Fix buffer overflow in string_vformat. CVE-2019-16928 Bug 2449

[exim.git] / src / src / string.c

/*************************************************
*     Exim - an Internet mail transport agent    *
*************************************************/

/* Copyright (c) University of Cambridge 1995 - 2018 */
/* See the file NOTICE for conditions of use and distribution. */

/* Miscellaneous string-handling functions. Some are not required for
utilities and tests, and are cut out by the COMPILE_UTILITY macro. */


#include "exim.h"
#include <assert.h>


#ifndef COMPILE_UTILITY
/*************************************************
*            Test for IP address                 *
*************************************************/

/* This used just to be a regular expression, but with IPv6 things are a bit
more complicated. If the address contains a colon, it is assumed to be a v6
address (assuming HAVE_IPV6 is set). If a mask is permitted and one is present,
and maskptr is not NULL, its offset is placed there.

Arguments:
  s         a string
  maskptr   NULL if no mask is permitted to follow
            otherwise, points to an int where the offset of '/' is placed
            if there is no / followed by trailing digits, *maskptr is set 0

Returns:    0 if the string is not a textual representation of an IP address
            4 if it is an IPv4 address
            6 if it is an IPv6 address
*/

int
string_is_ip_address(const uschar *s, int *maskptr)
{
int i;
int yield = 4;

/* If an optional mask is permitted, check for it. If found, pass back the
offset. */

if (maskptr)
  {
  const uschar *ss = s + Ustrlen(s);
  *maskptr = 0;
  if (s != ss && isdigit(*(--ss)))
    {
    while (ss > s && isdigit(ss[-1])) ss--;
    if (ss > s && *(--ss) == '/') *maskptr = ss - s;
    }
  }

/* A colon anywhere in the string => IPv6 address */

if (Ustrchr(s, ':') != NULL)
  {
  BOOL had_double_colon = FALSE;
  BOOL v4end = FALSE;
  int count = 0;

  yield = 6;

  /* An IPv6 address must start with hex digit or double colon. A single
  colon is invalid. */

  if (*s == ':' && *(++s) != ':') return 0;

  /* Now read up to 8 components consisting of up to 4 hex digits each. There
  may be one and only one appearance of double colon, which implies any number
  of binary zero bits. The number of preceding components is held in count. */

  for (count = 0; count < 8; count++)
    {
    /* If the end of the string is reached before reading 8 components, the
    address is valid provided a double colon has been read. This also applies
    if we hit the / that introduces a mask or the % that introduces the
    interface specifier (scope id) of a link-local address. */

    if (*s == 0 || *s == '%' || *s == '/') return had_double_colon ? yield : 0;

    /* If a component starts with an additional colon, we have hit a double
    colon. This is permitted to appear once only, and counts as at least
    one component. The final component may be of this form. */

    if (*s == ':')
      {
      if (had_double_colon) return 0;
      had_double_colon = TRUE;
      s++;
      continue;
      }

    /* If the remainder of the string contains a dot but no colons, we
    can expect a trailing IPv4 address. This is valid if either there has
    been no double-colon and this is the 7th component (with the IPv4 address
    being the 7th & 8th components), OR if there has been a double-colon
    and fewer than 6 components. */

    if (Ustrchr(s, ':') == NULL && Ustrchr(s, '.') != NULL)
      {
      if ((!had_double_colon && count != 6) ||
          (had_double_colon && count > 6)) return 0;
      v4end = TRUE;
      yield = 6;
      break;
      }

    /* Check for at least one and not more than 4 hex digits for this
    component. */

    if (!isxdigit(*s++)) return 0;
    if (isxdigit(*s) && isxdigit(*(++s)) && isxdigit(*(++s))) s++;

    /* If the component is terminated by colon and there is more to
    follow, skip over the colon. If there is no more to follow the address is
    invalid. */

    if (*s == ':' && *(++s) == 0) return 0;
    }

  /* If about to handle a trailing IPv4 address, drop through. Otherwise
  all is well if we are at the end of the string or at the mask or at a percent
  sign, which introduces the interface specifier (scope id) of a link local
  address. */

  if (!v4end)
    return (*s == 0 || *s == '%' ||
           (*s == '/' && maskptr != NULL && *maskptr != 0))? yield : 0;
  }

/* Test for IPv4 address, which may be the tail-end of an IPv6 address. */

for (i = 0; i < 4; i++)
  {
  long n;
  uschar * end;

  if (i != 0 && *s++ != '.') return 0;
  n = strtol(CCS s, CSS &end, 10);
  if (n > 255 || n < 0 || end <= s || end > s+3) return 0;
  s = end;
  }

return !*s || (*s == '/' && maskptr && *maskptr != 0) ? yield : 0;
}
#endif  /* COMPILE_UTILITY */


/*************************************************
*              Format message size               *
*************************************************/

/* Convert a message size in bytes to printing form, rounding
according to the magnitude of the number. A value of zero causes
a string of spaces to be returned.

Arguments:
  size        the message size in bytes
  buffer      where to put the answer

Returns:      pointer to the buffer
              a string of exactly 5 characters is normally returned
*/

uschar *
string_format_size(int size, uschar *buffer)
{
if (size == 0) Ustrcpy(buffer, "     ");
else if (size < 1024) sprintf(CS buffer, "%5d", size);
else if (size < 10*1024)
  sprintf(CS buffer, "%4.1fK", (double)size / 1024.0);
else if (size < 1024*1024)
  sprintf(CS buffer, "%4dK", (size + 512)/1024);
else if (size < 10*1024*1024)
  sprintf(CS buffer, "%4.1fM", (double)size / (1024.0 * 1024.0));
else
  sprintf(CS buffer, "%4dM", (size + 512 * 1024)/(1024*1024));
return buffer;
}



#ifndef COMPILE_UTILITY
/*************************************************
*       Convert a number to base 62 format       *
*************************************************/

/* Convert a long integer into an ASCII base 62 string. For Cygwin the value of
BASE_62 is actually 36. Always return exactly 6 characters plus zero, in a
static area.

Argument: a long integer
Returns:  pointer to base 62 string
*/

uschar *
string_base62(unsigned long int value)
{
static uschar yield[7];
uschar *p = yield + sizeof(yield) - 1;
*p = 0;
while (p > yield)
  {
  *(--p) = base62_chars[value % BASE_62];
  value /= BASE_62;
  }
return yield;
}
#endif  /* COMPILE_UTILITY */



/*************************************************
*          Interpret escape sequence             *
*************************************************/

/* This function is called from several places where escape sequences are to be
interpreted in strings.

Arguments:
  pp       points a pointer to the initiating "\" in the string;
           the pointer gets updated to point to the final character
           If the backslash is the last character in the string, it
           is not interpreted.
Returns:   the value of the character escape
*/

int
string_interpret_escape(const uschar **pp)
{
#ifdef COMPILE_UTILITY
const uschar *hex_digits= CUS"0123456789abcdef";
#endif
int ch;
const uschar *p = *pp;
ch = *(++p);
if (ch == '\0') return **pp;
if (isdigit(ch) && ch != '8' && ch != '9')
  {
  ch -= '0';
  if (isdigit(p[1]) && p[1] != '8' && p[1] != '9')
    {
    ch = ch * 8 + *(++p) - '0';
    if (isdigit(p[1]) && p[1] != '8' && p[1] != '9')
      ch = ch * 8 + *(++p) - '0';
    }
  }
else switch(ch)
  {
  case 'b':  ch = '\b'; break;
  case 'f':  ch = '\f'; break;
  case 'n':  ch = '\n'; break;
  case 'r':  ch = '\r'; break;
  case 't':  ch = '\t'; break;
  case 'v':  ch = '\v'; break;
  case 'x':
  ch = 0;
  if (isxdigit(p[1]))
    {
    ch = ch * 16 +
      Ustrchr(hex_digits, tolower(*(++p))) - hex_digits;
    if (isxdigit(p[1])) ch = ch * 16 +
      Ustrchr(hex_digits, tolower(*(++p))) - hex_digits;
    }
  break;
  }
*pp = p;
return ch;
}



#ifndef COMPILE_UTILITY
/*************************************************
*          Ensure string is printable            *
*************************************************/

/* This function is called for critical strings. It checks for any
non-printing characters, and if any are found, it makes a new copy
of the string with suitable escape sequences. It is most often called by the
macro string_printing(), which sets allow_tab TRUE.

Arguments:
  s             the input string
  allow_tab     TRUE to allow tab as a printing character

Returns:        string with non-printers encoded as printing sequences
*/

const uschar *
string_printing2(const uschar *s, BOOL allow_tab)
{
int nonprintcount = 0;
int length = 0;
const uschar *t = s;
uschar *ss, *tt;

while (*t != 0)
  {
  int c = *t++;
  if (!mac_isprint(c) || (!allow_tab && c == '\t')) nonprintcount++;
  length++;
  }

if (nonprintcount == 0) return s;

/* Get a new block of store guaranteed big enough to hold the
expanded string. */

ss = store_get(length + nonprintcount * 3 + 1);

/* Copy everything, escaping non printers. */

t = s;
tt = ss;

while (*t != 0)
  {
  int c = *t;
  if (mac_isprint(c) && (allow_tab || c != '\t')) *tt++ = *t++; else
    {
    *tt++ = '\\';
    switch (*t)
      {
      case '\n': *tt++ = 'n'; break;
      case '\r': *tt++ = 'r'; break;
      case '\b': *tt++ = 'b'; break;
      case '\v': *tt++ = 'v'; break;
      case '\f': *tt++ = 'f'; break;
      case '\t': *tt++ = 't'; break;
      default: sprintf(CS tt, "%03o", *t); tt += 3; break;
      }
    t++;
    }
  }
*tt = 0;
return ss;
}
#endif  /* COMPILE_UTILITY */

/*************************************************
*        Undo printing escapes in string         *
*************************************************/

/* This function is the reverse of string_printing2.  It searches for
backslash characters and if any are found, it makes a new copy of the
string with escape sequences parsed.  Otherwise it returns the original
string.

Arguments:
  s             the input string

Returns:        string with printing escapes parsed back
*/

uschar *
string_unprinting(uschar *s)
{
uschar *p, *q, *r, *ss;
int len, off;

p = Ustrchr(s, '\\');
if (!p) return s;

len = Ustrlen(s) + 1;
ss = store_get(len);

q = ss;
off = p - s;
if (off)
  {
  memcpy(q, s, off);
  q += off;
  }

while (*p)
  {
  if (*p == '\\')
    {
    *q++ = string_interpret_escape((const uschar **)&p);
    p++;
    }
  else
    {
    r = Ustrchr(p, '\\');
    if (!r)
      {
      off = Ustrlen(p);
      memcpy(q, p, off);
      p += off;
      q += off;
      break;
      }
    else
      {
      off = r - p;
      memcpy(q, p, off);
      q += off;
      p = r;
      }
    }
  }
*q = '\0';

return ss;
}




/*************************************************
*            Copy and save string                *
*************************************************/

/* This function assumes that memcpy() is faster than strcpy().

Argument: string to copy
Returns:  copy of string in new store
*/

uschar *
string_copy(const uschar *s)
{
int len = Ustrlen(s) + 1;
uschar *ss = store_get(len);
memcpy(ss, s, len);
return ss;
}



/*************************************************
*     Copy and save string in malloc'd store     *
*************************************************/

/* This function assumes that memcpy() is faster than strcpy().

Argument: string to copy
Returns:  copy of string in new store
*/

uschar *
string_copy_malloc(const uschar *s)
{
int len = Ustrlen(s) + 1;
uschar *ss = store_malloc(len);
memcpy(ss, s, len);
return ss;
}



/*************************************************
*       Copy, lowercase and save string          *
*************************************************/

/*
Argument: string to copy
Returns:  copy of string in new store, with letters lowercased
*/

uschar *
string_copylc(const uschar *s)
{
uschar *ss = store_get(Ustrlen(s) + 1);
uschar *p = ss;
while (*s != 0) *p++ = tolower(*s++);
*p = 0;
return ss;
}



/*************************************************
*       Copy and save string, given length       *
*************************************************/

/* It is assumed the data contains no zeros. A zero is added
onto the end.

Arguments:
  s         string to copy
  n         number of characters

Returns:    copy of string in new store
*/

uschar *
string_copyn(const uschar *s, int n)
{
uschar *ss = store_get(n + 1);
Ustrncpy(ss, s, n);
ss[n] = 0;
return ss;
}


/*************************************************
* Copy, lowercase, and save string, given length *
*************************************************/

/* It is assumed the data contains no zeros. A zero is added
onto the end.

Arguments:
  s         string to copy
  n         number of characters

Returns:    copy of string in new store, with letters lowercased
*/

uschar *
string_copynlc(uschar *s, int n)
{
uschar *ss = store_get(n + 1);
uschar *p = ss;
while (n-- > 0) *p++ = tolower(*s++);
*p = 0;
return ss;
}



/*************************************************
*    Copy string if long, inserting newlines     *
*************************************************/

/* If the given string is longer than 75 characters, it is copied, and within
the copy, certain space characters are converted into newlines.

Argument:  pointer to the string
Returns:   pointer to the possibly altered string
*/

uschar *
string_split_message(uschar *msg)
{
uschar *s, *ss;

if (msg == NULL || Ustrlen(msg) <= 75) return msg;
s = ss = msg = string_copy(msg);

for (;;)
  {
  int i = 0;
  while (i < 75 && *ss != 0 && *ss != '\n') ss++, i++;
  if (*ss == 0) break;
  if (*ss == '\n')
    s = ++ss;
  else
    {
    uschar *t = ss + 1;
    uschar *tt = NULL;
    while (--t > s + 35)
      {
      if (*t == ' ')
        {
        if (t[-1] == ':') { tt = t; break; }
        if (tt == NULL) tt = t;
        }
      }

    if (tt == NULL)          /* Can't split behind - try ahead */
      {
      t = ss + 1;
      while (*t != 0)
        {
        if (*t == ' ' || *t == '\n')
          { tt = t; break; }
        t++;
        }
      }

    if (tt == NULL) break;   /* Can't find anywhere to split */
    *tt = '\n';
    s = ss = tt+1;
    }
  }

return msg;
}



/*************************************************
*   Copy returned DNS domain name, de-escaping   *
*************************************************/

/* If a domain name contains top-bit characters, some resolvers return
the fully qualified name with those characters turned into escapes. The
convention is a backslash followed by _decimal_ digits. We convert these
back into the original binary values. This will be relevant when
allow_utf8_domains is set true and UTF-8 characters are used in domain
names. Backslash can also be used to escape other characters, though we
shouldn't come across them in domain names.

Argument:   the domain name string
Returns:    copy of string in new store, de-escaped
*/

uschar *
string_copy_dnsdomain(uschar *s)
{
uschar *yield;
uschar *ss = yield = store_get(Ustrlen(s) + 1);

while (*s != 0)
  {
  if (*s != '\\')
    {
    *ss++ = *s++;
    }
  else if (isdigit(s[1]))
    {
    *ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0';
    s += 4;
    }
  else if (*(++s) != 0)
    {
    *ss++ = *s++;
    }
  }

*ss = 0;
return yield;
}


#ifndef COMPILE_UTILITY
/*************************************************
*     Copy space-terminated or quoted string     *
*************************************************/

/* This function copies from a string until its end, or until whitespace is
encountered, unless the string begins with a double quote, in which case the
terminating quote is sought, and escaping within the string is done. The length
of a de-quoted string can be no longer than the original, since escaping always
turns n characters into 1 character.

Argument:  pointer to the pointer to the first character, which gets updated
Returns:   the new string
*/

uschar *
string_dequote(const uschar **sptr)
{
const uschar *s = *sptr;
uschar *t, *yield;

/* First find the end of the string */

if (*s != '\"')
  while (*s != 0 && !isspace(*s)) s++;
else
  {
  s++;
  while (*s && *s != '\"')
    {
    if (*s == '\\') (void)string_interpret_escape(&s);
    s++;
    }
  if (*s) s++;
  }

/* Get enough store to copy into */

t = yield = store_get(s - *sptr + 1);
s = *sptr;

/* Do the copy */

if (*s != '\"')
  {
  while (*s != 0 && !isspace(*s)) *t++ = *s++;
  }
else
  {
  s++;
  while (*s != 0 && *s != '\"')
    {
    if (*s == '\\') *t++ = string_interpret_escape(&s);
      else *t++ = *s;
    s++;
    }
  if (*s != 0) s++;
  }

/* Update the pointer and return the terminated copy */

*sptr = s;
*t = 0;
return yield;
}
#endif  /* COMPILE_UTILITY */



/*************************************************
*          Format a string and save it           *
*************************************************/

/* The formatting is done by string_vformat, which checks the length of
everything.

Arguments:
  format    a printf() format - deliberately char * rather than uschar *
              because it will most usually be a literal string
  ...       arguments for format

Returns:    pointer to fresh piece of store containing sprintf'ed string
*/

uschar *
string_sprintf(const char *format, ...)
{
#ifdef COMPILE_UTILITY
uschar buffer[STRING_SPRINTF_BUFFER_SIZE];
gstring g = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer };
gstring * gp = &g;
#else
gstring * gp = string_get(STRING_SPRINTF_BUFFER_SIZE);
#endif
gstring * gp2;
va_list ap;

va_start(ap, format);
gp2 = string_vformat(gp, FALSE, format, ap);
gp->s[gp->ptr] = '\0';
va_end(ap);

if (!gp2)
  log_write(0, LOG_MAIN|LOG_PANIC_DIE,
    "string_sprintf expansion was longer than %d; format string was (%s)\n"
    "expansion started '%.32s'",
    gp->size, format, gp->s);

#ifdef COMPILE_UTILITY
return string_copy(gp->s);
#else
gstring_reset_unused(gp);
return gp->s;
#endif
}



/*************************************************
*         Case-independent strncmp() function    *
*************************************************/

/*
Arguments:
  s         first string
  t         second string
  n         number of characters to compare

Returns:    < 0, = 0, or > 0, according to the comparison
*/

int
strncmpic(const uschar *s, const uschar *t, int n)
{
while (n--)
  {
  int c = tolower(*s++) - tolower(*t++);
  if (c) return c;
  }
return 0;
}


/*************************************************
*         Case-independent strcmp() function     *
*************************************************/

/*
Arguments:
  s         first string
  t         second string

Returns:    < 0, = 0, or > 0, according to the comparison
*/

int
strcmpic(const uschar *s, const uschar *t)
{
while (*s != 0)
  {
  int c = tolower(*s++) - tolower(*t++);
  if (c != 0) return c;
  }
return *t;
}


/*************************************************
*         Case-independent strstr() function     *
*************************************************/

/* The third argument specifies whether whitespace is required
to follow the matched string.

Arguments:
  s              string to search
  t              substring to search for
  space_follows  if TRUE, match only if whitespace follows

Returns:         pointer to substring in string, or NULL if not found
*/

uschar *
strstric(uschar *s, uschar *t, BOOL space_follows)
{
uschar *p = t;
uschar *yield = NULL;
int cl = tolower(*p);
int cu = toupper(*p);

while (*s)
  {
  if (*s == cl || *s == cu)
    {
    if (yield == NULL) yield = s;
    if (*(++p) == 0)
      {
      if (!space_follows || s[1] == ' ' || s[1] == '\n' ) return yield;
      yield = NULL;
      p = t;
      }
    cl = tolower(*p);
    cu = toupper(*p);
    s++;
    }
  else if (yield != NULL)
    {
    yield = NULL;
    p = t;
    cl = tolower(*p);
    cu = toupper(*p);
    }
  else s++;
  }
return NULL;
}



#ifdef COMPILE_UTILITY
/* Dummy version for this function; it should never be called */
static void
gstring_grow(gstring * g, int p, int count)
{
assert(FALSE);
}
#endif



#ifndef COMPILE_UTILITY
/*************************************************
*       Get next string from separated list      *
*************************************************/

/* Leading and trailing space is removed from each item. The separator in the
list is controlled by the int pointed to by the separator argument as follows:

  If the value is > 0 it is used as the separator. This is typically used for
  sublists such as slash-separated options. The value is always a printing
  character.

    (If the value is actually > UCHAR_MAX there is only one item in the list.
    This is used for some cases when called via functions that sometimes
    plough through lists, and sometimes are given single items.)

  If the value is <= 0, the string is inspected for a leading <x, where x is an
  ispunct() or an iscntrl() character. If found, x is used as the separator. If
  not found:

      (a) if separator == 0, ':' is used
      (b) if separator <0, -separator is used

  In all cases the value of the separator that is used is written back to the
  int so that it is used on subsequent calls as we progress through the list.

A literal ispunct() separator can be represented in an item by doubling, but
there is no way to include an iscntrl() separator as part of the data.

Arguments:
  listptr    points to a pointer to the current start of the list; the
             pointer gets updated to point after the end of the next item
  separator  a pointer to the separator character in an int (see above)
  buffer     where to put a copy of the next string in the list; or
               NULL if the next string is returned in new memory
  buflen     when buffer is not NULL, the size of buffer; otherwise ignored

Returns:     pointer to buffer, containing the next substring,
             or NULL if no more substrings
*/

uschar *
string_nextinlist(const uschar **listptr, int *separator, uschar *buffer, int buflen)
{
int sep = *separator;
const uschar *s = *listptr;
BOOL sep_is_special;

if (!s) return NULL;

/* This allows for a fixed specified separator to be an iscntrl() character,
but at the time of implementation, this is never the case. However, it's best
to be conservative. */

while (isspace(*s) && *s != sep) s++;

/* A change of separator is permitted, so look for a leading '<' followed by an
allowed character. */

if (sep <= 0)
  {
  if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1])))
    {
    sep = s[1];
    if (*++s) ++s;
    while (isspace(*s) && *s != sep) s++;
    }
  else
    sep = sep ? -sep : ':';
  *separator = sep;
  }

/* An empty string has no list elements */

if (!*s) return NULL;

/* Note whether whether or not the separator is an iscntrl() character. */

sep_is_special = iscntrl(sep);

/* Handle the case when a buffer is provided. */

if (buffer)
  {
  int p = 0;
  for (; *s; s++)
    {
    if (*s == sep && (*(++s) != sep || sep_is_special)) break;
    if (p < buflen - 1) buffer[p++] = *s;
    }
  while (p > 0 && isspace(buffer[p-1])) p--;
  buffer[p] = '\0';
  }

/* Handle the case when a buffer is not provided. */

else
  {
  const uschar *ss;
  gstring * g = NULL;

  /* We know that *s != 0 at this point. However, it might be pointing to a
  separator, which could indicate an empty string, or (if an ispunct()
  character) could be doubled to indicate a separator character as data at the
  start of a string. Avoid getting working memory for an empty item. */

  if (*s == sep)
    {
    s++;
    if (*s != sep || sep_is_special)
      {
      *listptr = s;
      return string_copy(US"");
      }
    }

  /* Not an empty string; the first character is guaranteed to be a data
  character. */

  for (;;)
    {
    for (ss = s + 1; *ss && *ss != sep; ss++) ;
    g = string_catn(g, s, ss-s);
    s = ss;
    if (!*s || *++s != sep || sep_is_special) break;
    }
  while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--;
  buffer = string_from_gstring(g);
  gstring_reset_unused(g);
  }

/* Update the current pointer and return the new string */

*listptr = s;
return buffer;
}


static const uschar *
Ustrnchr(const uschar * s, int c, unsigned * len)
{
unsigned siz = *len;
while (siz)
  {
  if (!*s) return NULL;
  if (*s == c)
    {
    *len = siz;
    return s;
    }
  s++;
  siz--;
  }
return NULL;
}


/************************************************
*       Add element to separated list           *
************************************************/
/* This function is used to build a list, returning an allocated null-terminated
growable string. The given element has any embedded separator characters
doubled.

Despite having the same growable-string interface as string_cat() the list is
always returned null-terminated.

Arguments:
  list  expanding-string for the list that is being built, or NULL
        if this is a new list that has no contents yet
  sep   list separator character
  ele   new element to be appended to the list

Returns:  pointer to the start of the list, changed if copied for expansion.
*/

gstring *
string_append_listele(gstring * list, uschar sep, const uschar * ele)
{
uschar * sp;

if (list && list->ptr)
  list = string_catn(list, &sep, 1);

while((sp = Ustrchr(ele, sep)))
  {
  list = string_catn(list, ele, sp-ele+1);
  list = string_catn(list, &sep, 1);
  ele = sp+1;
  }
list = string_cat(list, ele);
(void) string_from_gstring(list);
return list;
}


gstring *
string_append_listele_n(gstring * list, uschar sep, const uschar * ele,
 unsigned len)
{
const uschar * sp;

if (list && list->ptr)
  list = string_catn(list, &sep, 1);

while((sp = Ustrnchr(ele, sep, &len)))
  {
  list = string_catn(list, ele, sp-ele+1);
  list = string_catn(list, &sep, 1);
  ele = sp+1;
  len--;
  }
list = string_catn(list, ele, len);
(void) string_from_gstring(list);
return list;
}



/* A slightly-bogus listmaker utility; the separator is a string so
can be multiple chars - there is no checking for the element content
containing any of the separator. */

gstring *
string_append2_listele_n(gstring * list, const uschar * sepstr,
 const uschar * ele, unsigned len)
{
if (list && list->ptr)
  list = string_cat(list, sepstr);

list = string_catn(list, ele, len);
(void) string_from_gstring(list);
return list;
}



/************************************************/
/* Create a growable-string with some preassigned space */

gstring *
string_get(unsigned size)
{
gstring * g = store_get(sizeof(gstring) + size);
g->size = size;
g->ptr = 0;
g->s = US(g + 1);
return g;
}

/* NUL-terminate the C string in the growable-string, and return it. */

uschar *
string_from_gstring(gstring * g)
{
if (!g) return NULL;
g->s[g->ptr] = '\0';
return g->s;
}

void
gstring_reset_unused(gstring * g)
{
store_reset(g->s + (g->size = g->ptr + 1));
}


/* Add more space to a growable-string.

Arguments:
  g             the growable-string
  p             current end of data
  count         amount to grow by, offset from p
*/

static void
gstring_grow(gstring * g, int p, int count)
{
int oldsize = g->size;

/* Mostly, string_cat() is used to build small strings of a few hundred
characters at most. There are times, however, when the strings are very much
longer (for example, a lookup that returns a vast number of alias addresses).
To try to keep things reasonable, we use increments whose size depends on the
existing length of the string. */

unsigned inc = oldsize < 4096 ? 127 : 1023;
g->size = ((p + count + inc) & ~inc) + 1;

/* Try to extend an existing allocation. If the result of calling
store_extend() is false, either there isn't room in the current memory block,
or this string is not the top item on the dynamic store stack. We then have
to get a new chunk of store and copy the old string. When building large
strings, it is helpful to call store_release() on the old string, to release
memory blocks that have become empty. (The block will be freed if the string
is at its start.) However, we can do this only if we know that the old string
was the last item on the dynamic memory stack. This is the case if it matches
store_last_get. */

if (!store_extend(g->s, oldsize, g->size))
  g->s = store_newblock(g->s, g->size, p);
}



/*************************************************
*             Add chars to string                *
*************************************************/
/* This function is used when building up strings of unknown length. Room is
always left for a terminating zero to be added to the string that is being
built. This function does not require the string that is being added to be NUL
terminated, because the number of characters to add is given explicitly. It is
sometimes called to extract parts of other strings.

Arguments:
  string   points to the start of the string that is being built, or NULL
             if this is a new string that has no contents yet
  s        points to characters to add
  count    count of characters to add; must not exceed the length of s, if s
             is a C string.

Returns:   pointer to the start of the string, changed if copied for expansion.
           Note that a NUL is not added, though space is left for one. This is
           because string_cat() is often called multiple times to build up a
           string - there's no point adding the NUL till the end.

*/
/* coverity[+alloc] */

gstring *
string_catn(gstring * g, const uschar *s, int count)
{
int p;

if (!g)
  {
  unsigned inc = count < 4096 ? 127 : 1023;
  unsigned size = ((count + inc) &  ~inc) + 1;
  g = string_get(size);
  }

p = g->ptr;
if (p + count >= g->size)
  gstring_grow(g, p, count);

/* Because we always specify the exact number of characters to copy, we can
use memcpy(), which is likely to be more efficient than strncopy() because the
latter has to check for zero bytes. */

memcpy(g->s + p, s, count);
g->ptr = p + count;
return g;
}


gstring *
string_cat(gstring *string, const uschar *s)
{
return string_catn(string, s, Ustrlen(s));
}



/*************************************************
*        Append strings to another string        *
*************************************************/

/* This function can be used to build a string from many other strings.
It calls string_cat() to do the dirty work.

Arguments:
  string   expanding-string that is being built, or NULL
             if this is a new string that has no contents yet
  count    the number of strings to append
  ...      "count" uschar* arguments, which must be valid zero-terminated
             C strings

Returns:   pointer to the start of the string, changed if copied for expansion.
           The string is not zero-terminated - see string_cat() above.
*/

__inline__ gstring *
string_append(gstring *string, int count, ...)
{
va_list ap;

va_start(ap, count);
while (count-- > 0)
  {
  uschar *t = va_arg(ap, uschar *);
  string = string_cat(string, t);
  }
va_end(ap);

return string;
}
#endif



/*************************************************
*        Format a string with length checks      *
*************************************************/

/* This function is used to format a string with checking of the length of the
output for all conversions. It protects Exim from absent-mindedness when
calling functions like debug_printf and string_sprintf, and elsewhere. There
are two different entry points to what is actually the same function, depending
on whether the variable length list of data arguments are given explicitly or
as a va_list item.

The formats are the usual printf() ones, with some omissions (never used) and
three additions for strings: %S forces lower case, %T forces upper case, and
%#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed
(useful in debugging). There is also the addition of %D and %M, which insert
the date in the form used for datestamped log files.

Arguments:
  buffer       a buffer in which to put the formatted string
  buflen       the length of the buffer
  format       the format string - deliberately char * and not uschar *
  ... or ap    variable list of supplementary arguments

Returns:       TRUE if the result fitted in the buffer
*/

BOOL
string_format(uschar * buffer, int buflen, const char * format, ...)
{
gstring g = { .size = buflen, .ptr = 0, .s = buffer }, *gp;
va_list ap;
va_start(ap, format);
gp = string_vformat(&g, FALSE, format, ap);
va_end(ap);
g.s[g.ptr] = '\0';
return !!gp;
}





/* Bulid or append to a growing-string, sprintf-style.

If the "extend" argument is true, the string passed in can be NULL,
empty, or non-empty.

If the "extend" argument is false, the string passed in may not be NULL,
will not be grown, and is usable in the original place after return.
The return value can be NULL to signify overflow.

Returns the possibly-new (if copy for growth was needed) string,
not nul-terminated.
*/

gstring *
string_vformat(gstring * g, BOOL extend, const char *format, va_list ap)
{
enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 };

int width, precision, off, lim;
const char * fp = format;       /* Deliberately not unsigned */

string_datestamp_offset = -1;   /* Datestamp not inserted */
string_datestamp_length = 0;    /* Datestamp not inserted */
string_datestamp_type = 0;      /* Datestamp not inserted */

#ifdef COMPILE_UTILITY
assert(!extend);
assert(g);
#else

/* Ensure we have a string, to save on checking later */
if (!g) g = string_get(16);
#endif  /*!COMPILE_UTILITY*/

lim = g->size - 1;      /* leave one for a nul */
off = g->ptr;           /* remember initial offset in gstring */

/* Scan the format and handle the insertions */

while (*fp)
  {
  int length = L_NORMAL;
  int *nptr;
  int slen;
  const char *null = "NULL";            /* ) These variables */
  const char *item_start, *s;           /* ) are deliberately */
  char newformat[16];                   /* ) not unsigned */
  char * gp = CS g->s + g->ptr;         /* ) */

  /* Non-% characters just get copied verbatim */

  if (*fp != '%')
    {
    /* Avoid string_copyn() due to COMPILE_UTILITY */
    if (g->ptr >= lim - 1)
      {
      if (!extend) return NULL;
      gstring_grow(g, g->ptr, 1);
      lim = g->size - 1;
      }
    g->s[g->ptr++] = (uschar) *fp++;
    continue;
    }

  /* Deal with % characters. Pick off the width and precision, for checking
  strings, skipping over the flag and modifier characters. */

  item_start = fp;
  width = precision = -1;

  if (strchr("-+ #0", *(++fp)) != NULL)
    {
    if (*fp == '#') null = "";
    fp++;
    }

  if (isdigit((uschar)*fp))
    {
    width = *fp++ - '0';
    while (isdigit((uschar)*fp)) width = width * 10 + *fp++ - '0';
    }
  else if (*fp == '*')
    {
    width = va_arg(ap, int);
    fp++;
    }

  if (*fp == '.')
    if (*(++fp) == '*')
      {
      precision = va_arg(ap, int);
      fp++;
      }
    else
      for (precision = 0; isdigit((uschar)*fp); fp++)
        precision = precision*10 + *fp - '0';

  /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */

  if (*fp == 'h')
    { fp++; length = L_SHORT; }
  else if (*fp == 'L')
    { fp++; length = L_LONGDOUBLE; }
  else if (*fp == 'l')
    if (fp[1] == 'l')
      { fp += 2; length = L_LONGLONG; }
    else
      { fp++; length = L_LONG; }
  else if (*fp == 'z')
    { fp++; length = L_SIZE; }

  /* Handle each specific format type. */

  switch (*fp++)
    {
    case 'n':
      nptr = va_arg(ap, int *);
      *nptr = g->ptr - off;
      break;

    case 'd':
    case 'o':
    case 'u':
    case 'x':
    case 'X':
      width = length > L_LONG ? 24 : 12;
      if (g->ptr >= lim - width)
        {
        if (!extend) return NULL;
        gstring_grow(g, g->ptr, width);
        lim = g->size - 1;
        gp = CS g->s + g->ptr;
        }
      strncpy(newformat, item_start, fp - item_start);
      newformat[fp - item_start] = 0;

      /* Short int is promoted to int when passing through ..., so we must use
      int for va_arg(). */

      switch(length)
        {
        case L_SHORT:
        case L_NORMAL:
          g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break;
        case L_LONG:
          g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break;
        case L_LONGLONG:
          g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break;
        case L_SIZE:
          g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break;
        }
      break;

    case 'p':
      {
      void * ptr;
      if (g->ptr >= lim - 24)
        {
        if (!extend) return NULL;
        gstring_grow(g, g->ptr, 24);
        lim = g->size - 1;
        gp = CS g->s + g->ptr;
        }
      /* sprintf() saying "(nil)" for a null pointer seems unreliable.
      Handle it explicitly. */
      if ((ptr = va_arg(ap, void *)))
        {
        strncpy(newformat, item_start, fp - item_start);
        newformat[fp - item_start] = 0;
        g->ptr += sprintf(gp, newformat, ptr);
        }
      else
        g->ptr += sprintf(gp, "(nil)");
      }
    break;

    /* %f format is inherently insecure if the numbers that it may be
    handed are unknown (e.g. 1e300). However, in Exim, %f is used for
    printing load averages, and these are actually stored as integers
    (load average * 1000) so the size of the numbers is constrained.
    It is also used for formatting sending rates, where the simplicity
    of the format prevents overflow. */

    case 'f':
    case 'e':
    case 'E':
    case 'g':
    case 'G':
      if (precision < 0) precision = 6;
      if (g->ptr >= lim - precision - 8)
        {
        if (!extend) return NULL;
        gstring_grow(g, g->ptr, precision+8);
        lim = g->size - 1;
        gp = CS g->s + g->ptr;
        }
      strncpy(newformat, item_start, fp - item_start);
      newformat[fp-item_start] = 0;
      if (length == L_LONGDOUBLE)
        g->ptr += sprintf(gp, newformat, va_arg(ap, long double));
      else
        g->ptr += sprintf(gp, newformat, va_arg(ap, double));
      break;

    /* String types */

    case '%':
      if (g->ptr >= lim - 1)
        {
        if (!extend) return NULL;
        gstring_grow(g, g->ptr, 1);
        lim = g->size - 1;
        }
      g->s[g->ptr++] = (uschar) '%';
      break;

    case 'c':
      if (g->ptr >= lim - 1)
        {
        if (!extend) return NULL;
        gstring_grow(g, g->ptr, 1);
        lim = g->size - 1;
        }
      g->s[g->ptr++] = (uschar) va_arg(ap, int);
      break;

    case 'D':                   /* Insert daily datestamp for log file names */
      s = CS tod_stamp(tod_log_datestamp_daily);
      string_datestamp_offset = g->ptr;         /* Passed back via global */
      string_datestamp_length = Ustrlen(s);     /* Passed back via global */
      string_datestamp_type = tod_log_datestamp_daily;
      slen = string_datestamp_length;
      goto INSERT_STRING;

    case 'M':                   /* Insert monthly datestamp for log file names */
      s = CS tod_stamp(tod_log_datestamp_monthly);
      string_datestamp_offset = g->ptr;         /* Passed back via global */
      string_datestamp_length = Ustrlen(s);     /* Passed back via global */
      string_datestamp_type = tod_log_datestamp_monthly;
      slen = string_datestamp_length;
      goto INSERT_STRING;

    case 's':
    case 'S':                   /* Forces *lower* case */
    case 'T':                   /* Forces *upper* case */
      s = va_arg(ap, char *);

      if (!s) s = null;
      slen = Ustrlen(s);

    INSERT_STRING:              /* Come to from %D or %M above */

      {
      BOOL truncated = FALSE;

      /* If the width is specified, check that there is a precision
      set; if not, set it to the width to prevent overruns of long
      strings. */

      if (width >= 0)
        {
        if (precision < 0) precision = width;
        }

      /* If a width is not specified and the precision is specified, set
      the width to the precision, or the string length if shorted. */

      else if (precision >= 0)
        width = precision < slen ? precision : slen;

      /* If neither are specified, set them both to the string length. */

      else
        width = precision = slen;

      if (!extend)
        {
        if (g->ptr == lim) return NULL;
        if (g->ptr >= lim - width)
          {
          truncated = TRUE;
          width = precision = lim - g->ptr - 1;
          if (width < 0) width = 0;
          if (precision < 0) precision = 0;
          }
        }
      else if (g->ptr >= lim - width)
        {
        gstring_grow(g, g->ptr, width);
        lim = g->size - 1;
        gp = CS g->s + g->ptr;
        }

      g->ptr += sprintf(gp, "%*.*s", width, precision, s);
      if (fp[-1] == 'S')
        while (*gp) { *gp = tolower(*gp); gp++; }
      else if (fp[-1] == 'T')
        while (*gp) { *gp = toupper(*gp); gp++; }

      if (truncated) return NULL;
      break;
      }

    /* Some things are never used in Exim; also catches junk. */

    default:
      strncpy(newformat, item_start, fp - item_start);
      newformat[fp-item_start] = 0;
      log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type "
        "in \"%s\" in \"%s\"", newformat, format);
      break;
    }
  }

return g;
}



#ifndef COMPILE_UTILITY

gstring *
string_fmt_append(gstring * g, const char *format, ...)
{
va_list ap;
va_start(ap, format);
g = string_vformat(g, TRUE, format, ap);
va_end(ap);
return g;
}



/*************************************************
*       Generate an "open failed" message        *
*************************************************/

/* This function creates a message after failure to open a file. It includes a
string supplied as data, adds the strerror() text, and if the failure was
"Permission denied", reads and includes the euid and egid.

Arguments:
  eno           the value of errno after the failure
  format        a text format string - deliberately not uschar *
  ...           arguments for the format string

Returns:        a message, in dynamic store
*/

uschar *
string_open_failed(int eno, const char *format, ...)
{
va_list ap;
gstring * g = string_get(1024);

g = string_catn(g, US"failed to open ", 15);

/* Use the checked formatting routine to ensure that the buffer
does not overflow. It should not, since this is called only for internally
specified messages. If it does, the message just gets truncated, and there
doesn't seem much we can do about that. */

va_start(ap, format);
(void) string_vformat(g, FALSE, format, ap);
string_from_gstring(g);
gstring_reset_unused(g);
va_end(ap);

return eno == EACCES
  ? string_sprintf("%s: %s (euid=%ld egid=%ld)", g->s, strerror(eno),
    (long int)geteuid(), (long int)getegid())
  : string_sprintf("%s: %s", g->s, strerror(eno));
}
#endif  /* COMPILE_UTILITY */





#ifndef COMPILE_UTILITY
/* qsort(3), currently used to sort the environment variables
for -bP environment output, needs a function to compare two pointers to string
pointers. Here it is. */

int
string_compare_by_pointer(const void *a, const void *b)
{
return Ustrcmp(* CUSS a, * CUSS b);
}
#endif /* COMPILE_UTILITY */




/*************************************************
**************************************************
*             Stand-alone test program           *
**************************************************
*************************************************/

#ifdef STAND_ALONE
int main(void)
{
uschar buffer[256];

printf("Testing is_ip_address\n");

while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
  {
  int offset;
  buffer[Ustrlen(buffer) - 1] = 0;
  printf("%d\n", string_is_ip_address(buffer, NULL));
  printf("%d %d %s\n", string_is_ip_address(buffer, &offset), offset, buffer);
  }

printf("Testing string_nextinlist\n");

while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
  {
  uschar *list = buffer;
  uschar *lp1, *lp2;
  uschar item[256];
  int sep1 = 0;
  int sep2 = 0;

  if (*list == '<')
    {
    sep1 = sep2 = list[1];
    list += 2;
    }

  lp1 = lp2 = list;
  for (;;)
    {
    uschar *item1 = string_nextinlist(&lp1, &sep1, item, sizeof(item));
    uschar *item2 = string_nextinlist(&lp2, &sep2, NULL, 0);

    if (item1 == NULL && item2 == NULL) break;
    if (item == NULL || item2 == NULL || Ustrcmp(item1, item2) != 0)
      {
      printf("***ERROR\nitem1=\"%s\"\nitem2=\"%s\"\n",
        (item1 == NULL)? "NULL" : CS item1,
        (item2 == NULL)? "NULL" : CS item2);
      break;
      }
    else printf("  \"%s\"\n", CS item1);
    }
  }

/* This is a horrible lash-up, but it serves its purpose. */

printf("Testing string_format\n");

while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
  {
  void *args[3];
  long long llargs[3];
  double dargs[3];
  int dflag = 0;
  int llflag = 0;
  int n = 0;
  int count;
  int countset = 0;
  uschar format[256];
  uschar outbuf[256];
  uschar *s;
  buffer[Ustrlen(buffer) - 1] = 0;

  s = Ustrchr(buffer, ',');
  if (s == NULL) s = buffer + Ustrlen(buffer);

  Ustrncpy(format, buffer, s - buffer);
  format[s-buffer] = 0;

  if (*s == ',') s++;

  while (*s != 0)
    {
    uschar *ss = s;
    s = Ustrchr(ss, ',');
    if (s == NULL) s = ss + Ustrlen(ss);

    if (isdigit(*ss))
      {
      Ustrncpy(outbuf, ss, s-ss);
      if (Ustrchr(outbuf, '.') != NULL)
        {
        dflag = 1;
        dargs[n++] = Ustrtod(outbuf, NULL);
        }
      else if (Ustrstr(outbuf, "ll") != NULL)
        {
        llflag = 1;
        llargs[n++] = strtoull(CS outbuf, NULL, 10);
        }
      else
        {
        args[n++] = (void *)Uatoi(outbuf);
        }
      }

    else if (Ustrcmp(ss, "*") == 0)
      {
      args[n++] = (void *)(&count);
      countset = 1;
      }

    else
      {
      uschar *sss = malloc(s - ss + 1);
      Ustrncpy(sss, ss, s-ss);
      args[n++] = sss;
      }

    if (*s == ',') s++;
    }

  if (!dflag && !llflag)
    printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
      args[0], args[1], args[2])? "True" : "False");

  else if (dflag)
    printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
      dargs[0], dargs[1], dargs[2])? "True" : "False");

  else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format,
    llargs[0], llargs[1], llargs[2])? "True" : "False");

  printf("%s\n", CS outbuf);
  if (countset) printf("count=%d\n", count);
  }

return 0;
}
#endif

/* End of string.c */