1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) University of Cambridge 1995 - 2018 */
6 /* Copyright (c) The Exim Maintainers 2020 - 2021 */
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 /* As above, but explicitly specifying the result taint status
440 string_copy_taint_function(const uschar * s, BOOL tainted)
442 return string_copy_taint(s, tainted);
447 /*************************************************
448 * Copy and save string, given length *
449 *************************************************/
451 /* It is assumed the data contains no zeros. A zero is added
456 n number of characters
458 Returns: copy of string in new store
462 string_copyn_function(const uschar * s, int n)
464 return string_copyn(s, n);
469 /*************************************************
470 * Copy and save string in malloc'd store *
471 *************************************************/
473 /* This function assumes that memcpy() is faster than strcpy().
475 Argument: string to copy
476 Returns: copy of string in new store
480 string_copy_malloc(const uschar * s)
482 int len = Ustrlen(s) + 1;
483 uschar * ss = store_malloc(len);
490 /*************************************************
491 * Copy string if long, inserting newlines *
492 *************************************************/
494 /* If the given string is longer than 75 characters, it is copied, and within
495 the copy, certain space characters are converted into newlines.
497 Argument: pointer to the string
498 Returns: pointer to the possibly altered string
502 string_split_message(uschar * msg)
506 if (!msg || Ustrlen(msg) <= 75) return msg;
507 s = ss = msg = string_copy(msg);
512 while (i < 75 && *ss && *ss != '\n') ss++, i++;
524 if (t[-1] == ':') { tt = t; break; }
529 if (!tt) /* Can't split behind - try ahead */
534 if (*t == ' ' || *t == '\n')
540 if (!tt) break; /* Can't find anywhere to split */
551 /*************************************************
552 * Copy returned DNS domain name, de-escaping *
553 *************************************************/
555 /* If a domain name contains top-bit characters, some resolvers return
556 the fully qualified name with those characters turned into escapes. The
557 convention is a backslash followed by _decimal_ digits. We convert these
558 back into the original binary values. This will be relevant when
559 allow_utf8_domains is set true and UTF-8 characters are used in domain
560 names. Backslash can also be used to escape other characters, though we
561 shouldn't come across them in domain names.
563 Argument: the domain name string
564 Returns: copy of string in new store, de-escaped
568 string_copy_dnsdomain(uschar * s)
571 uschar * ss = yield = store_get(Ustrlen(s) + 1, TRUE); /* always treat as tainted */
577 else if (isdigit(s[1]))
579 *ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0';
591 #ifndef COMPILE_UTILITY
592 /*************************************************
593 * Copy space-terminated or quoted string *
594 *************************************************/
596 /* This function copies from a string until its end, or until whitespace is
597 encountered, unless the string begins with a double quote, in which case the
598 terminating quote is sought, and escaping within the string is done. The length
599 of a de-quoted string can be no longer than the original, since escaping always
600 turns n characters into 1 character.
602 Argument: pointer to the pointer to the first character, which gets updated
603 Returns: the new string
607 string_dequote(const uschar ** sptr)
609 const uschar * s = * sptr;
612 /* First find the end of the string */
615 while (*s && !isspace(*s)) s++;
619 while (*s && *s != '\"')
621 if (*s == '\\') (void)string_interpret_escape(&s);
627 /* Get enough store to copy into */
629 t = yield = store_get(s - *sptr + 1, is_tainted(*sptr));
635 while (*s && !isspace(*s)) *t++ = *s++;
639 while (*s && *s != '\"')
641 *t++ = *s == '\\' ? string_interpret_escape(&s) : *s;
647 /* Update the pointer and return the terminated copy */
653 #endif /* COMPILE_UTILITY */
657 /*************************************************
658 * Format a string and save it *
659 *************************************************/
661 /* The formatting is done by string_vformat, which checks the length of
662 everything. Taint is taken from the worst of the arguments.
665 format a printf() format - deliberately char * rather than uschar *
666 because it will most usually be a literal string
667 func caller, for debug
668 line caller, for debug
669 ... arguments for format
671 Returns: pointer to fresh piece of store containing sprintf'ed string
675 string_sprintf_trc(const char * format, const uschar * func, unsigned line, ...)
677 #ifdef COMPILE_UTILITY
678 uschar buffer[STRING_SPRINTF_BUFFER_SIZE];
679 gstring gs = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer };
684 unsigned flags = SVFMT_REBUFFER|SVFMT_EXTEND;
689 g = string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE,
694 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
695 "string_sprintf expansion was longer than %d; format string was (%s)\n"
696 " called from %s %d\n",
697 STRING_SPRINTF_BUFFER_SIZE, format, func, line);
699 #ifdef COMPILE_UTILITY
700 return string_copyn(g->s, g->ptr);
702 gstring_release_unused(g);
703 return string_from_gstring(g);
709 /*************************************************
710 * Case-independent strncmp() function *
711 *************************************************/
717 n number of characters to compare
719 Returns: < 0, = 0, or > 0, according to the comparison
723 strncmpic(const uschar * s, const uschar * t, int n)
727 int c = tolower(*s++) - tolower(*t++);
734 /*************************************************
735 * Case-independent strcmp() function *
736 *************************************************/
743 Returns: < 0, = 0, or > 0, according to the comparison
747 strcmpic(const uschar * s, const uschar * t)
751 int c = tolower(*s++) - tolower(*t++);
752 if (c != 0) return c;
758 /*************************************************
759 * Case-independent strstr() function *
760 *************************************************/
762 /* The third argument specifies whether whitespace is required
763 to follow the matched string.
767 t substring to search for
768 space_follows if TRUE, match only if whitespace follows
770 Returns: pointer to substring in string, or NULL if not found
774 strstric(uschar * s, uschar * t, BOOL space_follows)
777 uschar * yield = NULL;
778 int cl = tolower(*p);
779 int cu = toupper(*p);
783 if (*s == cl || *s == cu)
785 if (!yield) yield = s;
788 if (!space_follows || s[1] == ' ' || s[1] == '\n' ) return yield;
810 #ifdef COMPILE_UTILITY
811 /* Dummy version for this function; it should never be called */
813 gstring_grow(gstring * g, int count)
821 #ifndef COMPILE_UTILITY
822 /*************************************************
823 * Get next string from separated list *
824 *************************************************/
826 /* Leading and trailing space is removed from each item. The separator in the
827 list is controlled by the int pointed to by the separator argument as follows:
829 If the value is > 0 it is used as the separator. This is typically used for
830 sublists such as slash-separated options. The value is always a printing
833 (If the value is actually > UCHAR_MAX there is only one item in the list.
834 This is used for some cases when called via functions that sometimes
835 plough through lists, and sometimes are given single items.)
837 If the value is <= 0, the string is inspected for a leading <x, where x is an
838 ispunct() or an iscntrl() character. If found, x is used as the separator. If
841 (a) if separator == 0, ':' is used
842 (b) if separator <0, -separator is used
844 In all cases the value of the separator that is used is written back to the
845 int so that it is used on subsequent calls as we progress through the list.
847 A literal ispunct() separator can be represented in an item by doubling, but
848 there is no way to include an iscntrl() separator as part of the data.
851 listptr points to a pointer to the current start of the list; the
852 pointer gets updated to point after the end of the next item
853 separator a pointer to the separator character in an int (see above)
854 buffer where to put a copy of the next string in the list; or
855 NULL if the next string is returned in new memory
856 Note that if the list is tainted then a provided buffer must be
857 also (else we trap, with a message referencing the callsite).
858 If we do the allocation, taint is handled there.
859 buflen when buffer is not NULL, the size of buffer; otherwise ignored
861 func caller, for debug
862 line caller, for debug
864 Returns: pointer to buffer, containing the next substring,
865 or NULL if no more substrings
869 string_nextinlist_trc(const uschar ** listptr, int * separator, uschar * buffer,
870 int buflen, const uschar * func, int line)
872 int sep = *separator;
873 const uschar * s = *listptr;
878 /* This allows for a fixed specified separator to be an iscntrl() character,
879 but at the time of implementation, this is never the case. However, it's best
880 to be conservative. */
882 while (isspace(*s) && *s != sep) s++;
884 /* A change of separator is permitted, so look for a leading '<' followed by an
885 allowed character. */
889 if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1])))
893 while (isspace(*s) && *s != sep) s++;
896 sep = sep ? -sep : ':';
900 /* An empty string has no list elements */
902 if (!*s) return NULL;
904 /* Note whether whether or not the separator is an iscntrl() character. */
906 sep_is_special = iscntrl(sep);
908 /* Handle the case when a buffer is provided. */
913 if (is_tainted(s) && !is_tainted(buffer))
914 die_tainted(US"string_nextinlist", func, line);
917 if (*s == sep && (*(++s) != sep || sep_is_special)) break;
918 if (p < buflen - 1) buffer[p++] = *s;
920 while (p > 0 && isspace(buffer[p-1])) p--;
924 /* Handle the case when a buffer is not provided. */
930 /* We know that *s != 0 at this point. However, it might be pointing to a
931 separator, which could indicate an empty string, or (if an ispunct()
932 character) could be doubled to indicate a separator character as data at the
933 start of a string. Avoid getting working memory for an empty item. */
936 if (*++s != sep || sep_is_special)
939 return string_copy(US"");
942 /* Not an empty string; the first character is guaranteed to be a data
948 for (ss = s + 1; *ss && *ss != sep; ) ss++;
949 g = string_catn(g, s, ss-s);
951 if (!*s || *++s != sep || sep_is_special) break;
954 /* Trim trailing spaces from the returned string */
956 /* while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; */
957 while ( g->ptr > 0 && isspace(g->s[g->ptr-1])
958 && (g->ptr == 1 || g->s[g->ptr-2] != '\\') )
960 buffer = string_from_gstring(g);
961 gstring_release_unused_trc(g, CCS func, line);
964 /* Update the current pointer and return the new string */
971 static const uschar *
972 Ustrnchr(const uschar * s, int c, unsigned * len)
977 if (!*s) return NULL;
990 /************************************************
991 * Add element to separated list *
992 ************************************************/
993 /* This function is used to build a list, returning an allocated null-terminated
994 growable string. The given element has any embedded separator characters
997 Despite having the same growable-string interface as string_cat() the list is
998 always returned null-terminated.
1001 list expanding-string for the list that is being built, or NULL
1002 if this is a new list that has no contents yet
1003 sep list separator character
1004 ele new element to be appended to the list
1006 Returns: pointer to the start of the list, changed if copied for expansion.
1010 string_append_listele(gstring * list, uschar sep, const uschar * ele)
1014 if (list && list->ptr)
1015 list = string_catn(list, &sep, 1);
1017 while((sp = Ustrchr(ele, sep)))
1019 list = string_catn(list, ele, sp-ele+1);
1020 list = string_catn(list, &sep, 1);
1023 list = string_cat(list, ele);
1024 (void) string_from_gstring(list);
1030 string_append_listele_n(gstring * list, uschar sep, const uschar * ele,
1035 if (list && list->ptr)
1036 list = string_catn(list, &sep, 1);
1038 while((sp = Ustrnchr(ele, sep, &len)))
1040 list = string_catn(list, ele, sp-ele+1);
1041 list = string_catn(list, &sep, 1);
1045 list = string_catn(list, ele, len);
1046 (void) string_from_gstring(list);
1052 /* A slightly-bogus listmaker utility; the separator is a string so
1053 can be multiple chars - there is no checking for the element content
1054 containing any of the separator. */
1057 string_append2_listele_n(gstring * list, const uschar * sepstr,
1058 const uschar * ele, unsigned len)
1060 if (list && list->ptr)
1061 list = string_cat(list, sepstr);
1063 list = string_catn(list, ele, len);
1064 (void) string_from_gstring(list);
1070 /************************************************/
1071 /* Add more space to a growable-string. The caller should check
1072 first if growth is required. The gstring struct is modified on
1073 return; specifically, the string-base-pointer may have been changed.
1076 g the growable-string
1077 count amount needed for g->ptr to increase by
1081 gstring_grow(gstring * g, int count)
1084 int oldsize = g->size;
1085 BOOL tainted = is_tainted(g->s);
1087 /* Mostly, string_cat() is used to build small strings of a few hundred
1088 characters at most. There are times, however, when the strings are very much
1089 longer (for example, a lookup that returns a vast number of alias addresses).
1090 To try to keep things reasonable, we use increments whose size depends on the
1091 existing length of the string. */
1093 unsigned inc = oldsize < 4096 ? 127 : 1023;
1095 if (g->ptr < 0 || g->ptr > g->size || g->size >= INT_MAX/2)
1096 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1097 "internal error in gstring_grow (ptr %d size %d)", g->ptr, g->size);
1099 if (count <= 0) return;
1101 if (count >= INT_MAX/2 - g->ptr)
1102 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1103 "internal error in gstring_grow (ptr %d count %d)", g->ptr, count);
1105 g->size = (p + count + inc + 1) & ~inc; /* one for a NUL */
1107 /* Try to extend an existing allocation. If the result of calling
1108 store_extend() is false, either there isn't room in the current memory block,
1109 or this string is not the top item on the dynamic store stack. We then have
1110 to get a new chunk of store and copy the old string. When building large
1111 strings, it is helpful to call store_release() on the old string, to release
1112 memory blocks that have become empty. (The block will be freed if the string
1113 is at its start.) However, we can do this only if we know that the old string
1114 was the last item on the dynamic memory stack. This is the case if it matches
1117 if (!store_extend(g->s, tainted, oldsize, g->size))
1118 g->s = store_newblock(g->s, tainted, g->size, p);
1123 /*************************************************
1124 * Add chars to string *
1125 *************************************************/
1126 /* This function is used when building up strings of unknown length. Room is
1127 always left for a terminating zero to be added to the string that is being
1128 built. This function does not require the string that is being added to be NUL
1129 terminated, because the number of characters to add is given explicitly. It is
1130 sometimes called to extract parts of other strings.
1133 g growable-string that is being built, or NULL if not assigned yet
1134 s points to characters to add
1135 count count of characters to add; must not exceed the length of s, if s
1138 Returns: growable string, changed if copied for expansion.
1139 Note that a NUL is not added, though space is left for one. This is
1140 because string_cat() is often called multiple times to build up a
1141 string - there's no point adding the NUL till the end.
1144 /* coverity[+alloc] */
1147 string_catn(gstring * g, const uschar *s, int count)
1150 BOOL srctaint = is_tainted(s);
1153 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1154 "internal error in string_catn (count %d)", count);
1158 unsigned inc = count < 4096 ? 127 : 1023;
1159 unsigned size = ((count + inc) & ~inc) + 1; /* round up requested count */
1160 g = string_get_tainted(size, srctaint);
1162 else if (srctaint && !is_tainted(g->s))
1163 gstring_rebuffer(g);
1165 if (g->ptr < 0 || g->ptr > g->size)
1166 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1167 "internal error in string_catn (ptr %d size %d)", g->ptr, g->size);
1170 if (count >= g->size - p)
1171 gstring_grow(g, count);
1173 /* Because we always specify the exact number of characters to copy, we can
1174 use memcpy(), which is likely to be more efficient than strncopy() because the
1175 latter has to check for zero bytes. */
1177 memcpy(g->s + p, s, count);
1184 string_cat(gstring * g, const uschar * s)
1186 return string_catn(g, s, Ustrlen(s));
1191 /*************************************************
1192 * Append strings to another string *
1193 *************************************************/
1195 /* This function can be used to build a string from many other strings.
1196 It calls string_cat() to do the dirty work.
1199 g growable-string that is being built, or NULL if not yet assigned
1200 count the number of strings to append
1201 ... "count" uschar* arguments, which must be valid zero-terminated
1204 Returns: growable string, changed if copied for expansion.
1205 The string is not zero-terminated - see string_cat() above.
1208 __inline__ gstring *
1209 string_append(gstring * g, int count, ...)
1213 va_start(ap, count);
1216 uschar * t = va_arg(ap, uschar *);
1217 g = string_cat(g, t);
1227 /*************************************************
1228 * Format a string with length checks *
1229 *************************************************/
1231 /* This function is used to format a string with checking of the length of the
1232 output for all conversions. It protects Exim from absent-mindedness when
1233 calling functions like debug_printf and string_sprintf, and elsewhere. There
1234 are two different entry points to what is actually the same function, depending
1235 on whether the variable length list of data arguments are given explicitly or
1238 The formats are the usual printf() ones, with some omissions (never used) and
1239 three additions for strings: %S forces lower case, %T forces upper case, and
1240 %#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed
1241 (useful in debugging). There is also the addition of %D and %M, which insert
1242 the date in the form used for datestamped log files.
1245 buffer a buffer in which to put the formatted string
1246 buflen the length of the buffer
1247 format the format string - deliberately char * and not uschar *
1248 ... or ap variable list of supplementary arguments
1250 Returns: TRUE if the result fitted in the buffer
1254 string_format_trc(uschar * buffer, int buflen,
1255 const uschar * func, unsigned line, const char * format, ...)
1257 gstring g = { .size = buflen, .ptr = 0, .s = buffer }, * gp;
1259 va_start(ap, format);
1260 gp = string_vformat_trc(&g, func, line, STRING_SPRINTF_BUFFER_SIZE,
1270 /* Build or append to a growing-string, sprintf-style.
1274 func called-from function name, for debug
1275 line called-from file line number, for debug
1276 limit maximum string size
1278 format printf-like format string
1279 ap variable-args pointer
1282 SVFMT_EXTEND buffer can be created or exteded as needed
1283 SVFMT_REBUFFER buffer can be recopied to tainted mem as needed
1284 SVFMT_TAINT_NOCHK do not check inputs for taint
1286 If the "extend" flag is true, the string passed in can be NULL,
1287 empty, or non-empty. Growing is subject to an overall limit given
1288 by the limit argument.
1290 If the "extend" flag is false, the string passed in may not be NULL,
1291 will not be grown, and is usable in the original place after return.
1292 The return value can be NULL to signify overflow.
1294 Returns the possibly-new (if copy for growth or taint-handling was needed)
1295 string, not nul-terminated.
1299 string_vformat_trc(gstring * g, const uschar * func, unsigned line,
1300 unsigned size_limit, unsigned flags, const char * format, va_list ap)
1302 enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 };
1304 int width, precision, off, lim, need;
1305 const char * fp = format; /* Deliberately not unsigned */
1306 BOOL dest_tainted = FALSE;
1308 string_datestamp_offset = -1; /* Datestamp not inserted */
1309 string_datestamp_length = 0; /* Datestamp not inserted */
1310 string_datestamp_type = 0; /* Datestamp not inserted */
1312 #ifdef COMPILE_UTILITY
1313 assert(!(flags & SVFMT_EXTEND));
1317 /* Ensure we have a string, to save on checking later */
1318 if (!g) g = string_get(16);
1319 else if (!(flags & SVFMT_TAINT_NOCHK)) dest_tainted = is_tainted(g->s);
1321 if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(format))
1323 #ifndef MACRO_PREDEF
1324 if (!(flags & SVFMT_REBUFFER))
1325 die_tainted(US"string_vformat", func, line);
1327 gstring_rebuffer(g);
1328 dest_tainted = TRUE;
1330 #endif /*!COMPILE_UTILITY*/
1332 lim = g->size - 1; /* leave one for a nul */
1333 off = g->ptr; /* remember initial offset in gstring */
1335 /* Scan the format and handle the insertions */
1339 int length = L_NORMAL;
1342 const char *null = "NULL"; /* ) These variables */
1343 const char *item_start, *s; /* ) are deliberately */
1344 char newformat[16]; /* ) not unsigned */
1345 char * gp = CS g->s + g->ptr; /* ) */
1347 /* Non-% characters just get copied verbatim */
1351 /* Avoid string_copyn() due to COMPILE_UTILITY */
1352 if ((need = g->ptr + 1) > lim)
1354 if (!(flags & SVFMT_EXTEND) || need > size_limit) return NULL;
1358 g->s[g->ptr++] = (uschar) *fp++;
1362 /* Deal with % characters. Pick off the width and precision, for checking
1363 strings, skipping over the flag and modifier characters. */
1366 width = precision = -1;
1368 if (strchr("-+ #0", *(++fp)) != NULL)
1370 if (*fp == '#') null = "";
1374 if (isdigit((uschar)*fp))
1376 width = *fp++ - '0';
1377 while (isdigit((uschar)*fp)) width = width * 10 + *fp++ - '0';
1379 else if (*fp == '*')
1381 width = va_arg(ap, int);
1388 precision = va_arg(ap, int);
1392 for (precision = 0; isdigit((uschar)*fp); fp++)
1393 precision = precision*10 + *fp - '0';
1395 /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */
1398 { fp++; length = L_SHORT; }
1399 else if (*fp == 'L')
1400 { fp++; length = L_LONGDOUBLE; }
1401 else if (*fp == 'l')
1403 { fp += 2; length = L_LONGLONG; }
1405 { fp++; length = L_LONG; }
1406 else if (*fp == 'z')
1407 { fp++; length = L_SIZE; }
1409 /* Handle each specific format type. */
1414 nptr = va_arg(ap, int *);
1415 *nptr = g->ptr - off;
1423 width = length > L_LONG ? 24 : 12;
1424 if ((need = g->ptr + width) > lim)
1426 if (!(flags & SVFMT_EXTEND) || need >= size_limit) return NULL;
1427 gstring_grow(g, width);
1429 gp = CS g->s + g->ptr;
1431 strncpy(newformat, item_start, fp - item_start);
1432 newformat[fp - item_start] = 0;
1434 /* Short int is promoted to int when passing through ..., so we must use
1435 int for va_arg(). */
1441 g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break;
1443 g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break;
1445 g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break;
1447 g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break;
1454 if ((need = g->ptr + 24) > lim)
1456 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1457 gstring_grow(g, 24);
1459 gp = CS g->s + g->ptr;
1461 /* sprintf() saying "(nil)" for a null pointer seems unreliable.
1462 Handle it explicitly. */
1463 if ((ptr = va_arg(ap, void *)))
1465 strncpy(newformat, item_start, fp - item_start);
1466 newformat[fp - item_start] = 0;
1467 g->ptr += sprintf(gp, newformat, ptr);
1470 g->ptr += sprintf(gp, "(nil)");
1474 /* %f format is inherently insecure if the numbers that it may be
1475 handed are unknown (e.g. 1e300). However, in Exim, %f is used for
1476 printing load averages, and these are actually stored as integers
1477 (load average * 1000) so the size of the numbers is constrained.
1478 It is also used for formatting sending rates, where the simplicity
1479 of the format prevents overflow. */
1486 if (precision < 0) precision = 6;
1487 if ((need = g->ptr + precision + 8) > lim)
1489 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1490 gstring_grow(g, precision+8);
1492 gp = CS g->s + g->ptr;
1494 strncpy(newformat, item_start, fp - item_start);
1495 newformat[fp-item_start] = 0;
1496 if (length == L_LONGDOUBLE)
1497 g->ptr += sprintf(gp, newformat, va_arg(ap, long double));
1499 g->ptr += sprintf(gp, newformat, va_arg(ap, double));
1505 if ((need = g->ptr + 1) > lim)
1507 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1511 g->s[g->ptr++] = (uschar) '%';
1515 if ((need = g->ptr + 1) > lim)
1517 if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL;
1521 g->s[g->ptr++] = (uschar) va_arg(ap, int);
1524 case 'D': /* Insert daily datestamp for log file names */
1525 s = CS tod_stamp(tod_log_datestamp_daily);
1526 string_datestamp_offset = g->ptr; /* Passed back via global */
1527 string_datestamp_length = Ustrlen(s); /* Passed back via global */
1528 string_datestamp_type = tod_log_datestamp_daily;
1529 slen = string_datestamp_length;
1532 case 'M': /* Insert monthly datestamp for log file names */
1533 s = CS tod_stamp(tod_log_datestamp_monthly);
1534 string_datestamp_offset = g->ptr; /* Passed back via global */
1535 string_datestamp_length = Ustrlen(s); /* Passed back via global */
1536 string_datestamp_type = tod_log_datestamp_monthly;
1537 slen = string_datestamp_length;
1541 case 'S': /* Forces *lower* case */
1542 case 'T': /* Forces *upper* case */
1543 s = va_arg(ap, char *);
1548 if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(s))
1549 if (flags & SVFMT_REBUFFER)
1551 gstring_rebuffer(g);
1552 gp = CS g->s + g->ptr;
1553 dest_tainted = TRUE;
1555 #ifndef MACRO_PREDEF
1557 die_tainted(US"string_vformat", func, line);
1560 INSERT_STRING: /* Come to from %D or %M above */
1563 BOOL truncated = FALSE;
1565 /* If the width is specified, check that there is a precision
1566 set; if not, set it to the width to prevent overruns of long
1571 if (precision < 0) precision = width;
1574 /* If a width is not specified and the precision is specified, set
1575 the width to the precision, or the string length if shorted. */
1577 else if (precision >= 0)
1578 width = precision < slen ? precision : slen;
1580 /* If neither are specified, set them both to the string length. */
1583 width = precision = slen;
1585 if ((need = g->ptr + width) >= size_limit || !(flags & SVFMT_EXTEND))
1587 if (g->ptr == lim) return NULL;
1591 width = precision = lim - g->ptr - 1;
1592 if (width < 0) width = 0;
1593 if (precision < 0) precision = 0;
1596 else if (need > lim)
1598 gstring_grow(g, width);
1600 gp = CS g->s + g->ptr;
1603 g->ptr += sprintf(gp, "%*.*s", width, precision, s);
1605 while (*gp) { *gp = tolower(*gp); gp++; }
1606 else if (fp[-1] == 'T')
1607 while (*gp) { *gp = toupper(*gp); gp++; }
1609 if (truncated) return NULL;
1613 /* Some things are never used in Exim; also catches junk. */
1616 strncpy(newformat, item_start, fp - item_start);
1617 newformat[fp-item_start] = 0;
1618 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type "
1619 "in \"%s\" in \"%s\"", newformat, format);
1624 if (g->ptr > g->size)
1625 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
1626 "string_format internal error: caller %s %d", func, line);
1632 #ifndef COMPILE_UTILITY
1633 /*************************************************
1634 * Generate an "open failed" message *
1635 *************************************************/
1637 /* This function creates a message after failure to open a file. It includes a
1638 string supplied as data, adds the strerror() text, and if the failure was
1639 "Permission denied", reads and includes the euid and egid.
1642 format a text format string - deliberately not uschar *
1643 func caller, for debug
1644 line caller, for debug
1645 ... arguments for the format string
1647 Returns: a message, in dynamic store
1651 string_open_failed_trc(const uschar * func, unsigned line,
1652 const char * format, ...)
1655 gstring * g = string_get(1024);
1657 g = string_catn(g, US"failed to open ", 15);
1659 /* Use the checked formatting routine to ensure that the buffer
1660 does not overflow. It should not, since this is called only for internally
1661 specified messages. If it does, the message just gets truncated, and there
1662 doesn't seem much we can do about that. */
1664 va_start(ap, format);
1665 (void) string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE,
1666 SVFMT_REBUFFER, format, ap);
1669 g = string_catn(g, US": ", 2);
1670 g = string_cat(g, US strerror(errno));
1672 if (errno == EACCES)
1674 int save_errno = errno;
1675 g = string_fmt_append(g, " (euid=%ld egid=%ld)",
1676 (long int)geteuid(), (long int)getegid());
1679 gstring_release_unused(g);
1680 return string_from_gstring(g);
1687 /* qsort(3), currently used to sort the environment variables
1688 for -bP environment output, needs a function to compare two pointers to string
1689 pointers. Here it is. */
1692 string_compare_by_pointer(const void *a, const void *b)
1694 return Ustrcmp(* CUSS a, * CUSS b);
1696 #endif /* COMPILE_UTILITY */
1701 /*************************************************
1702 **************************************************
1703 * Stand-alone test program *
1704 **************************************************
1705 *************************************************/
1712 printf("Testing is_ip_address\n");
1715 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1718 buffer[Ustrlen(buffer) - 1] = 0;
1719 printf("%d\n", string_is_ip_address(buffer, NULL));
1720 printf("%d %d %s\n", string_is_ip_address(buffer, &offset), offset, buffer);
1723 printf("Testing string_nextinlist\n");
1725 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1727 uschar *list = buffer;
1735 sep1 = sep2 = list[1];
1742 uschar *item1 = string_nextinlist(&lp1, &sep1, item, sizeof(item));
1743 uschar *item2 = string_nextinlist(&lp2, &sep2, NULL, 0);
1745 if (item1 == NULL && item2 == NULL) break;
1746 if (item == NULL || item2 == NULL || Ustrcmp(item1, item2) != 0)
1748 printf("***ERROR\nitem1=\"%s\"\nitem2=\"%s\"\n",
1749 (item1 == NULL)? "NULL" : CS item1,
1750 (item2 == NULL)? "NULL" : CS item2);
1753 else printf(" \"%s\"\n", CS item1);
1757 /* This is a horrible lash-up, but it serves its purpose. */
1759 printf("Testing string_format\n");
1761 while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
1764 long long llargs[3];
1774 buffer[Ustrlen(buffer) - 1] = 0;
1776 s = Ustrchr(buffer, ',');
1777 if (s == NULL) s = buffer + Ustrlen(buffer);
1779 Ustrncpy(format, buffer, s - buffer);
1780 format[s-buffer] = 0;
1787 s = Ustrchr(ss, ',');
1788 if (s == NULL) s = ss + Ustrlen(ss);
1792 Ustrncpy(outbuf, ss, s-ss);
1793 if (Ustrchr(outbuf, '.') != NULL)
1796 dargs[n++] = Ustrtod(outbuf, NULL);
1798 else if (Ustrstr(outbuf, "ll") != NULL)
1801 llargs[n++] = strtoull(CS outbuf, NULL, 10);
1805 args[n++] = (void *)Uatoi(outbuf);
1809 else if (Ustrcmp(ss, "*") == 0)
1811 args[n++] = (void *)(&count);
1817 uschar *sss = malloc(s - ss + 1);
1818 Ustrncpy(sss, ss, s-ss);
1825 if (!dflag && !llflag)
1826 printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1827 args[0], args[1], args[2])? "True" : "False");
1830 printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1831 dargs[0], dargs[1], dargs[2])? "True" : "False");
1833 else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
1834 llargs[0], llargs[1], llargs[2])? "True" : "False");
1836 printf("%s\n", CS outbuf);
1837 if (countset) printf("count=%d\n", count);
1844 /* End of string.c */