utils.c

wget讓你可以在console介面下

   are currently used, which means that "%s %s %s" will work, but "%s
   %s %s %s" won't.  If you need to print more than three wgints,
   bump the RING_SIZE (or rethink your message.)  */

char *
number_to_static_string (wgint number)
{
  static char ring[RING_SIZE][24];
  static int ringpos;
  char *buf = ring[ringpos];
  number_to_string (buf, number);
  ringpos = (ringpos + 1) % RING_SIZE;
  return buf;
}

/* Determine the width of the terminal we're running on.  If that's
   not possible, return 0.  */

int
determine_screen_width (void)
{
  /* If there's a way to get the terminal size using POSIX
     tcgetattr(), somebody please tell me.  */
#ifdef TIOCGWINSZ
  int fd;
  struct winsize wsz;

  if (opt.lfilename != NULL)
    return 0;

  fd = fileno (stderr);
  if (ioctl (fd, TIOCGWINSZ, &wsz) < 0)
    return 0;			/* most likely ENOTTY */

  return wsz.ws_col;
#else  /* not TIOCGWINSZ */
# ifdef WINDOWS
  CONSOLE_SCREEN_BUFFER_INFO csbi;
  if (!GetConsoleScreenBufferInfo (GetStdHandle (STD_ERROR_HANDLE), &csbi))
    return 0;
  return csbi.dwSize.X;
# else /* neither WINDOWS nor TIOCGWINSZ */
  return 0;
#endif /* neither WINDOWS nor TIOCGWINSZ */
#endif /* not TIOCGWINSZ */
}

/* Return a random number between 0 and MAX-1, inclusive.

   If MAX is greater than the value of RAND_MAX+1 on the system, the
   returned value will be in the range [0, RAND_MAX].  This may be
   fixed in a future release.

   The random number generator is seeded automatically the first time
   it is called.

   This uses rand() for portability.  It has been suggested that
   random() offers better randomness, but this is not required for
   Wget, so I chose to go for simplicity and use rand
   unconditionally.

   DO NOT use this for cryptographic purposes.  It is only meant to be
   used in situations where quality of the random numbers returned
   doesn't really matter.  */

int
random_number (int max)
{
  static int seeded;
  double bounded;
  int rnd;

  if (!seeded)
    {
      srand (time (NULL));
      seeded = 1;
    }
  rnd = rand ();

  /* On systems that don't define RAND_MAX, assume it to be 2**15 - 1,
     and enforce that assumption by masking other bits.  */
#ifndef RAND_MAX
# define RAND_MAX 32767
  rnd &= RAND_MAX;
#endif

  /* This is equivalent to rand() % max, but uses the high-order bits
     for better randomness on architecture where rand() is implemented
     using a simple congruential generator.  */

  bounded = (double)max * rnd / (RAND_MAX + 1.0);
  return (int)bounded;
}

/* Return a random uniformly distributed floating point number in the
   [0, 1) range.  The precision of returned numbers is 9 digits.

   Modify this to use erand48() where available!  */

double
random_float (void)
{
  /* We can't rely on any specific value of RAND_MAX, but I'm pretty
     sure it's greater than 1000.  */
  int rnd1 = random_number (1000);
  int rnd2 = random_number (1000);
  int rnd3 = random_number (1000);
  return rnd1 / 1000.0 + rnd2 / 1000000.0 + rnd3 / 1000000000.0;
}

/* Implementation of run_with_timeout, a generic timeout-forcing
   routine for systems with Unix-like signal handling.  */

#ifdef USE_SIGNAL_TIMEOUT
# ifdef HAVE_SIGSETJMP
#  define SETJMP(env) sigsetjmp (env, 1)

static sigjmp_buf run_with_timeout_env;

static RETSIGTYPE
abort_run_with_timeout (int sig)
{
  assert (sig == SIGALRM);
  siglongjmp (run_with_timeout_env, -1);
}
# else /* not HAVE_SIGSETJMP */
#  define SETJMP(env) setjmp (env)

static jmp_buf run_with_timeout_env;

static RETSIGTYPE
abort_run_with_timeout (int sig)
{
  assert (sig == SIGALRM);
  /* We don't have siglongjmp to preserve the set of blocked signals;
     if we longjumped out of the handler at this point, SIGALRM would
     remain blocked.  We must unblock it manually. */
  int mask = siggetmask ();
  mask &= ~sigmask (SIGALRM);
  sigsetmask (mask);

  /* Now it's safe to longjump. */
  longjmp (run_with_timeout_env, -1);
}
# endif /* not HAVE_SIGSETJMP */

/* Arrange for SIGALRM to be delivered in TIMEOUT seconds.  This uses
   setitimer where available, alarm otherwise.

   TIMEOUT should be non-zero.  If the timeout value is so small that
   it would be rounded to zero, it is rounded to the least legal value
   instead (1us for setitimer, 1s for alarm).  That ensures that
   SIGALRM will be delivered in all cases.  */

static void
alarm_set (double timeout)
{
#ifdef ITIMER_REAL
  /* Use the modern itimer interface. */
  struct itimerval itv;
  xzero (itv);
  itv.it_value.tv_sec = (long) timeout;
  itv.it_value.tv_usec = 1000000 * (timeout - (long)timeout);
  if (itv.it_value.tv_sec == 0 && itv.it_value.tv_usec == 0)
    /* Ensure that we wait for at least the minimum interval.
       Specifying zero would mean "wait forever".  */
    itv.it_value.tv_usec = 1;
  setitimer (ITIMER_REAL, &itv, NULL);
#else  /* not ITIMER_REAL */
  /* Use the old alarm() interface. */
  int secs = (int) timeout;
  if (secs == 0)
    /* Round TIMEOUTs smaller than 1 to 1, not to zero.  This is
       because alarm(0) means "never deliver the alarm", i.e. "wait
       forever", which is not what someone who specifies a 0.5s
       timeout would expect.  */
    secs = 1;
  alarm (secs);
#endif /* not ITIMER_REAL */
}

/* Cancel the alarm set with alarm_set. */

static void
alarm_cancel (void)
{
#ifdef ITIMER_REAL
  struct itimerval disable;
  xzero (disable);
  setitimer (ITIMER_REAL, &disable, NULL);
#else  /* not ITIMER_REAL */
  alarm (0);
#endif /* not ITIMER_REAL */
}

/* Call FUN(ARG), but don't allow it to run for more than TIMEOUT
   seconds.  Returns non-zero if the function was interrupted with a
   timeout, zero otherwise.

   This works by setting up SIGALRM to be delivered in TIMEOUT seconds
   using setitimer() or alarm().  The timeout is enforced by
   longjumping out of the SIGALRM handler.  This has several
   advantages compared to the traditional approach of relying on
   signals causing system calls to exit with EINTR:

     * The callback function is *forcibly* interrupted after the
       timeout expires, (almost) regardless of what it was doing and
       whether it was in a syscall.  For example, a calculation that
       takes a long time is interrupted as reliably as an IO
       operation.

     * It works with both SYSV and BSD signals because it doesn't
       depend on the default setting of SA_RESTART.

     * It doesn't require special handler setup beyond a simple call
       to signal().  (It does use sigsetjmp/siglongjmp, but they're
       optional.)

   The only downside is that, if FUN allocates internal resources that
   are normally freed prior to exit from the functions, they will be
   lost in case of timeout.  */

int
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
  int saved_errno;

  if (timeout == 0)
    {
      fun (arg);
      return 0;
    }

  signal (SIGALRM, abort_run_with_timeout);
  if (SETJMP (run_with_timeout_env) != 0)
    {
      /* Longjumped out of FUN with a timeout. */
      signal (SIGALRM, SIG_DFL);
      return 1;
    }
  alarm_set (timeout);
  fun (arg);

  /* Preserve errno in case alarm() or signal() modifies it. */
  saved_errno = errno;
  alarm_cancel ();
  signal (SIGALRM, SIG_DFL);
  errno = saved_errno;

  return 0;
}

#else  /* not USE_SIGNAL_TIMEOUT */

#ifndef WINDOWS
/* A stub version of run_with_timeout that just calls FUN(ARG).  Don't
   define it under Windows, because Windows has its own version of
   run_with_timeout that uses threads.  */

int
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
  fun (arg);
  return 0;
}
#endif /* not WINDOWS */
#endif /* not USE_SIGNAL_TIMEOUT */

#ifndef WINDOWS

/* Sleep the specified amount of seconds.  On machines without
   nanosleep(), this may sleep shorter if interrupted by signals.  */

void
xsleep (double seconds)
{
#ifdef HAVE_NANOSLEEP
  /* nanosleep is the preferred interface because it offers high
     accuracy and, more importantly, because it allows us to reliably
     restart receiving a signal such as SIGWINCH.  (There was an
     actual Debian bug report about --limit-rate malfunctioning while
     the terminal was being resized.)  */
  struct timespec sleep, remaining;
  sleep.tv_sec = (long) seconds;
  sleep.tv_nsec = 1000000000 * (seconds - (long) seconds);
  while (nanosleep (&sleep, &remaining) < 0 && errno == EINTR)
    /* If nanosleep has been interrupted by a signal, adjust the
       sleeping period and return to sleep.  */
    sleep = remaining;
#else  /* not HAVE_NANOSLEEP */
#ifdef HAVE_USLEEP
  /* If usleep is available, use it in preference to select.  */
  if (seconds >= 1)
    {
      /* On some systems, usleep cannot handle values larger than
	 1,000,000.  If the period is larger than that, use sleep
	 first, then add usleep for subsecond accuracy.  */
      sleep (seconds);
      seconds -= (long) seconds;
    }
  usleep (seconds * 1000000);
#else  /* not HAVE_USLEEP */
#ifdef HAVE_SELECT
  /* Note that, although Windows supports select, this sleeping
     strategy doesn't work there because Winsock's select doesn't
     implement timeout when it is passed NULL pointers for all fd
     sets.  (But it does work under Cygwin, which implements its own
     select.)  */
  struct timeval sleep;
  sleep.tv_sec = (long) seconds;
  sleep.tv_usec = 1000000 * (seconds - (long) seconds);
  select (0, NULL, NULL, NULL, &sleep);
  /* If select returns -1 and errno is EINTR, it means we were
     interrupted by a signal.  But without knowing how long we've
     actually slept, we can't return to sleep.  Using gettimeofday to
     track sleeps is slow and unreliable due to clock skew.  */
#else  /* not HAVE_SELECT */
  sleep (seconds);
#endif /* not HAVE_SELECT */
#endif /* not HAVE_USLEEP */
#endif /* not HAVE_NANOSLEEP */
}

#endif /* not WINDOWS */

/* Encode the string STR of length LENGTH to base64 format and place it
   to B64STORE.  The output will be \0-terminated, and must point to a
   writable buffer of at least 1+BASE64_LENGTH(length) bytes.  It
   returns the length of the resulting base64 data, not counting the
   terminating zero.

   This implementation will not emit newlines after 76 characters of
   base64 data.  */

int
base64_encode (const char *str, int length, char *b64store)
{
  /* Conversion table.  */
  static char tbl[64] = {
    'A','B','C','D','E','F','G','H',
    'I','J','K','L','M','N','O','P',
    'Q','R','S','T','U','V','W','X',
    'Y','Z','a','b','c','d','e','f',
    'g','h','i','j','k','l','m','n',
    'o','p','q','r','s','t','u','v',
    'w','x','y','z','0','1','2','3',
    '4','5','6','7','8','9','+','/'
  };
  int i;
  const unsigned char *s = (const unsigned char *) str;
  char *p = b64store;

  /* Transform the 3x8 bits to 4x6 bits, as required by base64.  */
  for (i = 0; i < length; i += 3)
    {
      *p++ = tbl[s[0] >> 2];
      *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
      *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
      *p++ = tbl[s[2] & 0x3f];
      s += 3;
    }

  /* Pad the result if necessary...  */
  if (i == length + 1)
    *(p - 1) = '=';
  else if (i == length + 2)
    *(p - 1) = *(p - 2) = '=';

  /* ...and zero-terminate it.  */
  *p = '\0';

  return p - b64store;
}

#define IS_ASCII(c) (((c) & 0x80) == 0)
#define IS_BASE64(c) ((IS_ASCII (c) && base64_char_to_value[c] >= 0) || c == '=')

/* Get next character from the string, except that non-base64
   characters are ignored, as mandated by rfc2045.  */
#define NEXT_BASE64_CHAR(c, p) do {			\
  c = *p++;						\
} while (c != '\0' && !IS_BASE64 (c))

/* Decode data from BASE64 (assumed to be encoded as base64) into
   memory pointed to by TO.  TO should be large enough to accomodate
   the decoded data, which is guaranteed to be less than
   strlen(base64).

   Since TO is assumed to contain binary data, it is not
   NUL-terminated.  The function returns the length of the data
   written to TO.  -1 is returned in case of error caused by malformed
   base64 input.  */

int
base64_decode (const char *base64, char *to)
{
  /* Table of base64 values for first 128 characters.  Note that this
     assumes ASCII (but so does Wget in other places).  */
  static short base64_char_to_value[128] =
    {
      -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,	/*   0-  9 */
      -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,	/*  10- 19 */
      -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,	/*  20- 29 */
      -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,  -1,	/*  30- 39 */
      -1,  -1,  -1,  62,  -1,  -1,  -1,  63,  52,  53,	/*  40- 49 */
      54,  55,  56,  57,  58,  59,  60,  61,  -1,  -1,	/*  50- 59 */
      -1,  -1,  -1,  -1,  -1,  0,   1,   2,   3,   4,	/*  60- 69 */
      5,   6,   7,   8,   9,   10,  11,  12,  13,  14,	/*  70- 79 */
      15,  16,  17,  18,  19,  20,  21,  22,  23,  24,	/*  80- 89 */
      25,  -1,  -1,  -1,  -1,  -1,  -1,  26,  27,  28,	/*  90- 99 */
      29,  30,  31,  32,  33,  34,  35,  36,  37,  38,	/* 100-109 */
      39,  40,  41,  42,  43,  44,  45,  46,  47,  48,	/* 110-119 */
      49,  50,  51,  -1,  -1,  -1,  -1,  -1		/* 120-127 */
    };

  const char *p = base64;
  char *q = to;

  while (1)
    {
      unsigned char c;
      unsigned long value;

      /* Process first byte of a quadruplet.  */
      NEXT_BASE64_CHAR (c, p);
      if (!c)
	break;
      if (c == '=')
	return -1;		/* illegal '=' while decoding base64 */
      value = base64_char_to_value[c] << 18;

      /* Process scond byte of a quadruplet.  */
      NEXT_BASE64_CHAR (c, p);
      if (!c)
	return -1;		/* premature EOF while decoding base64 */
      if (c == '=')
	return -1;		/* illegal `=' while decoding base64 */
      value |= base64_char_to_value[c] << 12;
      *q++ = value >> 16;

      /* Process third byte of a quadruplet.  */
      NEXT_BASE64_CHAR (c, p);
      if (!c)
	return -1;		/* premature EOF while decoding base64 */

      if (c == '=')
	{
	  NEXT_BASE64_CHAR (c, p);
	  if (!c)
	    return -1;		/* premature EOF while decoding base64 */
	  if (c != '=')
	    return -1;		/* padding `=' expected but not found */
	  continue;
	}

      value |= base64_char_to_value[c] << 6;
      *q++ = 0xff & value >> 8;

      /* Process fourth byte of a quadruplet.  */
      NEXT_BASE64_CHAR (c, p);
      if (!c)
	return -1;		/* premature EOF while decoding base64 */
      if (c == '=')
	continue;

      value |= base64_char_to_value[c];
      *q++ = 0xff & value;
    }

  return q - to;
}

#undef IS_ASCII
#undef IS_BASE64
#undef NEXT_BASE64_CHAR

/* Simple merge sort for use by stable_sort.  Implementation courtesy
   Zeljko Vrba with additional debugging by Nenad Barbutov.  */

static void
mergesort_internal (void *base, void *temp, size_t size, size_t from, size_t to,
		    int (*cmpfun) PARAMS ((const void *, const void *)))
{
#define ELT(array, pos) ((char *)(array) + (pos) * size)
  if (from < to)
    {
      size_t i, j, k;
      size_t mid = (to + from) / 2;
      mergesort_internal (base, temp, size, from, mid, cmpfun);
      mergesort_internal (base, temp, size, mid + 1, to, cmpfun);
      i = from;
      j = mid + 1;
      for (k = from; (i <= mid) && (j <= to); k++)
	if (cmpfun (ELT (base, i), ELT (base, j)) <= 0)
	  memcpy (ELT (temp, k), ELT (base, i++), size);
	else
	  memcpy (ELT (temp, k), ELT (base, j++), size);
      while (i <= mid)
	memcpy (ELT (temp, k++), ELT (base, i++), size);
      while (j <= to)
	memcpy (ELT (temp, k++), ELT (base, j++), size);
      for (k = from; k <= to; k++)
	memcpy (ELT (base, k), ELT (temp, k), size);
    }
#undef ELT
}

/* Stable sort with interface exactly like standard library's qsort.
   Uses mergesort internally, allocating temporary storage with
   alloca.  */

void
stable_sort (void *base, size_t nmemb, size_t size,
	     int (*cmpfun) PARAMS ((const void *, const void *)))
{
  if (size > 1)
    {
      void *temp = alloca (nmemb * size * sizeof (void *));
      mergesort_internal (base, temp, size, 0, nmemb - 1, cmpfun);
    }
}