *
673 because it will most usually be a literal string
674 ... arguments for format
676 Returns: pointer to fresh piece of store containing sprintf'ed string
680 string_sprintf_trc(const char *format, const uschar * func, unsigned line, ...)
682 #ifdef COMPILE_UTILITY
683 uschar buffer[STRING_SPRINTF_BUFFER_SIZE];
684 gstring gs = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer };
689 unsigned flags = SVFMT_REBUFFER|SVFMT_EXTEND;
694 g = string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE,
699 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
700 "string_sprintf expansion was longer than %d; format string was (%s)\n"
701 " called from %s %d\n",
702 STRING_SPRINTF_BUFFER_SIZE, format, func, line);
704 #ifdef COMPILE_UTILITY
705 return string_copyn(g->s, g->ptr);
707 gstring_release_unused(g);
708 return string_from_gstring(g);
714 /*************************************************
715 * Case-independent strncmp() function *
716 *************************************************/
722 n number of characters to compare
724 Returns: < 0, = 0, or > 0, according to the comparison
728 strncmpic(const uschar *s, const uschar *t, int n)
732 int c = tolower(*s++) - tolower(*t++);
739 /*************************************************
740 * Case-independent strcmp() function *
741 *************************************************/
748 Returns: < 0, = 0, or > 0, according to the comparison
752 strcmpic(const uschar *s, const uschar *t)
756 int c = tolower(*s++) - tolower(*t++);
757 if (c != 0) return c;
763 /*************************************************
764 * Case-independent strstr() function *
765 *************************************************/
767 /* The third argument specifies whether whitespace is required
768 to follow the matched string.
772 t substring to search for
773 space_follows if TRUE, match only if whitespace follows
775 Returns: pointer to substring in string, or NULL if not found
779 strstric(uschar *s, uschar *t, BOOL space_follows)
782 uschar *yield = NULL;
783 int cl = tolower(*p);
784 int cu = toupper(*p);
788 if (*s == cl || *s == cu)
790 if (yield == NULL) yield = s;
793 if (!space_follows || s[1] == ' ' || s[1] == '\n' ) return yield;
801 else if (yield != NULL)
815 #ifdef COMPILE_UTILITY
816 /* Dummy version for this function; it should never be called */
818 gstring_grow(gstring * g, int count)
826 #ifndef COMPILE_UTILITY
827 /*************************************************
828 * Get next string from separated list *
829 *************************************************/
831 /* Leading and trailing space is removed from each item. The separator in the
832 list is controlled by the int pointed to by the separator argument as follows:
834 If the value is > 0 it is used as the separator. This is typically used for
835 sublists such as slash-separated options. The value is always a printing
838 (If the value is actually > UCHAR_MAX there is only one item in the list.
839 This is used for some cases when called via functions that sometimes
840 plough through lists, and sometimes are given single items.)
842 If the value is <= 0, the string is inspected for a leading <x, where x is an
843 ispunct() or an iscntrl() character. If found, x is used as the separator. If
846 (a) if separator == 0, ':' is used
847 (b) if separator <0, -separator is used
849 In all cases the value of the separator that is used is written back to the
850 int so that it is used on subsequent calls as we progress through the list.
852 A literal ispunct() separator can be represented in an item by doubling, but
853 there is no way to include an iscntrl() separator as part of the data.
856 listptr points to a pointer to the current start of the list; the
857 pointer gets updated to point after the end of the next item
858 separator a pointer to the separator character in an int (see above)
859 buffer where to put a copy of the next string in the list; or
860 NULL if the next string is returned in new memory
861 buflen when buffer is not NULL, the size of buffer; otherwise ignored
863 Returns: pointer to buffer, containing the next substring,
864 or NULL if no more substrings
868 string_nextinlist_trc(const uschar **listptr, int *separator, uschar *buffer, int buflen,
869 const uschar * func, int line)
871 int sep = *separator;
872 const uschar *s = *listptr;
877 /* This allows for a fixed specified separator to be an iscntrl() character,
878 but at the time of implementation, this is never the case. However, it's best
879 to be conservative. */
881 while (isspace(*s) && *s != sep) s++;
883 /* A change of separator is permitted, so look for a leading '<' followed by an
884 allowed character. */
888 if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1])))
892 while (isspace(*s) && *s != sep) s++;
895 sep = sep ? -sep : ':';
899 /* An empty string has no list elements */
901 if (!*s) return NULL;
903 /* Note whether whether or not the separator is an iscntrl() character. */
905 sep_is_special = iscntrl(sep);
907 /* Handle the case when a buffer is provided. */
912 if (is_tainted(s) && !is_tainted(buffer))
913 die_tainted(US"string_nextinlist", func, line);
916 if (*s == sep && (*(++s) != sep || sep_is_special)) break;
917 if (p < buflen - 1) buffer[p++] = *s;
919 while (p > 0 && isspace(buffer[p-1])) p--;
923 /* Handle the case when a buffer is not provided. */
929 /* We know that *s != 0 at this point. However, it might be pointing to a
930 separator, which could indicate an empty string, or (if an ispunct()
931 character) could be doubled to indicate a separator character as data at the
932 start of a string. Avoid getting working memory for an empty item. */
935 if (*++s != sep || sep_is_special)
938 return string_copy(US"");
941 /* Not an empty string; the first character is guaranteed to be a data
947 for (ss = s + 1; *ss && *ss != sep; ) ss++;
948 g = string_catn(g, s, ss-s);
950 if (!*s || *++s != sep || sep_is_special) break;
952 /* while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; */
953 while ( g->ptr > 0 && isspace(g->s[g->ptr-1])
954 && (g->ptr == 1 || g->s[g->ptr-2] != '\\') )
956 buffer = string_from_gstring(g);
957 gstring_release_unused(g);
960 /* Update the current pointer and return the new string */
967 static const uschar *
968 Ustrnchr(const uschar * s, int c, unsigned * len)
973 if (!*s) return NULL;
986 /************************************************
987 * Add element to separated list *
988 ************************************************/
989 /* This function is used to build a list, returning an allocated null-terminated
990 growable string. The given element has any embedded separator characters
993 Despite having the same growable-string interface as string_cat() the list is
994 always returned null-terminated.
997 list expanding-string for the list that is being built, or NULL
998 if this is a new list that has no contents yet
999 sep list separator character
1000 ele new element to be appended to the list
1002 Returns: pointer to the start of the list, changed if copied for expansion.
1006 string_append_listele(gstring * list, uschar sep, const uschar * ele)
1010 if (list && list->ptr)
1011 list = string_catn(list, &sep, 1);
1013 while((sp = Ustrchr(ele, sep)))
1015 list = string_catn(list, ele, sp-ele+1);
1016 list = string_catn(list, &sep, 1);
1019 list = string_cat(list, ele);
1020 (void) string_from_gstring(list);
1026 string_append_listele_n(gstring * list, uschar sep, const uschar * ele,
1031 if (list && list->ptr)
1032 list = string_catn(list, &sep, 1);
1034 while((sp = Ustrnchr(ele, sep, &len)))
1036 list = string_catn(list, ele, sp-ele+1);
1037 list = string_catn(list, &sep, 1);
1041 list = string_catn(list, ele, len);
1042 (void) string_from_gstring(list);
1048 /* A slightly-bogus listmaker utility; the separator is a string so
1049 can be multiple chars - there is no checking for the element content
1050 containing any of the separator. */
1053 string_append2_listele_n(gstring * list, const uschar * sepstr,
1054 const uschar * ele, unsigned len)
1056 if (list && list->ptr)
1057 list = string_cat(list, sepstr);
1059 list = string_catn(list, ele, len);
1060 (void) string_from_gstring(list);
1066 /************************************************/
1067 /* Add more space to a growable-string. The caller should check
1068 first if growth is required. The gstring struct is modified on
1069 return; specifically, the string-base-pointer may have been changed.
1072 g the growable-string
1073 count amount needed for g->ptr to increase by
1077 gstring_grow(gstring * g, int count)
1080 int oldsize = g->size;
1081 BOOL tainted = is_tainted(g->s);
1083 /* Mostly, string_cat() is used to build small strings of a few hundred
1084 characters at most. There are times, however, when the strings are very much
1085 longer (for example, a lookup that returns a vast number of alias addresses).
1086 To try to keep things reasonable, we use increments whose size depends on the
1087 existing length of the string. */
1089 unsigned inc = oldsize < 4096 ? 127 : 1023;
1091 if (count <= 0) return;
1092 g->size = (p + count + inc + 1) & ~inc; /* one for a NUL */
1094 /* Try to extend an existing allocation. If the result of calling
1095 store_extend() is false, either there isn't room in the current memory block,
1096 or this string is not the top item on the dynamic store stack. We then have
1097 to get a new chunk of store and copy the old string. When building large
1098 strings, it is helpful to call store_release() on the old string, to release
1099 memory blocks that have become empty. (The block will be freed if the string
1100 is at its start.) However, we can do this only if we know that the old string
1101 was the last item on the dynamic memory stack. This is the case if it matches
1104 if (!store_extend(g->s, tainted, oldsize, g->size))
1105 g->s = store_newblock(g->s, tainted, g->size, p);
1110 /*************************************************
1111 * Add chars to string *
1112 *************************************************/
1113 /* This function is used when building up strings of unknown length. Room is
1114 always left for a terminating zero to be added to the string that is being
1115 built. This function does not require the string that is being added to be NUL
1116 terminated, because the number of characters to add is given explicitly. It is
1117 sometimes called to extract parts of other strings.
1120 string points to the start of the string that is being built, or NULL
1121 if this is a new string that has no contents yet
1122 s points to characters to add
1123 count count of characters to add; must not exceed the length of s, if s
1126 Returns: pointer to the start of the string, changed if copied for expansion.
1127 Note that a NUL is not added, though space is left for one. This is
1128 because string_cat() is often called multiple times to build up a
1129 string - there's no point adding the NUL till the end.
1132 /* coverity[+alloc] */
1135 string_catn(gstring * g, const uschar *s, int count)
1138 BOOL srctaint = is_tainted(s);
1142 unsigned inc = count < 4096 ? 127 : 1023;
1143 unsigned size = ((count + inc) & ~inc) + 1;
1144 g = string_get_tainted(size, srctaint);
1146 else if (srctaint && !is_tainted(g->s))
1147 gstring_rebuffer(g);
1150 if (p + count >= g->size)
1151 gstring_grow(g, count);
1153 /* Because we always specify the exact number of characters to copy, we can
1154 use memcpy(), which is likely to be more efficient than strncopy() because the
1155 latter has to check for zero bytes. */
1157 memcpy(g->s + p, s, count);
1164 string_cat(gstring *string, const uschar *s)
1166 return string_catn(string, s, Ustrlen(s));
1171 /*************************************************
1172 * Append strings to another string *
1173 *************************************************/
1175 /* This function can be used to build a string from many other strings.
1176 It calls string_cat() to do the dirty work.
1179 string expanding-string that is being built, or NULL
1180 if this is a new string that has no contents yet
1181 count the number of strings to append
1182 ... "count" uschar* arguments, which must be valid zero-terminated
1185 Returns: pointer to the start of the string, changed if copied for expansion.
1186 The string is not zero-terminated - see string_cat() above.
1189 __inline__ gstring *
1190 string_append(gstring *string, int count, ...)
1194 va_start(ap, count);
1197 uschar *t = va_arg(ap, uschar *);
1198 string = string_cat(string, t);
1208 /*************************************************
1209 * Format a string with length checks *
1210 *************************************************/
1212 /* This function is used to format a string with checking of the length of the
1213 output for all conversions. It protects Exim from absent-mindedness when
1214 calling functions like debug_printf and string_sprintf, and elsewhere. There
1215 are two different entry points to what is actually the same function, depending
1216 on whether the variable length list of data arguments are given explicitly or
1219 The formats are the usual printf() ones, with some omissions (never used) and
1220 three additions for strings: %S forces lower case, %T forces upper case, and
1221 %#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed
1222 (useful in debugging). There is also the addition of %D and %M, which insert
1223 the date in the form used for datestamped log files.
1226 buffer a buffer in which to put the formatted string
1227 buflen the length of the buffer
1228 format the format string - deliberately char * and not uschar *
1229 ... or ap variable list of supplementary arguments
1231 Returns: TRUE if the result fitted in the buffer
1235 string_format_trc(uschar * buffer, int buflen,
1236 const uschar * func, unsigned line, const char * format, ...)
1238 gstring g = { .size = buflen, .ptr = 0, .s = buffer }, *gp;
1240 va_start(ap, format);
1241 gp = string_vformat_trc(&g, func, line, STRING_SPRINTF_BUFFER_SIZE,
1251 /* Build or append to a growing-string, sprintf-style.
1255 func called-from function name, for debug
1256 line called-from file line number, for debug
1257 limit maximum string size
1259 format printf-like format string
1260 ap variable-args pointer
1263 SVFMT_EXTEND buffer can be created or exteded as needed
1264 SVFMT_REBUFFER buffer can be recopied to tainted mem as needed
1265 SVFMT_TAINT_NOCHK do not check inputs for taint
1267 If the "extend" flag is true, the string passed in can be NULL,
1268 empty, or non-empty. Growing is subject to an overall limit given
1269 by the limit argument.
1271 If the "extend" flag is false, the string passed in may not be NULL,
1272 will not be grown, and is usable in the original place after return.
1273 The return value can be NULL to signify overflow.
1275 Returns the possibly-new (if copy for growth or taint-handling was needed)
1276 string, not nul-terminated.
1280 string_vformat_trc(gstring * g, const uschar * func, unsigned line,
1281 unsigned size_limit, unsigned flags, const char *format, va_list ap)
1283 enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 };
1285 int width, precision, off, lim, need;
1286 const char * fp = format; /* Deliberately not unsigned */
1287 BOOL dest_tainted = FALSE;
1289 string_datestamp_offset = -1; /* Datestamp not inserted */
1290 string_datestamp_length = 0; /* Datestamp not inserted */
1291 string_datestamp_type = 0; /* Datestamp not inserted */
1293 #ifdef COMPILE_UTILITY
1294 assert(!(flags & SVFMT_EXTEND));
1298 /* Ensure we have a string, to save on checking later */
1299 if (!g) g = string_get(16);
1300 else if (!(flags & SVFMT_TAINT_NOCHK)) dest_tainted = is_tainted(g->s);
1302 if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(format))
1304 #ifndef MACRO_PREDEF
1305 if (!(flags & SVFMT_REBUFFER))
1306 die_tainted(US"string_vformat", func, line);
1308 gstring_rebuffer(g);
1309 dest_tainted = TRUE;
1311 #endif /*!COMPILE_UTILITY*/
1313 lim = g->size - 1; /* leave one for a nul */
1314 off = g->ptr; /* remember initial offset in gstring */
1316 /* Scan the format and handle the insertions */
1320 int length = L_NORMAL;
1323 const char *null = "NULL"; /* ) These variables */
1324 const char *item_start, *s; /* ) are deliberately */
1325 char newformat[16]; /* ) not unsigned */
1326 char * gp = CS g->s + g->ptr; /* ) */
1328 /* Non-% characters just get copied verbatim */
1332 /* Avoid string_copyn() due to COMPILE_UTILITY */
1333 if ((need = g->ptr + 1) > lim)
1335 if (!(flags & SVFMT_EXTEND) || need > size_limit) return NULL;
1339 g->s[g->ptr++] = (uschar) *fp++;
1343 /* Deal with % characters. Pick off the width and precision, for checking
1344 strings, skipping over the flag and modifier characters. */
1347 width = precision = -1;
1349 if (strchr("-+ #0", *(++fp)) != NULL)
1351 if (*fp == '#') null = "";
1355 if (isdigit((uschar)*fp))
1357 width = *fp++ - '0';
1358 while (isdigit((uschar)*fp)) width = width * 10 + *fp++ - '0';
1360 else if (*fp == '*')
1362 width = va_arg(ap, int);
1369 precision = va_arg(ap, int);
1373 for (precision = 0; isdigit((uschar)*fp); fp++)
1374 precision = precision*10 + *fp - '0';
1376 /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */
1379 { fp++; length = L_SHORT; }
1380 else if (*fp == 'L')
1381 { fp++; length = L_LONGDOUBLE; }
1382 else if (*fp == 'l')
1384 { fp += 2; length = L_LONGLONG; }
1386 { fp++; length = L_LONG; }
1387 else if (*fp == 'z')
1388 { fp++; length = L_SIZE; }
1390 /* Handle each specific format type. */
1395 nptr = va_arg(ap, int *);
1396 *nptr = g->ptr - off;
1404 width = length > L_LONG ? 24 : 12;
1405 if ((need = g->ptr + width) > lim)
1407 if (!(flags & SVFMT_EXTEND) || need >= size_limit) return NULL;
1408 gstring_grow(g, width);
1410 gp = CS g->s + g->ptr;
1412 strncpy(newformat, item_start, fp - item_start);
1413 newformat[fp - item_start] = 0;
1415 /* Short int is promoted to int when passing through ..., so we must use
1416 int for va_arg(). */
1422 g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break;
1424 g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break;
1426 g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break;
1428 g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break;
1435 if ((need = g->ptr + 24) > lim)
1437 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1438 gstring_grow(g, 24);
1440 gp = CS g->s + g->ptr;
1442 /* sprintf() saying "(nil)" for a null pointer seems unreliable.
1443 Handle it explicitly. */
1444 if ((ptr = va_arg(ap, void *)))
1446 strncpy(newformat, item_start, fp - item_start);
1447 newformat[fp - item_start] = 0;
1448 g->ptr += sprintf(gp, newformat, ptr);
1451 g->ptr += sprintf(gp, "(nil)");
1455 /* %f format is inherently insecure if the numbers that it may be
1456 handed are unknown (e.g. 1e300). However, in Exim, %f is used for
1457 printing load averages, and these are actually stored as integers
1458 (load average * 1000) so the size of the numbers is constrained.
1459 It is also used for formatting sending rates, where the simplicity
1460 of the format prevents overflow. */
1467 if (precision < 0) precision = 6;
1468 if ((need = g->ptr + precision + 8) > lim)
1470 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1471 gstring_grow(g, precision+8);
1473 gp = CS g->s + g->ptr;
1475 strncpy(newformat, item_start, fp - item_start);
1476 newformat[fp-item_start] = 0;
1477 if (length == L_LONGDOUBLE)
1478 g->ptr += sprintf(gp, newformat, va_arg(ap, long double));
1480 g->ptr += sprintf(gp, newformat, va_arg(ap, double));
1486 if ((need = g->ptr + 1) > lim)
1488 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1492 g->s[g->ptr++] = (uschar) '%';
1496 if ((need = g->ptr + 1) > lim)
1498 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1502 g->s[g->ptr++] = (uschar) va_arg(ap, int);
1505 case 'D': /* Insert daily datestamp for log file names */
1506 s = CS tod_stamp(tod_log_datestamp_daily);
1507 string_datestamp_offset = g->ptr; /* Passed back via global */
1508 string_datestamp_length = Ustrlen(s); /* Passed back via global */
1509 string_datestamp_type = tod_log_datestamp_daily;
1510 slen = string_datestamp_length;
1513 case 'M': /* Insert monthly datestamp for log file names */
1514 s = CS tod_stamp(tod_log_datestamp_monthly);
1515 string_datestamp_offset = g->ptr; /* Passed back via global */
1516 string_datestamp_length = Ustrlen(s); /* Passed back via global */
1517 string_datestamp_type = tod_log_datestamp_monthly;
1518 slen = string_datestamp_length;
1522 case 'S': /* Forces *lower* case */
1523 case 'T': /* Forces *upper* case */
1524 s = va_arg(ap, char *);
1529 if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(s))
1530 if (flags & SVFMT_REBUFFER)
1532 gstring_rebuffer(g);
1533 gp = CS g->s + g->ptr;
1534 dest_tainted = TRUE;
1536 #ifndef MACRO_PREDEF
1538 die_tainted(US"string_vformat", func, line);
1541 INSERT_STRING: /* Come to from %D or %M above */
1544 BOOL truncated = FALSE;
1546 /* If the width is specified, check that there is a precision
1547 set; if not, set it to the width to prevent overruns of long
1552 if (precision < 0) precision = width;
1555 /* If a width is not specified and the precision is specified, set
1556 the width to the precision, or the string length if shorted. */
1558 else if (precision >= 0)
1559 width = precision < slen ? precision : slen;
1561 /* If neither are specified, set them both to the string length. */
1564 width = precision = slen;
1566 if ((need = g->ptr + width) >= size_limit || !(flags & SVFMT_EXTEND))
1568 if (g->ptr == lim) return NULL;
1572 width = precision = lim - g->ptr - 1;
1573 if (width < 0) width = 0;
1574 if (precision < 0) precision = 0;
1577 else if (need > lim)
1579 gstring_grow(g, width);
1581 gp = CS g->s + g->ptr;
1584 g->ptr += sprintf(gp, "%*.*s", width, precision, s);
1586 while (*gp) { *gp = tolower(*gp); gp++; }
1587 else if (fp[-1] == 'T')
1588 while (*gp) { *gp = toupper(*gp); gp++; }
1590 if (truncated) return NULL;
1594 /* Some things are never used in Exim; also catches junk. */
1597 strncpy(newformat, item_start, fp - item_start);
1598 newformat[fp-item_start] = 0;
1599 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type "
1600 "in \"%s\" in \"%s\"", newformat, format);
1605 if (g->ptr > g->size)
1606 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1607 "string_format internal error: caller %s %d", func, line);
1613 #ifndef COMPILE_UTILITY
1614 /*************************************************
1615 * Generate an "open failed" message *
1616 *************************************************/
1618 /* This function creates a message after failure to open a file. It includes a
1619 string supplied as data, adds the strerror() text, and if the failure was
1620 "Permission denied", reads and includes the euid and egid.
1623 format a text format string - deliberately not uschar *
1624 ... arguments for the format string
1626 Returns: a message, in dynamic store
1630 string_open_failed_trc(const uschar * func, unsigned line,
1631 const char *format, ...)
1634 gstring * g = string_get(1024);
1636 g = string_catn(g, US"failed to open ", 15);
1638 /* Use the checked formatting routine to ensure that the buffer
1639 does not overflow. It should not, since this is called only for internally
1640 specified messages. If it does, the message just gets truncated, and there
1641 doesn't seem much we can do about that. */
1643 va_start(ap, format);
1644 (void) string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE,
1645 SVFMT_REBUFFER, format, ap);
1648 g = string_catn(g, US": ", 2);
1649 g = string_cat(g, US strerror(errno));
1651 if (errno == EACCES)
1653 int save_errno = errno;
1654 g = string_fmt_append(g, " (euid=%ld egid=%ld)",
1655 (long int)geteuid(), (long int)getegid());
1658 gstring_release_unused(g);
1659 return string_from_gstring(g);
1666 /* qsort(3), currently used to sort the environment variables
1667 for -bP environment output, needs a function to compare two pointers to string
1668 pointers. Here it is. */
1671 string_compare_by_pointer(const void *a, const void *b)
1673 return Ustrcmp(* CUSS a, * CUSS b);
1675 #endif /* COMPILE_UTILITY */
1680 /*************************************************
1681 **************************************************
1682 * Stand-alone test program *
1683 **************************************************
1684 *************************************************/
1691 printf("Testing is_ip_address\n");
1693 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1696 buffer[Ustrlen(buffer) - 1] = 0;
1697 printf("%d\n", string_is_ip_address(buffer, NULL));
1698 printf("%d %d %s\n", string_is_ip_address(buffer, &offset), offset, buffer);
1701 printf("Testing string_nextinlist\n");
1703 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1705 uschar *list = buffer;
1713 sep1 = sep2 = list[1];
1720 uschar *item1 = string_nextinlist(&lp1, &sep1, item, sizeof(item));
1721 uschar *item2 = string_nextinlist(&lp2, &sep2, NULL, 0);
1723 if (item1 == NULL && item2 == NULL) break;
1724 if (item == NULL || item2 == NULL || Ustrcmp(item1, item2) != 0)
1726 printf("***ERROR\nitem1=\"%s\"\nitem2=\"%s\"\n",
1727 (item1 == NULL)? "NULL" : CS item1,
1728 (item2 == NULL)? "NULL" : CS item2);
1731 else printf(" \"%s\"\n", CS item1);
1735 /* This is a horrible lash-up, but it serves its purpose. */
1737 printf("Testing string_format\n");
1739 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1742 long long llargs[3];
1752 buffer[Ustrlen(buffer) - 1] = 0;
1754 s = Ustrchr(buffer, ',');
1755 if (s == NULL) s = buffer + Ustrlen(buffer);
1757 Ustrncpy(format, buffer, s - buffer);
1758 format[s-buffer] = 0;
1765 s = Ustrchr(ss, ',');
1766 if (s == NULL) s = ss + Ustrlen(ss);
1770 Ustrncpy(outbuf, ss, s-ss);
1771 if (Ustrchr(outbuf, '.') != NULL)
1774 dargs[n++] = Ustrtod(outbuf, NULL);
1776 else if (Ustrstr(outbuf, "ll") != NULL)
1779 llargs[n++] = strtoull(CS outbuf, NULL, 10);
1783 args[n++] = (void *)Uatoi(outbuf);
1787 else if (Ustrcmp(ss, "*") == 0)
1789 args[n++] = (void *)(&count);
1795 uschar *sss = malloc(s - ss + 1);
1796 Ustrncpy(sss, ss, s-ss);
1803 if (!dflag && !llflag)
1804 printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1805 args[0], args[1], args[2])? "True" : "False");
1808 printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1809 dargs[0], dargs[1], dargs[2])? "True" : "False");
1811 else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1812 llargs[0], llargs[1], llargs[2])? "True" : "False");
1814 printf("%s\n", CS outbuf);
1815 if (countset) printf("count=%d\n", count);
1822 /* End of string.c */