#include <stdio.h>
#include <stdlib.h>
-#ifdef HAVE_STRING_H
-# include <string.h>
-#else /* not HAVE_STRING_H */
-# include <strings.h>
-#endif /* not HAVE_STRING_H */
-#include <sys/types.h>
+#include <string.h>
+#include <time.h>
+#ifdef HAVE_SYS_TIME_H
+# include <sys/time.h>
+#endif
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#ifdef HAVE_PWD_H
# include <pwd.h>
#endif
-#ifdef HAVE_LIMITS_H
-# include <limits.h>
-#endif
#ifdef HAVE_UTIME_H
# include <utime.h>
#endif
#endif
#include <fcntl.h>
#include <assert.h>
-#ifdef WGET_USE_STDARG
-# include <stdarg.h>
-#else
-# include <varargs.h>
+#include <stdarg.h>
+#ifdef HAVE_LOCALE_H
+# include <locale.h>
#endif
/* For TIOCGWINSZ and friends: */
#endif
/* Needed for run_with_timeout. */
-#undef USE_SIGNAL_TIMEOUT
-#ifdef HAVE_SIGNAL_H
-# include <signal.h>
-#endif
+#include <signal.h>
#ifdef HAVE_SETJMP_H
# include <setjmp.h>
#endif
# endif
#endif
-#ifdef HAVE_SIGNAL
-# ifdef HAVE_SIGSETJMP
-# define USE_SIGNAL_TIMEOUT
-# endif
-# ifdef HAVE_SIGBLOCK
-# define USE_SIGNAL_TIMEOUT
-# endif
+#undef USE_SIGNAL_TIMEOUT
+#if defined(HAVE_SIGSETJMP) || defined(HAVE_SIGBLOCK)
+# define USE_SIGNAL_TIMEOUT
#endif
#include "wget.h"
#include "utils.h"
#include "hash.h"
-#ifndef errno
-extern int errno;
-#endif
-
/* Utility function: like xstrdup(), but also lowercases S. */
char *
char *
strdupdelim (const char *beg, const char *end)
{
- char *res = (char *)xmalloc (end - beg + 1);
+ char *res = xmalloc (end - beg + 1);
memcpy (res, beg, end - beg);
res[end - beg] = '\0';
return res;
{
if (*s == ',')
{
- res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
+ res = xrealloc (res, (i + 2) * sizeof (char *));
res[i] = strdupdelim (p, s);
res[++i] = NULL;
++s;
else
++s;
}
- res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
+ res = xrealloc (res, (i + 2) * sizeof (char *));
res[i] = strdupdelim (p, s);
res[i + 1] = NULL;
return res;
}
\f
-#ifdef WGET_USE_STDARG
-# define VA_START(args, arg1) va_start (args, arg1)
-#else
-# define VA_START(args, ignored) va_start (args)
-#endif
-
/* Like sprintf, but allocates a string of sufficient size with malloc
and returns it. GNU libc has a similar function named asprintf,
which requires the pointer to the string to be passed. */
/* See log_vprintf_internal for explanation why it's OK to rely
on the return value of vsnprintf. */
- VA_START (args, fmt);
+ va_start (args, fmt);
n = vsnprintf (str, size, fmt, args);
va_end (args);
size <<= 1; /* twice the old size */
str = xrealloc (str, size);
}
- return NULL; /* unreached */
}
/* Concatenate the NULL-terminated list of string arguments into
/* Calculate the length of and allocate the resulting string. */
argcount = 0;
- VA_START (args, str0);
+ va_start (args, str0);
for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
{
int len = strlen (next_str);
/* Copy the strings into the allocated space. */
argcount = 0;
- VA_START (args, str0);
+ va_start (args, str0);
for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
{
int len;
{
pid_t pid;
/* Whether we arrange our own version of opt.lfilename here. */
- int logfile_changed = 0;
+ bool logfile_changed = false;
if (!opt.lfilename)
{
use fopen_excl) or lying to the user about the log file name
(which arises from using unique_name, printing the name, and
using fopen_excl later on.) */
- FILE *new_log_fp = unique_create (DEFAULT_LOGFILE, 0, &opt.lfilename);
+ FILE *new_log_fp = unique_create (DEFAULT_LOGFILE, false, &opt.lfilename);
if (new_log_fp)
{
- logfile_changed = 1;
+ logfile_changed = true;
fclose (new_log_fp);
}
}
else if (pid != 0)
{
/* parent, no error */
- printf (_("Continuing in background, pid %d.\n"), (int)pid);
+ printf (_("Continuing in background, pid %d.\n"), (int) pid);
if (logfile_changed)
printf (_("Output will be written to `%s'.\n"), opt.lfilename);
exit (0); /* #### should we use _exit()? */
}
#endif /* not WINDOWS */
\f
-/* "Touch" FILE, i.e. make its atime and mtime equal to the time
- specified with TM. */
+/* "Touch" FILE, i.e. make its mtime ("modified time") equal the time
+ specified with TM. The atime ("access time") is set to the current
+ time. */
+
void
touch (const char *file, time_t tm)
{
#ifdef HAVE_STRUCT_UTIMBUF
struct utimbuf times;
- times.actime = times.modtime = tm;
#else
- time_t times[2];
- times[0] = times[1] = tm;
+ struct {
+ time_t actime;
+ time_t modtime;
+ } times;
#endif
-
+ times.modtime = tm;
+ times.actime = time (NULL);
if (utime (file, ×) == -1)
logprintf (LOG_NOTQUIET, "utime(%s): %s\n", file, strerror (errno));
}
proper way should, of course, be to have a third, error state,
other than true/false, but that would introduce uncalled-for
additional complexity to the callers. */
-int
+bool
file_exists_p (const char *filename)
{
#ifdef HAVE_ACCESS
/* Returns 0 if PATH is a directory, 1 otherwise (any kind of file).
Returns 0 on error. */
-int
+bool
file_non_directory_p (const char *path)
{
struct_stat buf;
/* Use lstat() rather than stat() so that symbolic links pointing to
directories can be identified correctly. */
if (lstat (path, &buf) != 0)
- return 0;
- return S_ISDIR (buf.st_mode) ? 0 : 1;
+ return false;
+ return S_ISDIR (buf.st_mode) ? false : true;
}
/* Return the size of file named by FILENAME, or -1 if it cannot be
(and therefore doesn't need changing). */
char *
-unique_name (const char *file, int allow_passthrough)
+unique_name (const char *file, bool allow_passthrough)
{
/* If the FILE itself doesn't exist, return it without
modification. */
opening the file returned by unique_name. */
FILE *
-unique_create (const char *name, int binary, char **opened_name)
+unique_create (const char *name, bool binary, char **opened_name)
{
/* unique file name, based on NAME */
- char *uname = unique_name (name, 0);
+ char *uname = unique_name (name, false);
FILE *fp;
while ((fp = fopen_excl (uname, binary)) == NULL && errno == EEXIST)
{
xfree (uname);
- uname = unique_name (name, 0);
+ uname = unique_name (name, false);
}
if (opened_name && fp != NULL)
{
appropriately. */
FILE *
-fopen_excl (const char *fname, int binary)
+fopen_excl (const char *fname, bool binary)
{
int fd;
#ifdef O_EXCL
if (!cut)
return xstrdup (file);
- result = (char *)xmalloc (cut - base + 1 + strlen (file) + 1);
+ result = xmalloc (cut - base + 1 + strlen (file) + 1);
memcpy (result, base, cut - base);
result[cut - base] = '/';
strcpy (result + (cut - base) + 1, file);
return result;
}
\f
-static int in_acclist PARAMS ((const char *const *, const char *, int));
+static bool in_acclist (const char *const *, const char *, bool);
/* Determine whether a file is acceptable to be followed, according to
lists of patterns to accept/reject. */
-int
+bool
acceptable (const char *s)
{
int l = strlen (s);
if (opt.accepts)
{
if (opt.rejects)
- return (in_acclist ((const char *const *)opt.accepts, s, 1)
- && !in_acclist ((const char *const *)opt.rejects, s, 1));
+ return (in_acclist ((const char *const *)opt.accepts, s, true)
+ && !in_acclist ((const char *const *)opt.rejects, s, true));
else
- return in_acclist ((const char *const *)opt.accepts, s, 1);
+ return in_acclist ((const char *const *)opt.accepts, s, true);
}
else if (opt.rejects)
- return !in_acclist ((const char *const *)opt.rejects, s, 1);
- return 1;
+ return !in_acclist ((const char *const *)opt.rejects, s, true);
+ return true;
}
/* Compare S1 and S2 frontally; S2 must begin with S1. E.g. if S1 is
`/something', frontcmp() will return 1 only if S2 begins with
`/something'. Otherwise, 0 is returned. */
-int
+bool
frontcmp (const char *s1, const char *s2)
{
for (; *s1 && *s2 && (*s1 == *s2); ++s1, ++s2);
- return !*s1;
+ return *s1 == '\0';
}
/* Iterate through STRLIST, and return the first element that matches
proclist (char **strlist, const char *s, enum accd flags)
{
char **x;
-
for (x = strlist; *x; x++)
- if (has_wildcards_p (*x))
- {
- if (fnmatch (*x, s, FNM_PATHNAME) == 0)
- break;
- }
- else
- {
- char *p = *x + ((flags & ALLABS) && (**x == '/')); /* Remove '/' */
- if (frontcmp (p, s))
- break;
- }
+ {
+ /* Remove leading '/' if ALLABS */
+ char *p = *x + ((flags & ALLABS) && (**x == '/'));
+ if (has_wildcards_p (p))
+ {
+ if (fnmatch (p, s, FNM_PATHNAME) == 0)
+ break;
+ }
+ else
+ {
+ if (frontcmp (p, s))
+ break;
+ }
+ }
return *x;
}
If FLAGS is ALLABS, the leading `/' is ignored in paths; relative
and absolute paths may be freely intermixed. */
-int
+bool
accdir (const char *directory, enum accd flags)
{
/* Remove starting '/'. */
if (opt.includes)
{
if (!proclist (opt.includes, directory, flags))
- return 0;
+ return false;
}
if (opt.excludes)
{
if (proclist (opt.excludes, directory, flags))
- return 0;
+ return false;
}
- return 1;
+ return true;
}
-/* Return non-zero if STRING ends with TAIL. For instance:
+/* Return true if STRING ends with TAIL. For instance:
- match_tail ("abc", "bc", 0) -> 1
- match_tail ("abc", "ab", 0) -> 0
- match_tail ("abc", "abc", 0) -> 1
+ match_tail ("abc", "bc", false) -> 1
+ match_tail ("abc", "ab", false) -> 0
+ match_tail ("abc", "abc", false) -> 1
- If FOLD_CASE_P is non-zero, the comparison will be
- case-insensitive. */
+ If FOLD_CASE is true, the comparison will be case-insensitive. */
-int
-match_tail (const char *string, const char *tail, int fold_case_p)
+bool
+match_tail (const char *string, const char *tail, bool fold_case)
{
int i, j;
/* We want this to be fast, so we code two loops, one with
case-folding, one without. */
- if (!fold_case_p)
+ if (!fold_case)
{
for (i = strlen (string), j = strlen (tail); i >= 0 && j >= 0; i--, j--)
if (string[i] != tail[j])
/* If the tail was exhausted, the match was succesful. */
if (j == -1)
- return 1;
+ return true;
else
- return 0;
+ return false;
}
/* Checks whether string S matches each element of ACCEPTS. A list
element are matched either with fnmatch() or match_tail(),
according to whether the element contains wildcards or not.
- If the BACKWARD is 0, don't do backward comparison -- just compare
+ If the BACKWARD is false, don't do backward comparison -- just compare
them normally. */
-static int
-in_acclist (const char *const *accepts, const char *s, int backward)
+static bool
+in_acclist (const char *const *accepts, const char *s, bool backward)
{
for (; *accepts; accepts++)
{
/* fnmatch returns 0 if the pattern *does* match the
string. */
if (fnmatch (*accepts, s, 0) == 0)
- return 1;
+ return true;
}
else
{
if (backward)
{
if (match_tail (s, *accepts, 0))
- return 1;
+ return true;
}
else
{
if (!strcmp (s, *accepts))
- return 1;
+ return true;
}
}
}
- return 0;
+ return false;
}
/* Return the location of STR's suffix (file extension). Examples:
return NULL;
}
-/* Return non-zero if S contains globbing wildcards (`*', `?', `[' or
+/* Return true if S contains globbing wildcards (`*', `?', `[' or
`]'). */
-int
+bool
has_wildcards_p (const char *s)
{
for (; *s; s++)
if (*s == '*' || *s == '?' || *s == '[' || *s == ']')
- return 1;
- return 0;
+ return true;
+ return false;
}
-/* Return non-zero if FNAME ends with a typical HTML suffix. The
- following (case-insensitive) suffixes are presumed to be HTML files:
+/* Return true if FNAME ends with a typical HTML suffix. The
+ following (case-insensitive) suffixes are presumed to be HTML
+ files:
html
htm
#### CAVEAT. This is not necessarily a good indication that FNAME
refers to a file that contains HTML! */
-int
+bool
has_html_suffix_p (const char *fname)
{
char *suf;
if ((suf = suffix (fname)) == NULL)
- return 0;
+ return false;
if (!strcasecmp (suf, "html"))
- return 1;
+ return true;
if (!strcasecmp (suf, "htm"))
- return 1;
+ return true;
if (suf[0] && !strcasecmp (suf + 1, "html"))
- return 1;
- return 0;
+ return true;
+ return false;
}
/* Read a line from FP and return the pointer to freshly allocated
{
int length = 0;
int bufsize = 82;
- char *line = (char *)xmalloc (bufsize);
+ char *line = xmalloc (bufsize);
while (fgets (line + length, bufsize - length, fp))
{
int fd;
struct file_memory *fm;
long size;
- int inhibit_close = 0;
+ bool inhibit_close = false;
/* Some magic in the finest tradition of Perl and its kin: if FILE
is "-", just use stdin. */
if (HYPHENP (file))
{
fd = fileno (stdin);
- inhibit_close = 1;
+ inhibit_close = true;
/* Note that we don't inhibit mmap() in this case. If stdin is
redirected from a regular file, mmap() will still work. */
}
#ifdef HAVE_MMAP
{
- struct_stat buf;
+ struct_fstat buf;
if (fstat (fd, &buf) < 0)
goto mmap_lose;
fm->length = buf.st_size;
/* Count v2. */
for (j = 0; v2[j]; j++);
/* Reallocate v1. */
- v1 = (char **)xrealloc (v1, (i + j + 1) * sizeof (char **));
+ v1 = xrealloc (v1, (i + j + 1) * sizeof (char **));
memcpy (v1 + i, v2, (j + 1) * sizeof (char *));
xfree (v2);
return v1;
}
+
+/* Append a freshly allocated copy of STR to VEC. If VEC is NULL, it
+ is allocated as needed. Return the new value of the vector. */
+
+char **
+vec_append (char **vec, const char *str)
+{
+ int cnt; /* count of vector elements, including
+ the one we're about to append */
+ if (vec != NULL)
+ {
+ for (cnt = 0; vec[cnt]; cnt++)
+ ;
+ ++cnt;
+ }
+ else
+ cnt = 1;
+ /* Reallocate the array to fit the new element and the NULL. */
+ vec = xrealloc (vec, (cnt + 1) * sizeof (char *));
+ /* Append a copy of STR to the vector. */
+ vec[cnt - 1] = xstrdup (str);
+ vec[cnt] = NULL;
+ return vec;
+}
\f
/* Sometimes it's useful to create "sets" of strings, i.e. special
hash tables where you want to store strings as keys and merely
}
\f
-/* Add thousand separators to a number already in string form. Used
- by with_thousand_seps and with_thousand_seps_large. */
+/* Return a printed representation of N with thousand separators.
+ This should respect locale settings, with the exception of the "C"
+ locale which mandates no separator, but we use one anyway.
-static char *
-add_thousand_seps (const char *repr)
+ Unfortunately, we cannot use %'d (in fact it would be %'j) to get
+ the separators because it's too non-portable, and it's hard to test
+ for this feature at configure time. Besides, it wouldn't work in
+ the "C" locale, which many Unix users still work in. */
+
+const char *
+with_thousand_seps (wgint n)
{
static char outbuf[48];
- int i, i1, mod;
- char *outptr;
- const char *inptr;
- /* Reset the pointers. */
- outptr = outbuf;
- inptr = repr;
+ static char loc_sepchar;
+ static const char *loc_grouping;
- /* Ignore the sign for the purpose of adding thousand
- separators. */
- if (*inptr == '-')
- {
- *outptr++ = '-';
- ++inptr;
- }
- /* How many digits before the first separator? */
- mod = strlen (inptr) % 3;
- /* Insert them. */
- for (i = 0; i < mod; i++)
- *outptr++ = inptr[i];
- /* Now insert the rest of them, putting separator before every
- third digit. */
- for (i1 = i, i = 0; inptr[i1]; i++, i1++)
+ int i = 0, groupsize;
+ char *p;
+ const char *atgroup;
+
+ if (!loc_sepchar)
{
- if (i % 3 == 0 && i1 != 0)
- *outptr++ = ',';
- *outptr++ = inptr[i1];
+#ifdef LC_NUMERIC
+ /* Get the grouping character from the locale. */
+ struct lconv *lconv;
+ const char *oldlocale = setlocale (LC_NUMERIC, "");
+ lconv = localeconv ();
+ loc_sepchar = *lconv->thousands_sep;
+ loc_grouping = xstrdup (lconv->grouping);
+ /* Restore the C locale semantics of printing and reading numbers */
+ setlocale (LC_NUMERIC, oldlocale);
+ if (!loc_sepchar)
+#endif
+ /* defaults for C locale or no locale */
+ loc_sepchar = ',', loc_grouping = "\x03";
}
- /* Zero-terminate the string. */
- *outptr = '\0';
- return outbuf;
-}
-
-/* Return a static pointer to the number printed with thousand
- separators inserted at the right places. */
-
-char *
-with_thousand_seps (wgint l)
-{
- char inbuf[24];
- /* Print the number into the buffer. */
- number_to_string (inbuf, l);
- return add_thousand_seps (inbuf);
-}
-
-/* Write a string representation of LARGE_INT NUMBER into the provided
- buffer.
+ atgroup = loc_grouping;
- It would be dangerous to use sprintf, because the code wouldn't
- work on a machine with gcc-provided long long support, but without
- libc support for "%lld". However, such old systems platforms
- typically lack snprintf and will end up using our version, which
- does support "%lld" whereever long longs are available. */
+ p = outbuf + sizeof outbuf;
+ *--p = '\0';
+ groupsize = *atgroup++;
-static void
-large_int_to_string (char *buffer, int bufsize, LARGE_INT number)
-{
- snprintf (buffer, bufsize, LARGE_INT_FMT, number);
-}
-
-/* The same as with_thousand_seps, but works on LARGE_INT. */
-
-char *
-with_thousand_seps_large (LARGE_INT l)
-{
- char inbuf[48];
- large_int_to_string (inbuf, sizeof (inbuf), l);
- return add_thousand_seps (inbuf);
+ while (1)
+ {
+ *--p = n % 10 + '0';
+ n /= 10;
+ if (n == 0)
+ break;
+ /* Insert the separator on every groupsize'd digit, and get the
+ new groupsize. */
+ if (++i == groupsize)
+ {
+ *--p = loc_sepchar;
+ i = 0;
+ if (*atgroup)
+ groupsize = *atgroup++;
+ }
+ }
+ return p;
}
/* N, a byte quantity, is converted to a human-readable abberviated
usually improves readability."
This intentionally uses kilobyte (KB), megabyte (MB), etc. in their
- original computer science meaning of "multiples of 1024".
- Multiples of 1000 would be useless since Wget already adds thousand
- separators for legibility. We don't use the "*bibyte" names
- invented in 1998, and seldom used in practice. Wikipedia's entry
- on kilobyte discusses this in some detail. */
+ original computer-related meaning of "powers of 1024". Powers of
+ 1000 would be useless since Wget already displays sizes with
+ thousand separators. We don't use the "*bibyte" names invented in
+ 1998, and seldom used in practice. Wikipedia's entry on kilobyte
+ discusses this in some detail. */
char *
-human_readable (wgint n)
+human_readable (HR_NUMTYPE n)
{
/* These suffixes are compatible with those of GNU `ls -lh'. */
static char powers[] =
/* At each iteration N is greater than the *subsequent* power.
That way N/1024.0 produces a decimal number in the units of
*this* power. */
- if ((n >> 10) < 1024 || i == countof (powers) - 1)
+ if ((n / 1024) < 1024 || i == countof (powers) - 1)
{
/* Must cast to long first because MS VC can't directly cast
__int64 to double. (This is safe because N is known to
- be <2**20.) */
+ be < 1024^2, so always fits into long.) */
double val = (double) (long) n / 1024.0;
/* Print values smaller than 10 with one decimal digits, and
others without any decimals. */
val < 10 ? 1 : 0, val, powers[i]);
return buf;
}
- n >>= 10;
+ n /= 1024;
}
return NULL; /* unreached */
}
would just use "%j" and intmax_t, but many systems don't support
it, so it's used only if nothing else works. */
#if SIZEOF_LONG >= SIZEOF_WGINT
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%ld", (long) (n))
+# define SPRINTF_WGINT(buf, n) sprintf (buf, "%ld", (long) (n))
+#elif SIZEOF_LONG_LONG >= SIZEOF_WGINT
+# define SPRINTF_WGINT(buf, n) sprintf (buf, "%lld", (long long) (n))
+#elif defined(WINDOWS)
+# define SPRINTF_WGINT(buf, n) sprintf (buf, "%I64d", (__int64) (n))
#else
-# if SIZEOF_LONG_LONG >= SIZEOF_WGINT
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%lld", (long long) (n))
-# else
-# ifdef WINDOWS
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%I64", (__int64) (n))
-# else
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%j", (intmax_t) (n))
-# endif
-# endif
+# define SPRINTF_WGINT(buf, n) sprintf (buf, "%j", (intmax_t) (n))
#endif
/* Shorthand for casting to wgint. */
/* wgint is 32 bits wide: no number has more than 10 digits. */
else DIGITS_10 (1000000000);
#else
- /* wgint is 64 bits wide: handle numbers with more than 9 decimal
- digits. Constants are constructed by compile-time multiplication
- to avoid dealing with different notations for 64-bit constants
- (nnnL, nnnLL, and nnnI64, depending on the compiler). */
+ /* wgint is 64 bits wide: handle numbers with 9-19 decimal digits.
+ Constants are constructed by compile-time multiplication to avoid
+ dealing with different notations for 64-bit constants
+ (nL/nLL/nI64, depending on the compiler and architecture). */
else if (n < 10*(W)1000000000) DIGITS_10 (1000000000);
else if (n < 100*(W)1000000000) DIGITS_11 (10*(W)1000000000);
else if (n < 1000*(W)1000000000) DIGITS_12 (100*(W)1000000000);
return buf;
}
\f
-/* Support for timers. */
-
-#undef TIMER_WINDOWS
-#undef TIMER_GETTIMEOFDAY
-#undef TIMER_TIME
-
-/* Depending on the OS and availability of gettimeofday(), one and
- only one of the above constants will be defined. Virtually all
- modern Unix systems will define TIMER_GETTIMEOFDAY; Windows will
- use TIMER_WINDOWS. TIMER_TIME is a catch-all method for
- non-Windows systems without gettimeofday. */
-
-#ifdef WINDOWS
-# define TIMER_WINDOWS
-#else /* not WINDOWS */
-# ifdef HAVE_GETTIMEOFDAY
-# define TIMER_GETTIMEOFDAY
-# else
-# define TIMER_TIME
-# endif
-#endif /* not WINDOWS */
-
-#ifdef TIMER_GETTIMEOFDAY
-typedef struct timeval wget_sys_time;
-#endif
-
-#ifdef TIMER_TIME
-typedef time_t wget_sys_time;
-#endif
-
-#ifdef TIMER_WINDOWS
-typedef union {
- DWORD lores; /* In case GetTickCount is used */
- LARGE_INTEGER hires; /* In case high-resolution timer is used */
-} wget_sys_time;
-#endif
-
-struct wget_timer {
- /* Whether the start time has been initialized. */
- int initialized;
-
- /* The starting point in time which, subtracted from the current
- time, yields elapsed time. */
- wget_sys_time start;
-
- /* The most recent elapsed time, calculated by wtimer_update().
- Measured in milliseconds. */
- double elapsed_last;
-
- /* Approximately, the time elapsed between the true start of the
- measurement and the time represented by START. */
- double elapsed_pre_start;
-};
-
-#ifdef TIMER_WINDOWS
-
-/* Whether high-resolution timers are used. Set by wtimer_initialize_once
- the first time wtimer_allocate is called. */
-static int using_hires_timers;
-
-/* Frequency of high-resolution timers -- number of updates per
- millisecond. Calculated the first time wtimer_allocate is called
- provided that high-resolution timers are available. */
-static double hires_millisec_freq;
-
-/* The first time a timer is created, determine whether to use
- high-resolution timers. */
-
-static void
-wtimer_initialize_once (void)
-{
- static int init_done;
- if (!init_done)
- {
- LARGE_INTEGER freq;
- init_done = 1;
- freq.QuadPart = 0;
- QueryPerformanceFrequency (&freq);
- if (freq.QuadPart != 0)
- {
- using_hires_timers = 1;
- hires_millisec_freq = (double) freq.QuadPart / 1000.0;
- }
- }
-}
-#endif /* TIMER_WINDOWS */
-
-/* Allocate a timer. Calling wtimer_read on the timer will return
- zero. It is not legal to call wtimer_update with a freshly
- allocated timer -- use wtimer_reset first. */
-
-struct wget_timer *
-wtimer_allocate (void)
-{
- struct wget_timer *wt = xnew (struct wget_timer);
- xzero (*wt);
-
-#ifdef TIMER_WINDOWS
- wtimer_initialize_once ();
-#endif
-
- return wt;
-}
-
-/* Allocate a new timer and reset it. Return the new timer. */
-
-struct wget_timer *
-wtimer_new (void)
-{
- struct wget_timer *wt = wtimer_allocate ();
- wtimer_reset (wt);
- return wt;
-}
-
-/* Free the resources associated with the timer. Its further use is
- prohibited. */
-
-void
-wtimer_delete (struct wget_timer *wt)
-{
- xfree (wt);
-}
-
-/* Store system time to WST. */
-
-static void
-wtimer_sys_set (wget_sys_time *wst)
-{
-#ifdef TIMER_GETTIMEOFDAY
- gettimeofday (wst, NULL);
-#endif
-
-#ifdef TIMER_TIME
- time (wst);
-#endif
-
-#ifdef TIMER_WINDOWS
- if (using_hires_timers)
- {
- QueryPerformanceCounter (&wst->hires);
- }
- else
- {
- /* Where hires counters are not available, use GetTickCount rather
- GetSystemTime, because it is unaffected by clock skew and simpler
- to use. Note that overflows don't affect us because we never use
- absolute values of the ticker, only the differences. */
- wst->lores = GetTickCount ();
- }
-#endif
-}
-
-/* Reset timer WT. This establishes the starting point from which
- wtimer_read() will return the number of elapsed milliseconds.
- It is allowed to reset a previously used timer. */
-
-void
-wtimer_reset (struct wget_timer *wt)
-{
- /* Set the start time to the current time. */
- wtimer_sys_set (&wt->start);
- wt->elapsed_last = 0;
- wt->elapsed_pre_start = 0;
- wt->initialized = 1;
-}
-
-static double
-wtimer_sys_diff (wget_sys_time *wst1, wget_sys_time *wst2)
-{
-#ifdef TIMER_GETTIMEOFDAY
- return ((double)(wst1->tv_sec - wst2->tv_sec) * 1000
- + (double)(wst1->tv_usec - wst2->tv_usec) / 1000);
-#endif
-
-#ifdef TIMER_TIME
- return 1000 * (*wst1 - *wst2);
-#endif
-
-#ifdef WINDOWS
- if (using_hires_timers)
- return (wst1->hires.QuadPart - wst2->hires.QuadPart) / hires_millisec_freq;
- else
- return wst1->lores - wst2->lores;
-#endif
-}
-
-/* Update the timer's elapsed interval. This function causes the
- timer to call gettimeofday (or time(), etc.) to update its idea of
- current time. To get the elapsed interval in milliseconds, use
- wtimer_read.
-
- This function handles clock skew, i.e. time that moves backwards is
- ignored. */
-
-void
-wtimer_update (struct wget_timer *wt)
-{
- wget_sys_time now;
- double elapsed;
-
- assert (wt->initialized != 0);
-
- wtimer_sys_set (&now);
- elapsed = wt->elapsed_pre_start + wtimer_sys_diff (&now, &wt->start);
-
- /* Ideally we'd just return the difference between NOW and
- wt->start. However, the system timer can be set back, and we
- could return a value smaller than when we were last called, even
- a negative value. Both of these would confuse the callers, which
- expect us to return monotonically nondecreasing values.
-
- Therefore: if ELAPSED is smaller than its previous known value,
- we reset wt->start to the current time and effectively start
- measuring from this point. But since we don't want the elapsed
- value to start from zero, we set elapsed_pre_start to the last
- elapsed time and increment all future calculations by that
- amount. */
-
- if (elapsed < wt->elapsed_last)
- {
- wt->start = now;
- wt->elapsed_pre_start = wt->elapsed_last;
- elapsed = wt->elapsed_last;
- }
-
- wt->elapsed_last = elapsed;
-}
-
-/* Return the elapsed time in milliseconds between the last call to
- wtimer_reset and the last call to wtimer_update.
-
- A typical use of the timer interface would be:
-
- struct wtimer *timer = wtimer_new ();
- ... do something that takes a while ...
- wtimer_update ();
- double msecs = wtimer_read (); */
-
-double
-wtimer_read (const struct wget_timer *wt)
-{
- return wt->elapsed_last;
-}
-
-/* Return the assessed granularity of the timer implementation, in
- milliseconds. This is used by code that tries to substitute a
- better value for timers that have returned zero. */
-
-double
-wtimer_granularity (void)
-{
-#ifdef TIMER_GETTIMEOFDAY
- /* Granularity of gettimeofday varies wildly between architectures.
- However, it appears that on modern machines it tends to be better
- than 1ms. Assume 100 usecs. (Perhaps the configure process
- could actually measure this?) */
- return 0.1;
-#endif
-
-#ifdef TIMER_TIME
- return 1000;
-#endif
-
-#ifdef TIMER_WINDOWS
- if (using_hires_timers)
- return 1.0 / hires_millisec_freq;
- else
- return 10; /* according to MSDN */
-#endif
-}
-\f
-/* This should probably be at a better place, but it doesn't really
- fit into html-parse.c. */
-
-/* The function returns the pointer to the malloc-ed quoted version of
- string s. It will recognize and quote numeric and special graphic
- entities, as per RFC1866:
-
- `&' -> `&'
- `<' -> `<'
- `>' -> `>'
- `"' -> `"'
- SP -> ` '
-
- No other entities are recognized or replaced. */
-char *
-html_quote_string (const char *s)
-{
- const char *b = s;
- char *p, *res;
- int i;
-
- /* Pass through the string, and count the new size. */
- for (i = 0; *s; s++, i++)
- {
- if (*s == '&')
- i += 4; /* `amp;' */
- else if (*s == '<' || *s == '>')
- i += 3; /* `lt;' and `gt;' */
- else if (*s == '\"')
- i += 5; /* `quot;' */
- else if (*s == ' ')
- i += 4; /* #32; */
- }
- res = (char *)xmalloc (i + 1);
- s = b;
- for (p = res; *s; s++)
- {
- switch (*s)
- {
- case '&':
- *p++ = '&';
- *p++ = 'a';
- *p++ = 'm';
- *p++ = 'p';
- *p++ = ';';
- break;
- case '<': case '>':
- *p++ = '&';
- *p++ = (*s == '<' ? 'l' : 'g');
- *p++ = 't';
- *p++ = ';';
- break;
- case '\"':
- *p++ = '&';
- *p++ = 'q';
- *p++ = 'u';
- *p++ = 'o';
- *p++ = 't';
- *p++ = ';';
- break;
- case ' ':
- *p++ = '&';
- *p++ = '#';
- *p++ = '3';
- *p++ = '2';
- *p++ = ';';
- break;
- default:
- *p++ = *s;
- }
- }
- *p = '\0';
- return res;
-}
-
/* Determine the width of the terminal we're running on. If that's
not possible, return 0. */
return 0; /* most likely ENOTTY */
return wsz.ws_col;
-#else /* not TIOCGWINSZ */
-# ifdef WINDOWS
+#elif defined(WINDOWS)
CONSOLE_SCREEN_BUFFER_INFO csbi;
if (!GetConsoleScreenBufferInfo (GetStdHandle (STD_ERROR_HANDLE), &csbi))
return 0;
return csbi.dwSize.X;
-# else /* neither WINDOWS nor TIOCGWINSZ */
+#else /* neither TIOCGWINSZ nor WINDOWS */
return 0;
-#endif /* neither WINDOWS nor TIOCGWINSZ */
-#endif /* not TIOCGWINSZ */
+#endif /* neither TIOCGWINSZ nor WINDOWS */
}
/* Return a random number between 0 and MAX-1, inclusive.
static sigjmp_buf run_with_timeout_env;
-static RETSIGTYPE
+static void
abort_run_with_timeout (int sig)
{
assert (sig == SIGALRM);
static jmp_buf run_with_timeout_env;
-static RETSIGTYPE
+static void
abort_run_with_timeout (int sig)
{
assert (sig == SIGALRM);
}
/* 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.
+ seconds. Returns true if the function was interrupted with a
+ timeout, false otherwise.
This works by setting up SIGALRM to be delivered in TIMEOUT seconds
using setitimer() or alarm(). The timeout is enforced by
are normally freed prior to exit from the functions, they will be
lost in case of timeout. */
-int
+bool
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
int saved_errno;
if (timeout == 0)
{
fun (arg);
- return 0;
+ return false;
}
signal (SIGALRM, abort_run_with_timeout);
{
/* Longjumped out of FUN with a timeout. */
signal (SIGALRM, SIG_DFL);
- return 1;
+ return true;
}
alarm_set (timeout);
fun (arg);
signal (SIGALRM, SIG_DFL);
errno = saved_errno;
- return 0;
+ return false;
}
#else /* not USE_SIGNAL_TIMEOUT */
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
fun (arg);
- return 0;
+ return false;
}
#endif /* not WINDOWS */
#endif /* not USE_SIGNAL_TIMEOUT */
/* 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
+#elif defined(HAVE_USLEEP)
/* If usleep is available, use it in preference to select. */
if (seconds >= 1)
{
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.) */
+#else /* fall back select */
+ /* Note that, although Windows supports select, it can't be used to
+ implement sleeping because Winsock's select doesn't implement
+ timeout when it is passed NULL pointers for all fd sets. (But it
+ does under Cygwin, which implements Unix-compatible select.) */
struct timeval sleep;
sleep.tv_sec = (long) seconds;
sleep.tv_usec = 1000000 * (seconds - (long) seconds);
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
}
#endif /* not WINDOWS */
-/* Encode the string S of length LENGTH to base64 format and place it
- to STORE. STORE will be 0-terminated, and must point to a writable
- buffer of at least 1+BASE64_LENGTH(length) bytes. */
+/* 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.
-void
-base64_encode (const char *s, char *store, int length)
+ 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] = {
'4','5','6','7','8','9','+','/'
};
int i;
- unsigned char *p = (unsigned char *)store;
+ 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[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)
int
base64_decode (const char *base64, char *to)
{
- /* Table of base64 values for first 128 characters. */
+ /* 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 */
{
NEXT_BASE64_CHAR (c, p);
if (!c)
- return -1; /* premature EOF while dcoding base64 */
+ return -1; /* premature EOF while decoding base64 */
if (c != '=')
return -1; /* padding `=' expected but not found */
continue;
/* Process fourth byte of a quadruplet. */
NEXT_BASE64_CHAR (c, p);
if (!c)
- return -1; /* premature EOF while dcoding base64 */
+ return -1; /* premature EOF while decoding base64 */
if (c == '=')
continue;
#undef IS_ASCII
#undef IS_BASE64
#undef NEXT_BASE64_CHAR
+\f
+/* 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) (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) (const void *, const void *))
+{
+ if (size > 1)
+ {
+ void *temp = alloca (nmemb * size * sizeof (void *));
+ mergesort_internal (base, temp, size, 0, nmemb - 1, cmpfun);
+ }
+}
projects / wget / blobdiff