-/* Various functions of utilitarian nature.
- Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001
- Free Software Foundation, Inc.
+/* Various utility functions.
+ Copyright (C) 1996-2006 Free Software Foundation, Inc.
This file is part of GNU Wget.
GNU Wget is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
GNU Wget is distributed in the hope that it will be useful,
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
-along with Wget; if not, write to the Free Software
-Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+along with Wget. If not, see <http://www.gnu.org/licenses/>.
In addition, as a special exception, the Free Software Foundation
gives permission to link the code of its release of Wget with the
#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_MMAP
# include <sys/mman.h>
#endif
-#ifdef HAVE_PWD_H
-# include <pwd.h>
-#endif
-#include <limits.h>
#ifdef HAVE_UTIME_H
# include <utime.h>
#endif
# include <sys/utime.h>
#endif
#include <errno.h>
-#ifdef NeXT
-# include <libc.h> /* for access() */
-#endif
#include <fcntl.h>
#include <assert.h>
+#include <stdarg.h>
+#include <locale.h>
/* For TIOCGWINSZ and friends: */
#ifdef HAVE_SYS_IOCTL_H
# include <termios.h>
#endif
-/* Needed for run_with_timeout. */
-#undef USE_SIGNAL_TIMEOUT
-#ifdef HAVE_SIGNAL_H
-# include <signal.h>
-#endif
-#ifdef HAVE_SETJMP_H
-# include <setjmp.h>
-#endif
+/* Needed for Unix version of run_with_timeout. */
+#include <signal.h>
+#include <setjmp.h>
+
+#ifndef HAVE_SIGSETJMP
/* If sigsetjmp is a macro, configure won't pick it up. */
-#ifdef sigsetjmp
-# define HAVE_SIGSETJMP
-#endif
-#ifdef HAVE_SIGNAL
-# ifdef HAVE_SIGSETJMP
-# define USE_SIGNAL_TIMEOUT
-# endif
-# ifdef HAVE_SIGBLOCK
-# define USE_SIGNAL_TIMEOUT
+# ifdef sigsetjmp
+# define HAVE_SIGSETJMP
# endif
#endif
+#if defined HAVE_SIGSETJMP || defined HAVE_SIGBLOCK
+# define USE_SIGNAL_TIMEOUT
+#endif
+
#include "wget.h"
#include "utils.h"
-#include "fnmatch.h"
#include "hash.h"
-#ifndef errno
-extern int errno;
-#endif
-
-/* This section implements several wrappers around the basic
- allocation routines. This is done for two reasons: first, so that
- the callers of these functions need not consistently check for
- errors. If there is not enough virtual memory for running Wget,
- something is seriously wrong, and Wget exits with an appropriate
- error message.
-
- The second reason why these are useful is that, if DEBUG_MALLOC is
- defined, they also provide a handy (if crude) malloc debugging
- interface that checks memory leaks. */
-
-/* Croak the fatal memory error and bail out with non-zero exit
- status. */
-static void
-memfatal (const char *what)
-{
- /* Make sure we don't try to store part of the log line, and thus
- call malloc. */
- log_set_save_context (0);
- logprintf (LOG_ALWAYS, _("%s: %s: Not enough memory.\n"), exec_name, what);
- exit (1);
-}
-
-/* These functions end with _real because they need to be
- distinguished from the debugging functions, and from the macros.
- Explanation follows:
-
- If memory debugging is not turned on, wget.h defines these:
-
- #define xmalloc xmalloc_real
- #define xrealloc xrealloc_real
- #define xstrdup xstrdup_real
- #define xfree free
-
- In case of memory debugging, the definitions are a bit more
- complex, because we want to provide more information, *and* we want
- to call the debugging code. (The former is the reason why xmalloc
- and friends need to be macros in the first place.) Then it looks
- like this:
-
- #define xmalloc(a) xmalloc_debug (a, __FILE__, __LINE__)
- #define xfree(a) xfree_debug (a, __FILE__, __LINE__)
- #define xrealloc(a, b) xrealloc_debug (a, b, __FILE__, __LINE__)
- #define xstrdup(a) xstrdup_debug (a, __FILE__, __LINE__)
-
- Each of the *_debug function does its magic and calls the real one. */
-
-#ifdef DEBUG_MALLOC
-# define STATIC_IF_DEBUG static
-#else
-# define STATIC_IF_DEBUG
-#endif
-
-STATIC_IF_DEBUG void *
-xmalloc_real (size_t size)
-{
- void *ptr = malloc (size);
- if (!ptr)
- memfatal ("malloc");
- return ptr;
-}
-
-STATIC_IF_DEBUG void *
-xrealloc_real (void *ptr, size_t newsize)
-{
- void *newptr;
-
- /* Not all Un*xes have the feature of realloc() that calling it with
- a NULL-pointer is the same as malloc(), but it is easy to
- simulate. */
- if (ptr)
- newptr = realloc (ptr, newsize);
- else
- newptr = malloc (newsize);
- if (!newptr)
- memfatal ("realloc");
- return newptr;
-}
-
-STATIC_IF_DEBUG char *
-xstrdup_real (const char *s)
-{
- char *copy;
-
-#ifndef HAVE_STRDUP
- int l = strlen (s);
- copy = malloc (l + 1);
- if (!copy)
- memfatal ("strdup");
- memcpy (copy, s, l + 1);
-#else /* HAVE_STRDUP */
- copy = strdup (s);
- if (!copy)
- memfatal ("strdup");
-#endif /* HAVE_STRDUP */
-
- return copy;
-}
-
-#ifdef DEBUG_MALLOC
-
-/* Crude home-grown routines for debugging some malloc-related
- problems. Featured:
-
- * Counting the number of malloc and free invocations, and reporting
- the "balance", i.e. how many times more malloc was called than it
- was the case with free.
-
- * Making malloc store its entry into a simple array and free remove
- stuff from that array. At the end, print the pointers which have
- not been freed, along with the source file and the line number.
- This also has the side-effect of detecting freeing memory that
- was never allocated.
-
- Note that this kind of memory leak checking strongly depends on
- every malloc() being followed by a free(), even if the program is
- about to finish. Wget is careful to free the data structure it
- allocated in init.c. */
-
-static int malloc_count, free_count;
-
-static struct {
- char *ptr;
- const char *file;
- int line;
-} malloc_debug[100000];
-
-/* Both register_ptr and unregister_ptr take O(n) operations to run,
- which can be a real problem. It would be nice to use a hash table
- for malloc_debug, but the functions in hash.c are not suitable
- because they can call malloc() themselves. Maybe it would work if
- the hash table were preallocated to a huge size, and if we set the
- rehash threshold to 1.0. */
-
-/* Register PTR in malloc_debug. Abort if this is not possible
- (presumably due to the number of current allocations exceeding the
- size of malloc_debug.) */
-
-static void
-register_ptr (void *ptr, const char *file, int line)
-{
- int i;
- for (i = 0; i < countof (malloc_debug); i++)
- if (malloc_debug[i].ptr == NULL)
- {
- malloc_debug[i].ptr = ptr;
- malloc_debug[i].file = file;
- malloc_debug[i].line = line;
- return;
- }
- abort ();
-}
-
-/* Unregister PTR from malloc_debug. Abort if PTR is not present in
- malloc_debug. (This catches calling free() with a bogus pointer.) */
-
-static void
-unregister_ptr (void *ptr)
-{
- int i;
- for (i = 0; i < countof (malloc_debug); i++)
- if (malloc_debug[i].ptr == ptr)
- {
- malloc_debug[i].ptr = NULL;
- return;
- }
- abort ();
-}
-
-/* Print the malloc debug stats that can be gathered from the above
- information. Currently this is the count of mallocs, frees, the
- difference between the two, and the dump of the contents of
- malloc_debug. The last part are the memory leaks. */
-
-void
-print_malloc_debug_stats (void)
-{
- int i;
- printf ("\nMalloc: %d\nFree: %d\nBalance: %d\n\n",
- malloc_count, free_count, malloc_count - free_count);
- for (i = 0; i < countof (malloc_debug); i++)
- if (malloc_debug[i].ptr != NULL)
- printf ("0x%08ld: %s:%d\n", (long)malloc_debug[i].ptr,
- malloc_debug[i].file, malloc_debug[i].line);
-}
-
-void *
-xmalloc_debug (size_t size, const char *source_file, int source_line)
-{
- void *ptr = xmalloc_real (size);
- ++malloc_count;
- register_ptr (ptr, source_file, source_line);
- return ptr;
-}
-
-void
-xfree_debug (void *ptr, const char *source_file, int source_line)
-{
- assert (ptr != NULL);
- ++free_count;
- unregister_ptr (ptr);
- free (ptr);
-}
+#ifdef TESTING
+#include "test.h"
+#endif
-void *
-xrealloc_debug (void *ptr, size_t newsize, const char *source_file, int source_line)
-{
- void *newptr = xrealloc_real (ptr, newsize);
- if (!ptr)
- {
- ++malloc_count;
- register_ptr (newptr, source_file, source_line);
- }
- else if (newptr != ptr)
- {
- unregister_ptr (ptr);
- register_ptr (newptr, source_file, source_line);
- }
- return newptr;
-}
-
-char *
-xstrdup_debug (const char *s, const char *source_file, int source_line)
-{
- char *copy = xstrdup_real (s);
- ++malloc_count;
- register_ptr (copy, source_file, source_line);
- return copy;
-}
-
-#endif /* DEBUG_MALLOC */
-\f
/* Utility function: like xstrdup(), but also lowercases S. */
char *
return copy;
}
-/* Return a count of how many times CHR occurs in STRING. */
-
-int
-count_char (const char *string, char chr)
-{
- const char *p;
- int count = 0;
- for (p = string; *p; p++)
- if (*p == chr)
- ++count;
- return count;
-}
-
/* Copy the string formed by two pointers (one on the beginning, other
on the char after the last char) to a new, malloc-ed location.
0-terminate it. */
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
-/* Return pointer to a static char[] buffer in which zero-terminated
- string-representation of TM (in form hh:mm:ss) is printed.
-
- If TM is non-NULL, the current time-in-seconds will be stored
- there.
+/* Like sprintf, but prints into a string of sufficient size freshly
+ allocated with malloc, which is returned. If unable to print due
+ to invalid format, returns NULL. Inability to allocate needed
+ memory results in abort, as with xmalloc. This is in spirit
+ similar to the GNU/BSD extension asprintf, but somewhat easier to
+ use.
- (#### This is misleading: one would expect TM would be used instead
- of the current time in that case. This design was probably
- influenced by the design time(2), and should be changed at some
- points. No callers use non-NULL TM anyway.) */
+ Internally the function either calls vasprintf or loops around
+ vsnprintf until the correct size is found. Since Wget also ships a
+ fallback implementation of vsnprintf, this should be portable. */
char *
-time_str (time_t *tm)
-{
- static char output[15];
- struct tm *ptm;
- time_t secs = time (tm);
+aprintf (const char *fmt, ...)
+{
+#if defined HAVE_VASPRINTF && !defined DEBUG_MALLOC
+ /* Use vasprintf. */
+ int ret;
+ va_list args;
+ char *str;
+ va_start (args, fmt);
+ ret = vasprintf (&str, fmt, args);
+ va_end (args);
+ if (ret < 0 && errno == ENOMEM)
+ abort (); /* for consistency with xmalloc/xrealloc */
+ else if (ret < 0)
+ return NULL;
+ return str;
+#else /* not HAVE_VASPRINTF */
+
+ /* vasprintf is unavailable. snprintf into a small buffer and
+ resize it as necessary. */
+ int size = 32;
+ char *str = xmalloc (size);
+
+ /* #### This code will infloop and eventually abort in xrealloc if
+ passed a FMT that causes snprintf to consistently return -1. */
- if (secs == -1)
+ while (1)
{
- /* In case of error, return the empty string. Maybe we should
- just abort if this happens? */
- *output = '\0';
- return output;
+ int n;
+ va_list args;
+
+ va_start (args, fmt);
+ n = vsnprintf (str, size, fmt, args);
+ va_end (args);
+
+ /* If the printing worked, return the string. */
+ if (n > -1 && n < size)
+ return str;
+
+ /* Else try again with a larger buffer. */
+ if (n > -1) /* C99 */
+ size = n + 1; /* precisely what is needed */
+ else
+ size <<= 1; /* twice the old size */
+ str = xrealloc (str, size);
}
- ptm = localtime (&secs);
- sprintf (output, "%02d:%02d:%02d", ptm->tm_hour, ptm->tm_min, ptm->tm_sec);
- return output;
+#endif /* not HAVE_VASPRINTF */
}
-/* Like the above, but include the date: YYYY-MM-DD hh:mm:ss. */
+/* Concatenate the NULL-terminated list of string arguments into
+ freshly allocated space. */
char *
-datetime_str (time_t *tm)
+concat_strings (const char *str0, ...)
{
- static char output[20]; /* "YYYY-MM-DD hh:mm:ss" + \0 */
- struct tm *ptm;
- time_t secs = time (tm);
+ va_list args;
+ int saved_lengths[5]; /* inspired by Apache's apr_pstrcat */
+ char *ret, *p;
+
+ const char *next_str;
+ int total_length = 0;
+ int argcount;
- if (secs == -1)
+ /* Calculate the length of and allocate the resulting string. */
+
+ argcount = 0;
+ va_start (args, str0);
+ for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
+ {
+ int len = strlen (next_str);
+ if (argcount < countof (saved_lengths))
+ saved_lengths[argcount++] = len;
+ total_length += len;
+ }
+ va_end (args);
+ p = ret = xmalloc (total_length + 1);
+
+ /* Copy the strings into the allocated space. */
+
+ argcount = 0;
+ va_start (args, str0);
+ for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
{
- /* In case of error, return the empty string. Maybe we should
- just abort if this happens? */
- *output = '\0';
- return output;
+ int len;
+ if (argcount < countof (saved_lengths))
+ len = saved_lengths[argcount++];
+ else
+ len = strlen (next_str);
+ memcpy (p, next_str, len);
+ p += len;
}
- ptm = localtime (&secs);
- sprintf (output, "%04d-%02d-%02d %02d:%02d:%02d",
- ptm->tm_year + 1900, ptm->tm_mon + 1, ptm->tm_mday,
- ptm->tm_hour, ptm->tm_min, ptm->tm_sec);
+ va_end (args);
+ *p = '\0';
+
+ return ret;
+}
+\f
+/* Format the provided time according to the specified format. The
+ format is a string with format elements supported by strftime. */
+
+static char *
+fmttime (time_t t, const char *fmt)
+{
+ static char output[32];
+ struct tm *tm = localtime(&t);
+ if (!tm)
+ abort ();
+ if (!strftime(output, sizeof(output), fmt, tm))
+ abort ();
return output;
}
+
+/* Return pointer to a static char[] buffer in which zero-terminated
+ string-representation of TM (in form hh:mm:ss) is printed.
+
+ If TM is NULL, the current time will be used. */
+
+char *
+time_str (time_t t)
+{
+ return fmttime(t, "%H:%M:%S");
+}
+
+/* Like the above, but include the date: YYYY-MM-DD hh:mm:ss. */
+
+char *
+datetime_str (time_t t)
+{
+ return fmttime(t, "%Y-%m-%d %H:%M:%S");
+}
\f
/* The Windows versions of the following two functions are defined in
mswindows.c. */
{
pid_t pid;
/* Whether we arrange our own version of opt.lfilename here. */
- int changedp = 0;
+ bool logfile_changed = false;
if (!opt.lfilename)
{
- opt.lfilename = unique_name (DEFAULT_LOGFILE, 0);
- changedp = 1;
+ /* We must create the file immediately to avoid either a race
+ condition (which arises from using unique_name and failing to
+ 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, false, &opt.lfilename);
+ if (new_log_fp)
+ {
+ logfile_changed = true;
+ fclose (new_log_fp);
+ }
}
pid = fork ();
if (pid < 0)
else if (pid != 0)
{
/* parent, no error */
- printf (_("Continuing in background, pid %d.\n"), (int)pid);
- if (changedp)
+ 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));
}
remove_link (const char *file)
{
int err = 0;
- struct stat st;
+ struct_stat st;
if (lstat (file, &st) == 0 && S_ISLNK (st.st_mode))
{
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
return access (filename, F_OK) >= 0;
#else
- struct stat buf;
+ struct_stat buf;
return stat (filename, &buf) >= 0;
#endif
}
/* 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;
+ 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
opened or seeked into. */
-long
+wgint
file_size (const char *filename)
{
- long size;
+#if defined(HAVE_FSEEKO) && defined(HAVE_FTELLO)
+ wgint size;
/* We use fseek rather than stat to determine the file size because
- that way we can also verify whether the file is readable.
- Inspired by the POST patch by Arnaud Wylie. */
+ that way we can also verify that the file is readable without
+ explicitly checking for permissions. Inspired by the POST patch
+ by Arnaud Wylie. */
FILE *fp = fopen (filename, "rb");
- fseek (fp, 0, SEEK_END);
- size = ftell (fp);
+ if (!fp)
+ return -1;
+ fseeko (fp, 0, SEEK_END);
+ size = ftello (fp);
fclose (fp);
return size;
+#else
+ struct_stat st;
+ if (stat (filename, &st) < 0)
+ return -1;
+ return st.st_size;
+#endif
}
/* stat file names named PREFIX.1, PREFIX.2, etc., until one that
exist at the point in time when the function was called.
Therefore, where security matters, don't rely that the file created
by this function exists until you open it with O_EXCL or
- something.
+ equivalent.
If ALLOW_PASSTHROUGH is 0, it always returns a freshly allocated
string. Otherwise, it may return FILE if the file doesn't exist
(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. */
and return it. */
return unique_name_1 (file);
}
+
+/* Create a file based on NAME, except without overwriting an existing
+ file with that name. Providing O_EXCL is correctly implemented,
+ this function does not have the race condition associated with
+ opening the file returned by unique_name. */
+
+FILE *
+unique_create (const char *name, bool binary, char **opened_name)
+{
+ /* unique file name, based on NAME */
+ char *uname = unique_name (name, false);
+ FILE *fp;
+ while ((fp = fopen_excl (uname, binary)) == NULL && errno == EEXIST)
+ {
+ xfree (uname);
+ uname = unique_name (name, false);
+ }
+ if (opened_name && fp != NULL)
+ {
+ if (fp)
+ *opened_name = uname;
+ else
+ {
+ *opened_name = NULL;
+ xfree (uname);
+ }
+ }
+ else
+ xfree (uname);
+ return fp;
+}
+
+/* Open the file for writing, with the addition that the file is
+ opened "exclusively". This means that, if the file already exists,
+ this function will *fail* and errno will be set to EEXIST. If
+ BINARY is set, the file will be opened in binary mode, equivalent
+ to fopen's "wb".
+
+ If opening the file fails for any reason, including the file having
+ previously existed, this function returns NULL and sets errno
+ appropriately. */
+
+FILE *
+fopen_excl (const char *fname, bool binary)
+{
+ int fd;
+#ifdef O_EXCL
+ int flags = O_WRONLY | O_CREAT | O_EXCL;
+# ifdef O_BINARY
+ if (binary)
+ flags |= O_BINARY;
+# endif
+ fd = open (fname, flags, 0666);
+ if (fd < 0)
+ return NULL;
+ return fdopen (fd, binary ? "wb" : "w");
+#else /* not O_EXCL */
+ /* Manually check whether the file exists. This is prone to race
+ conditions, but systems without O_EXCL haven't deserved
+ better. */
+ if (file_exists_p (fname))
+ {
+ errno = EEXIST;
+ return NULL;
+ }
+ return fopen (fname, binary ? "wb" : "w");
+#endif /* not O_EXCL */
+}
\f
/* Create DIRECTORY. If some of the pathname components of DIRECTORY
are missing, create them first. In case any mkdir() call fails,
int
make_directory (const char *directory)
{
- int quit = 0;
- int i;
- int ret = 0;
+ int i, ret, quit = 0;
char *dir;
/* Make a copy of dir, to be able to write to it. Otherwise, the
file_merge("/foo/bar/", "baz") => "/foo/bar/baz"
file_merge("foo", "bar") => "bar"
- In other words, it's a simpler and gentler version of uri_merge_1. */
+ In other words, it's a simpler and gentler version of uri_merge. */
char *
file_merge (const char *base, const char *file)
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));
+/* Like fnmatch, but performs a case-insensitive match. */
+
+int
+fnmatch_nocase (const char *pattern, const char *string, int flags)
+{
+#ifdef FNM_CASEFOLD
+ /* The FNM_CASEFOLD flag started as a GNU extension, but it is now
+ also present on *BSD platforms, and possibly elsewhere. */
+ return fnmatch (pattern, string, flags | FNM_CASEFOLD);
+#else
+ /* Turn PATTERN and STRING to lower case and call fnmatch on them. */
+ char *patcopy = (char *) alloca (strlen (pattern) + 1);
+ char *strcopy = (char *) alloca (strlen (string) + 1);
+ char *p;
+ for (p = patcopy; *pattern; pattern++, p++)
+ *p = TOLOWER (*pattern);
+ *p = '\0';
+ for (p = strcopy; *string; string++, p++)
+ *p = TOLOWER (*string);
+ *p = '\0';
+ return fnmatch (patcopy, strcopy, flags);
+#endif
+}
+
+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
-frontcmp (const char *s1, const char *s2)
+/* Check if D2 is a subdirectory of D1. E.g. if D1 is `/something', subdir_p()
+ will return true if and only if D2 begins with `/something/' or is exactly
+ '/something'. */
+bool
+subdir_p (const char *d1, const char *d2)
{
- for (; *s1 && *s2 && (*s1 == *s2); ++s1, ++s2);
- return !*s1;
+ if (!opt.ignore_case)
+ for (; *d1 && *d2 && (*d1 == *d2); ++d1, ++d2)
+ ;
+ else
+ for (; *d1 && *d2 && (TOLOWER (*d1) == TOLOWER (*d2)); ++d1, ++d2)
+ ;
+
+ return *d1 == '\0' && (*d2 == '\0' || *d2 == '/');
}
-/* Iterate through STRLIST, and return the first element that matches
- S, through wildcards or front comparison (as appropriate). */
-static char *
-proclist (char **strlist, const char *s, enum accd flags)
+/* Iterate through DIRLIST (which must be NULL-terminated), and return the
+ first element that matches DIR, through wildcards or front comparison (as
+ appropriate). */
+static bool
+dir_matches_p (char **dirlist, const char *dir)
{
char **x;
+ int (*matcher) (const char *, const char *, int)
+ = opt.ignore_case ? fnmatch_nocase : fnmatch;
- 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;
- }
- return *x;
+ for (x = dirlist; *x; x++)
+ {
+ /* Remove leading '/' */
+ char *p = *x + (**x == '/');
+ if (has_wildcards_p (p))
+ {
+ if (matcher (p, dir, FNM_PATHNAME) == 0)
+ break;
+ }
+ else
+ {
+ if (subdir_p (p, dir))
+ break;
+ }
+ }
+
+ return *x ? true : false;
}
/* Returns whether DIRECTORY is acceptable for download, wrt the
include/exclude lists.
- If FLAGS is ALLABS, the leading `/' is ignored in paths; relative
- and absolute paths may be freely intermixed. */
-int
-accdir (const char *directory, enum accd flags)
+ The leading `/' is ignored in paths; relative and absolute paths
+ may be freely intermixed. */
+
+bool
+accdir (const char *directory)
{
/* Remove starting '/'. */
- if (flags & ALLABS && *directory == '/')
+ if (*directory == '/')
++directory;
if (opt.includes)
{
- if (!proclist (opt.includes, directory, flags))
- return 0;
+ if (!dir_matches_p (opt.includes, directory))
+ return false;
}
if (opt.excludes)
{
- if (proclist (opt.excludes, directory, flags))
- return 0;
+ if (dir_matches_p (opt.excludes, directory))
+ 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++)
{
if (has_wildcards_p (*accepts))
{
- /* fnmatch returns 0 if the pattern *does* match the
- string. */
- if (fnmatch (*accepts, s, 0) == 0)
- return 1;
+ int res = opt.ignore_case
+ ? fnmatch_nocase (*accepts, s, 0) : fnmatch (*accepts, s, 0);
+ /* fnmatch returns 0 if the pattern *does* match the string. */
+ if (res == 0)
+ return true;
}
else
{
if (backward)
{
- if (match_tail (s, *accepts, 0))
- return 1;
+ if (match_tail (s, *accepts, opt.ignore_case))
+ return true;
}
else
{
- if (!strcmp (s, *accepts))
- return 1;
+ int cmp = opt.ignore_case
+ ? strcasecmp (s, *accepts) : strcmp (s, *accepts);
+ if (cmp == 0)
+ return true;
}
}
}
- return 0;
+ return false;
}
/* Return the location of STR's suffix (file extension). Examples:
return NULL;
}
-/* 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 S contains globbing wildcards (`*', `?', `[' or
+ `]'). */
+
+bool
+has_wildcards_p (const char *s)
+{
+ for (; *s; s++)
+ if (*s == '*' || *s == '?' || *s == '[' || *s == ']')
+ return true;
+ return false;
+}
+
+/* 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
- storage. The stoarage space is obtained through malloc() and
- should be freed with free() when it is no longer needed.
+ storage. The storage space is obtained through malloc() and should
+ be freed with free() when it is no longer needed.
The length of the line is not limited, except by available memory.
The newline character at the end of line is retained. The line is
{
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. */
}
fd = open (file, O_RDONLY);
if (fd < 0)
return NULL;
- fm = xmalloc (sizeof (struct file_memory));
+ fm = xnew (struct file_memory);
#ifdef HAVE_MMAP
{
- struct stat buf;
+ struct_fstat buf;
if (fstat (fd, &buf) < 0)
goto mmap_lose;
fm->length = buf.st_size;
fm->content = xmalloc (size);
while (1)
{
- long nread;
+ wgint nread;
if (fm->length > size / 2)
{
/* #### I'm not sure whether the whole exponential-growth
/* Normally, we grow SIZE exponentially to make the number
of calls to read() and realloc() logarithmic in relation
to file size. However, read() can read an amount of data
- smaller than requested, and it would be unreasonably to
+ smaller than requested, and it would be unreasonable to
double SIZE every time *something* was read. Therefore,
we double SIZE only when the length exceeds half of the
entire allocated size. */
return v1;
}
/* Count v1. */
- for (i = 0; v1[i]; i++);
+ for (i = 0; v1[i]; i++)
+ ;
/* Count v2. */
- for (j = 0; v2[j]; j++);
+ 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;
}
-/* A set of simple-minded routines to store strings in a linked list.
- This used to also be used for searching, but now we have hash
- tables for that. */
-
-/* It's a shame that these simple things like linked lists and hash
- tables (see hash.c) need to be implemented over and over again. It
- would be nice to be able to use the routines from glib -- see
- www.gtk.org for details. However, that would make Wget depend on
- glib, and I want to avoid dependencies to external libraries for
- reasons of convenience and portability (I suspect Wget is more
- portable than anything ever written for Gnome). */
+/* 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. */
-/* Append an element to the list. If the list has a huge number of
- elements, this can get slow because it has to find the list's
- ending. If you think you have to call slist_append in a loop,
- think about calling slist_prepend() followed by slist_nreverse(). */
-
-slist *
-slist_append (slist *l, const char *s)
-{
- slist *newel = (slist *)xmalloc (sizeof (slist));
- slist *beg = l;
-
- newel->string = xstrdup (s);
- newel->next = NULL;
-
- if (!l)
- return newel;
- /* Find the last element. */
- while (l->next)
- l = l->next;
- l->next = newel;
- return beg;
-}
-
-/* Prepend S to the list. Unlike slist_append(), this is O(1). */
-
-slist *
-slist_prepend (slist *l, const char *s)
-{
- slist *newel = (slist *)xmalloc (sizeof (slist));
- newel->string = xstrdup (s);
- newel->next = l;
- return newel;
-}
-
-/* Destructively reverse L. */
-
-slist *
-slist_nreverse (slist *l)
-{
- slist *prev = NULL;
- while (l)
- {
- slist *next = l->next;
- l->next = prev;
- prev = l;
- l = next;
- }
- return prev;
-}
-
-/* Is there a specific entry in the list? */
-int
-slist_contains (slist *l, const char *s)
-{
- for (; l; l = l->next)
- if (!strcmp (l->string, s))
- return 1;
- return 0;
-}
-
-/* Free the whole slist. */
-void
-slist_free (slist *l)
+char **
+vec_append (char **vec, const char *str)
{
- while (l)
+ int cnt; /* count of vector elements, including
+ the one we're about to append */
+ if (vec != NULL)
{
- slist *n = l->next;
- xfree (l->string);
- xfree (l);
- l = n;
+ 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
return hash_table_contains (ht, s);
}
-static int
-string_set_free_mapper (void *key, void *value_ignored, void *arg_ignored)
+/* Convert the specified string set to array. ARRAY should be large
+ enough to hold hash_table_count(ht) char pointers. */
+
+void string_set_to_array (struct hash_table *ht, char **array)
{
- xfree (key);
- return 0;
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ *array++ = iter.key;
}
+/* Free the string set. This frees both the storage allocated for
+ keys and the actual hash table. (hash_table_destroy would only
+ destroy the hash table.) */
+
void
string_set_free (struct hash_table *ht)
{
- hash_table_map (ht, string_set_free_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ xfree (iter.key);
hash_table_destroy (ht);
}
-static int
-free_keys_and_values_mapper (void *key, void *value, void *arg_ignored)
-{
- xfree (key);
- xfree (value);
- return 0;
-}
-
-/* Another utility function: call free() on all keys and values of HT. */
+/* Utility function: simply call xfree() on all keys and values of HT. */
void
free_keys_and_values (struct hash_table *ht)
{
- hash_table_map (ht, free_keys_and_values_mapper, NULL);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ {
+ xfree (iter.key);
+ xfree (iter.value);
+ }
}
-
\f
-/* Engine for legible and legible_very_long; this function works on
- strings. */
+/* Get digit grouping data for thousand separors by calling
+ localeconv(). The data includes separator string and grouping info
+ and is cached after the first call to the function.
-static char *
-legible_1 (const char *repr)
-{
- static char outbuf[128];
- int i, i1, mod;
- char *outptr;
- const char *inptr;
-
- /* Reset the pointers. */
- outptr = outbuf;
- inptr = repr;
- /* If the number is negative, shift the pointers. */
- 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++)
+ In locales that don't set a thousand separator (such as the "C"
+ locale), this forces it to be ",". We are now only showing
+ thousand separators in one place, so this shouldn't be a problem in
+ practice. */
+
+static void
+get_grouping_data (const char **sep, const char **grouping)
+{
+ static const char *cached_sep;
+ static const char *cached_grouping;
+ static bool initialized;
+ if (!initialized)
{
- if (i % 3 == 0 && i1 != 0)
- *outptr++ = ',';
- *outptr++ = inptr[i1];
+ /* Get the grouping info from the locale. */
+ struct lconv *lconv = localeconv ();
+ cached_sep = lconv->thousands_sep;
+ cached_grouping = lconv->grouping;
+ if (!*cached_sep)
+ {
+ /* Many locales (such as "C" or "hr_HR") don't specify
+ grouping, which we still want to use it for legibility.
+ In those locales set the sep char to ',', unless that
+ character is used for decimal point, in which case set it
+ to ".". */
+ if (*lconv->decimal_point != ',')
+ cached_sep = ",";
+ else
+ cached_sep = ".";
+ cached_grouping = "\x03";
+ }
+ initialized = true;
}
- /* Zero-terminate the string. */
- *outptr = '\0';
- return outbuf;
+ *sep = cached_sep;
+ *grouping = cached_grouping;
}
-/* Legible -- return a static pointer to the legibly printed long. */
-char *
-legible (long l)
+/* 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.
+
+ 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 display
+ separators in the "C" locale, still used by many Unix users. */
+
+const char *
+with_thousand_seps (wgint n)
{
- char inbuf[24];
- /* Print the number into the buffer. */
- number_to_string (inbuf, l);
- return legible_1 (inbuf);
-}
+ static char outbuf[48];
+ char *p = outbuf + sizeof outbuf;
-/* Write a string representation of NUMBER into the provided buffer.
- We cannot use sprintf() because we cannot be sure whether the
- platform supports printing of what we chose for VERY_LONG_TYPE.
+ /* Info received from locale */
+ const char *grouping, *sep;
+ int seplen;
- Example: Gcc supports `long long' under many platforms, but on many
- of those the native libc knows nothing of it and therefore cannot
- print it.
+ /* State information */
+ int i = 0, groupsize;
+ const char *atgroup;
- How long BUFFER needs to be depends on the platform and the content
- of NUMBER. For 64-bit VERY_LONG_TYPE (the most common case), 24
- bytes are sufficient. Using more might be a good idea.
+ bool negative = n < 0;
- This function does not go through the hoops that long_to_string
- goes to because it doesn't aspire to be fast. (It's called perhaps
- once in a Wget run.) */
+ /* Initialize grouping data. */
+ get_grouping_data (&sep, &grouping);
+ seplen = strlen (sep);
+ atgroup = grouping;
+ groupsize = *atgroup++;
-static void
-very_long_to_string (char *buffer, VERY_LONG_TYPE number)
-{
- int i = 0;
- int j;
+ /* This would overflow on WGINT_MIN, but printing negative numbers
+ is not an important goal of this fuinction. */
+ if (negative)
+ n = -n;
- /* Print the number backwards... */
- do
+ /* Write the number into the buffer, backwards, inserting the
+ separators as necessary. */
+ *--p = '\0';
+ while (1)
{
- buffer[i++] = '0' + number % 10;
- number /= 10;
+ *--p = n % 10 + '0';
+ n /= 10;
+ if (n == 0)
+ break;
+ /* Prepend SEP to every groupsize'd digit and get new groupsize. */
+ if (++i == groupsize)
+ {
+ if (seplen == 1)
+ *--p = *sep;
+ else
+ memcpy (p -= seplen, sep, seplen);
+ i = 0;
+ if (*atgroup)
+ groupsize = *atgroup++;
+ }
}
- while (number);
+ if (negative)
+ *--p = '-';
- /* ...and reverse the order of the digits. */
- for (j = 0; j < i / 2; j++)
- {
- char c = buffer[j];
- buffer[j] = buffer[i - 1 - j];
- buffer[i - 1 - j] = c;
- }
- buffer[i] = '\0';
+ return p;
}
-/* The same as legible(), but works on VERY_LONG_TYPE. See sysdep.h. */
+/* N, a byte quantity, is converted to a human-readable abberviated
+ form a la sizes printed by `ls -lh'. The result is written to a
+ static buffer, a pointer to which is returned.
+
+ Unlike `with_thousand_seps', this approximates to the nearest unit.
+ Quoting GNU libit: "Most people visually process strings of 3-4
+ digits effectively, but longer strings of digits are more prone to
+ misinterpretation. Hence, converting to an abbreviated form
+ usually improves readability."
+
+ This intentionally uses kilobyte (KB), megabyte (MB), etc. in their
+ original computer-related meaning of "powers of 1024". We don't
+ use the "*bibyte" names invented in 1998, and seldom used in
+ practice. Wikipedia's entry on "binary prefix" discusses this in
+ some detail. */
+
char *
-legible_very_long (VERY_LONG_TYPE l)
+human_readable (HR_NUMTYPE n)
{
- char inbuf[128];
- /* Print the number into the buffer. */
- very_long_to_string (inbuf, l);
- return legible_1 (inbuf);
+ /* These suffixes are compatible with those of GNU `ls -lh'. */
+ static char powers[] =
+ {
+ 'K', /* kilobyte, 2^10 bytes */
+ 'M', /* megabyte, 2^20 bytes */
+ 'G', /* gigabyte, 2^30 bytes */
+ 'T', /* terabyte, 2^40 bytes */
+ 'P', /* petabyte, 2^50 bytes */
+ 'E', /* exabyte, 2^60 bytes */
+ };
+ static char buf[8];
+ int i;
+
+ /* If the quantity is smaller than 1K, just print it. */
+ if (n < 1024)
+ {
+ snprintf (buf, sizeof (buf), "%d", (int) n);
+ return buf;
+ }
+
+ /* Loop over powers, dividing N with 1024 in each iteration. This
+ works unchanged for all sizes of wgint, while still avoiding
+ non-portable `long double' arithmetic. */
+ for (i = 0; i < countof (powers); i++)
+ {
+ /* 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 / 1024) < 1024 || i == countof (powers) - 1)
+ {
+ double val = n / 1024.0;
+ /* Print values smaller than 10 with one decimal digits, and
+ others without any decimals. */
+ snprintf (buf, sizeof (buf), "%.*f%c",
+ val < 10 ? 1 : 0, val, powers[i]);
+ return buf;
+ }
+ n /= 1024;
+ }
+ return NULL; /* unreached */
}
-/* Count the digits in a (long) integer. */
+/* Count the digits in the provided number. Used to allocate space
+ when printing numbers. */
+
int
-numdigit (long number)
+numdigit (wgint number)
{
int cnt = 1;
if (number < 0)
- {
- number = -number;
- ++cnt;
- }
- while ((number /= 10) > 0)
+ ++cnt; /* accomodate '-' */
+ while ((number /= 10) != 0)
++cnt;
return cnt;
}
-/* A half-assed implementation of INT_MAX on machines that don't
- bother to define one. */
-#ifndef INT_MAX
-# define INT_MAX ((int) ~((unsigned)1 << 8 * sizeof (int) - 1))
-#endif
+#define PR(mask) *p++ = n / (mask) + '0'
+
+/* DIGITS_<D> is used to print a D-digit number and should be called
+ with mask==10^(D-1). It prints n/mask (the first digit), reducing
+ n to n%mask (the remaining digits), and calling DIGITS_<D-1>.
+ Recursively this continues until DIGITS_1 is invoked. */
+
+#define DIGITS_1(mask) PR (mask)
+#define DIGITS_2(mask) PR (mask), n %= (mask), DIGITS_1 ((mask) / 10)
+#define DIGITS_3(mask) PR (mask), n %= (mask), DIGITS_2 ((mask) / 10)
+#define DIGITS_4(mask) PR (mask), n %= (mask), DIGITS_3 ((mask) / 10)
+#define DIGITS_5(mask) PR (mask), n %= (mask), DIGITS_4 ((mask) / 10)
+#define DIGITS_6(mask) PR (mask), n %= (mask), DIGITS_5 ((mask) / 10)
+#define DIGITS_7(mask) PR (mask), n %= (mask), DIGITS_6 ((mask) / 10)
+#define DIGITS_8(mask) PR (mask), n %= (mask), DIGITS_7 ((mask) / 10)
+#define DIGITS_9(mask) PR (mask), n %= (mask), DIGITS_8 ((mask) / 10)
+#define DIGITS_10(mask) PR (mask), n %= (mask), DIGITS_9 ((mask) / 10)
+
+/* DIGITS_<11-20> are only used on machines with 64-bit wgints. */
-#define ONE_DIGIT(figure) *p++ = n / (figure) + '0'
-#define ONE_DIGIT_ADVANCE(figure) (ONE_DIGIT (figure), n %= (figure))
-
-#define DIGITS_1(figure) ONE_DIGIT (figure)
-#define DIGITS_2(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_1 ((figure) / 10)
-#define DIGITS_3(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_2 ((figure) / 10)
-#define DIGITS_4(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_3 ((figure) / 10)
-#define DIGITS_5(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_4 ((figure) / 10)
-#define DIGITS_6(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_5 ((figure) / 10)
-#define DIGITS_7(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_6 ((figure) / 10)
-#define DIGITS_8(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_7 ((figure) / 10)
-#define DIGITS_9(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_8 ((figure) / 10)
-#define DIGITS_10(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_9 ((figure) / 10)
-
-/* DIGITS_<11-20> are only used on machines with 64-bit longs. */
-
-#define DIGITS_11(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_10 ((figure) / 10)
-#define DIGITS_12(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_11 ((figure) / 10)
-#define DIGITS_13(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_12 ((figure) / 10)
-#define DIGITS_14(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_13 ((figure) / 10)
-#define DIGITS_15(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_14 ((figure) / 10)
-#define DIGITS_16(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_15 ((figure) / 10)
-#define DIGITS_17(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_16 ((figure) / 10)
-#define DIGITS_18(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_17 ((figure) / 10)
-#define DIGITS_19(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_18 ((figure) / 10)
-
-/* Print NUMBER to BUFFER in base 10. This should be completely
- equivalent to `sprintf(buffer, "%ld", number)', only much faster.
+#define DIGITS_11(mask) PR (mask), n %= (mask), DIGITS_10 ((mask) / 10)
+#define DIGITS_12(mask) PR (mask), n %= (mask), DIGITS_11 ((mask) / 10)
+#define DIGITS_13(mask) PR (mask), n %= (mask), DIGITS_12 ((mask) / 10)
+#define DIGITS_14(mask) PR (mask), n %= (mask), DIGITS_13 ((mask) / 10)
+#define DIGITS_15(mask) PR (mask), n %= (mask), DIGITS_14 ((mask) / 10)
+#define DIGITS_16(mask) PR (mask), n %= (mask), DIGITS_15 ((mask) / 10)
+#define DIGITS_17(mask) PR (mask), n %= (mask), DIGITS_16 ((mask) / 10)
+#define DIGITS_18(mask) PR (mask), n %= (mask), DIGITS_17 ((mask) / 10)
+#define DIGITS_19(mask) PR (mask), n %= (mask), DIGITS_18 ((mask) / 10)
+
+/* Shorthand for casting to wgint. */
+#define W wgint
+
+/* Print NUMBER to BUFFER in base 10. This is equivalent to
+ `sprintf(buffer, "%lld", (long long) number)', only typically much
+ faster and portable to machines without long long.
The speedup may make a difference in programs that frequently
convert numbers to strings. Some implementations of sprintf,
- particularly the one in GNU libc, have been known to be extremely
- slow compared to this function.
+ particularly the one in some versions of GNU libc, have been known
+ to be quite slow when converting integers to strings.
Return the pointer to the location where the terminating zero was
printed. (Equivalent to calling buffer+strlen(buffer) after the
function is done.)
- BUFFER should be big enough to accept as many bytes as you expect
- the number to take up. On machines with 64-bit longs the maximum
+ BUFFER should be large enough to accept as many bytes as you expect
+ the number to take up. On machines with 64-bit wgints the maximum
needed size is 24 bytes. That includes the digits needed for the
largest 64-bit number, the `-' sign in case it's negative, and the
terminating '\0'. */
char *
-number_to_string (char *buffer, long number)
+number_to_string (char *buffer, wgint number)
{
char *p = buffer;
- long n = number;
+ wgint n = number;
-#if (SIZEOF_LONG != 4) && (SIZEOF_LONG != 8)
- /* We are running in a strange or misconfigured environment. Let
- sprintf cope with it. */
- sprintf (buffer, "%ld", n);
- p += strlen (buffer);
-#else /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */
+ int last_digit_char = 0;
+
+#if (SIZEOF_WGINT != 4) && (SIZEOF_WGINT != 8)
+ /* We are running in a very strange environment. Leave the correct
+ printing to sprintf. */
+ p += sprintf (buf, "%j", (intmax_t) (n));
+#else /* (SIZEOF_WGINT == 4) || (SIZEOF_WGINT == 8) */
if (n < 0)
{
- if (n < -INT_MAX)
+ if (n < -WGINT_MAX)
{
- /* We cannot print a '-' and assign -n to n because -n would
- overflow. Let sprintf deal with this border case. */
- sprintf (buffer, "%ld", n);
- p += strlen (buffer);
- return p;
+ /* n = -n would overflow because -n would evaluate to a
+ wgint value larger than WGINT_MAX. Need to make n
+ smaller and handle the last digit separately. */
+ int last_digit = n % 10;
+ /* The sign of n%10 is implementation-defined. */
+ if (last_digit < 0)
+ last_digit_char = '0' - last_digit;
+ else
+ last_digit_char = '0' + last_digit;
+ /* After n is made smaller, -n will not overflow. */
+ n /= 10;
}
*p++ = '-';
n = -n;
}
- if (n < 10) { DIGITS_1 (1); }
- else if (n < 100) { DIGITS_2 (10); }
- else if (n < 1000) { DIGITS_3 (100); }
- else if (n < 10000) { DIGITS_4 (1000); }
- else if (n < 100000) { DIGITS_5 (10000); }
- else if (n < 1000000) { DIGITS_6 (100000); }
- else if (n < 10000000) { DIGITS_7 (1000000); }
- else if (n < 100000000) { DIGITS_8 (10000000); }
- else if (n < 1000000000) { DIGITS_9 (100000000); }
-#if SIZEOF_LONG == 4
- /* ``if (1)'' serves only to preserve editor indentation. */
- else if (1) { DIGITS_10 (1000000000); }
-#else /* SIZEOF_LONG != 4 */
- else if (n < 10000000000L) { DIGITS_10 (1000000000L); }
- else if (n < 100000000000L) { DIGITS_11 (10000000000L); }
- else if (n < 1000000000000L) { DIGITS_12 (100000000000L); }
- else if (n < 10000000000000L) { DIGITS_13 (1000000000000L); }
- else if (n < 100000000000000L) { DIGITS_14 (10000000000000L); }
- else if (n < 1000000000000000L) { DIGITS_15 (100000000000000L); }
- else if (n < 10000000000000000L) { DIGITS_16 (1000000000000000L); }
- else if (n < 100000000000000000L) { DIGITS_17 (10000000000000000L); }
- else if (n < 1000000000000000000L) { DIGITS_18 (100000000000000000L); }
- else { DIGITS_19 (1000000000000000000L); }
-#endif /* SIZEOF_LONG != 4 */
+ /* Use the DIGITS_ macro appropriate for N's number of digits. That
+ way printing any N is fully open-coded without a loop or jump.
+ (Also see description of DIGITS_*.) */
+
+ if (n < 10) DIGITS_1 (1);
+ else if (n < 100) DIGITS_2 (10);
+ else if (n < 1000) DIGITS_3 (100);
+ else if (n < 10000) DIGITS_4 (1000);
+ else if (n < 100000) DIGITS_5 (10000);
+ else if (n < 1000000) DIGITS_6 (100000);
+ else if (n < 10000000) DIGITS_7 (1000000);
+ else if (n < 100000000) DIGITS_8 (10000000);
+ else if (n < 1000000000) DIGITS_9 (100000000);
+#if SIZEOF_WGINT == 4
+ /* 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 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);
+ else if (n < 10000*(W)1000000000) DIGITS_13 (1000*(W)1000000000);
+ else if (n < 100000*(W)1000000000) DIGITS_14 (10000*(W)1000000000);
+ else if (n < 1000000*(W)1000000000) DIGITS_15 (100000*(W)1000000000);
+ else if (n < 10000000*(W)1000000000) DIGITS_16 (1000000*(W)1000000000);
+ else if (n < 100000000*(W)1000000000) DIGITS_17 (10000000*(W)1000000000);
+ else if (n < 1000000000*(W)1000000000) DIGITS_18 (100000000*(W)1000000000);
+ else DIGITS_19 (1000000000*(W)1000000000);
+#endif
+
+ if (last_digit_char)
+ *p++ = last_digit_char;
*p = '\0';
-#endif /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */
+#endif /* (SIZEOF_WGINT == 4) || (SIZEOF_WGINT == 8) */
return p;
}
-#undef ONE_DIGIT
-#undef ONE_DIGIT_ADVANCE
-
+#undef PR
+#undef W
+#undef SPRINTF_WGINT
#undef DIGITS_1
#undef DIGITS_2
#undef DIGITS_3
#undef DIGITS_17
#undef DIGITS_18
#undef DIGITS_19
-\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.
-
- #### Perhaps we should also support ftime(), which exists on old
- BSD 4.2-influenced systems? (It also existed under MS DOS Borland
- C, if memory serves me.) */
-
-#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 ULARGE_INTEGER wget_sys_time;
-#endif
-
-struct wget_timer {
- /* 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_elapsed().
- 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;
-};
-/* Allocate a timer. It is not legal to do anything with a freshly
- allocated timer, except call wtimer_reset() or wtimer_delete(). */
+#define RING_SIZE 3
+
+/* Print NUMBER to a statically allocated string and return a pointer
+ to the printed representation.
+
+ This function is intended to be used in conjunction with printf.
+ It is hard to portably print wgint values:
+ a) you cannot use printf("%ld", number) because wgint can be long
+ long on 32-bit machines with LFS.
+ b) you cannot use printf("%lld", number) because NUMBER could be
+ long on 32-bit machines without LFS, or on 64-bit machines,
+ which do not require LFS. Also, Windows doesn't support %lld.
+ c) you cannot use printf("%j", (int_max_t) number) because not all
+ versions of printf support "%j", the most notable being the one
+ on Windows.
+ d) you cannot #define WGINT_FMT to the appropriate format and use
+ printf(WGINT_FMT, number) because that would break translations
+ for user-visible messages, such as printf("Downloaded: %d
+ bytes\n", number).
+
+ What you should use instead is printf("%s", number_to_static_string
+ (number)).
+
+ CAVEAT: since the function returns pointers to static data, you
+ must be careful to copy its result before calling it again.
+ However, to make it more useful with printf, the function maintains
+ an internal ring of static buffers to return. That way things like
+ printf("%s %s", number_to_static_string (num1),
+ number_to_static_string (num2)) work as expected. Three buffers
+ 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.) */
-struct wget_timer *
-wtimer_allocate (void)
-{
- struct wget_timer *wt =
- (struct wget_timer *)xmalloc (sizeof (struct wget_timer));
- 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
- /* We use GetSystemTime to get the elapsed time. MSDN warns that
- system clock adjustments can skew the output of GetSystemTime
- when used as a timer and gives preference to GetTickCount and
- high-resolution timers. But GetTickCount can overflow, and hires
- timers are typically used for profiling, not for regular time
- measurement. Since we handle clock skew anyway, we just use
- GetSystemTime. */
- FILETIME ft;
- SYSTEMTIME st;
- GetSystemTime (&st);
-
- /* As recommended by MSDN, we convert SYSTEMTIME to FILETIME, copy
- FILETIME to ULARGE_INTEGER, and use regular 64-bit integer
- arithmetic on that. */
- SystemTimeToFileTime (&st, &ft);
- wst->HighPart = ft.dwHighDateTime;
- wst->LowPart = ft.dwLowDateTime;
-#endif
-}
-
-/* Reset timer WT. This establishes the starting point from which
- wtimer_elapsed() 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;
-}
-
-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
- /* VC++ 6 doesn't support direct cast of uint64 to double. To work
- around this, we subtract, then convert to signed, then finally to
- double. */
- return (double)(signed __int64)(wst1->QuadPart - wst2->QuadPart) / 10000;
-#endif
-}
-
-/* Return the number of milliseconds elapsed since the timer was last
- reset. It is allowed to call this function more than once to get
- increasingly higher elapsed values. These timers handle clock
- skew. */
-
-double
-wtimer_elapsed (struct wget_timer *wt)
-{
- wget_sys_time now;
- double elapsed;
-
- 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 elapsed;
-}
-
-/* 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
- /* According to MSDN, GetSystemTime returns a broken-down time
- structure the smallest member of which are milliseconds. */
- return 1;
-#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)
+number_to_static_string (wgint number)
{
- 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;
+ 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;
}
-
+\f
/* Determine the width of the terminal we're running on. If that's
not possible, return 0. */
{
/* If there's a way to get the terminal size using POSIX
tcgetattr(), somebody please tell me. */
-#ifndef TIOCGWINSZ
- return 0;
-#else /* TIOCGWINSZ */
+#ifdef TIOCGWINSZ
int fd;
struct winsize wsz;
return 0; /* most likely ENOTTY */
return wsz.ws_col;
-#endif /* TIOCGWINSZ */
+#elif defined(WINDOWS)
+ CONSOLE_SCREEN_BUFFER_INFO csbi;
+ if (!GetConsoleScreenBufferInfo (GetStdHandle (STD_ERROR_HANDLE), &csbi))
+ return 0;
+ return csbi.dwSize.X;
+#else /* neither TIOCGWINSZ nor WINDOWS */
+ return 0;
+#endif /* neither TIOCGWINSZ nor WINDOWS */
}
+\f
+/* Whether the rnd system (either rand or [dl]rand48) has been
+ seeded. */
+static int rnd_seeded;
/* 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.
-
+ If the system does not support lrand48 and 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. */
+ This uses lrand48 where available, rand elsewhere. DO NOT use it
+ for cryptography. 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;
+#ifdef HAVE_DRAND48
+ if (!rnd_seeded)
+ {
+ srand48 ((long) time (NULL) ^ (long) getpid ());
+ rnd_seeded = 1;
+ }
+ return lrand48 () % max;
+#else /* not HAVE_DRAND48 */
+
double bounded;
int rnd;
-
- if (!seeded)
+ if (!rnd_seeded)
{
- srand (time (NULL));
- seeded = 1;
+ srand ((unsigned) time (NULL) ^ (unsigned) getpid ());
+ rnd_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
+ /* Like rand() % max, but uses the high-order bits for better
+ randomness on architectures where rand() is implemented using a
+ simple congruential generator. */
- /* 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;
- bounded = (double)max * rnd / (RAND_MAX + 1.0);
- return (int)bounded;
+#endif /* not HAVE_DRAND48 */
}
/* 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! */
+ [0, 1) range. Uses drand48 where available, and a really lame
+ kludge elsewhere. */
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;
-}
-
-#if 0
-/* A debugging function for checking whether an MD5 library works. */
-
-#include "gen-md5.h"
-
-char *
-debug_test_md5 (char *buf)
-{
- unsigned char raw[16];
- static char res[33];
- unsigned char *p1;
- char *p2;
- int cnt;
- ALLOCA_MD5_CONTEXT (ctx);
-
- gen_md5_init (ctx);
- gen_md5_update ((unsigned char *)buf, strlen (buf), ctx);
- gen_md5_finish (ctx, raw);
-
- p1 = raw;
- p2 = res;
- cnt = 16;
- while (cnt--)
+#ifdef HAVE_DRAND48
+ if (!rnd_seeded)
{
- *p2++ = XNUM_TO_digit (*p1 >> 4);
- *p2++ = XNUM_TO_digit (*p1 & 0xf);
- ++p1;
+ srand48 ((long) time (NULL) ^ (long) getpid ());
+ rnd_seeded = 1;
}
- *p2 = '\0';
-
- return res;
+ return drand48 ();
+#else /* not HAVE_DRAND48 */
+ return ( random_number (10000) / 10000.0
+ + random_number (10000) / (10000.0 * 10000.0)
+ + random_number (10000) / (10000.0 * 10000.0 * 10000.0)
+ + random_number (10000) / (10000.0 * 10000.0 * 10000.0 * 10000.0));
+#endif /* not HAVE_DRAND48 */
}
-#endif
\f
/* Implementation of run_with_timeout, a generic timeout-forcing
routine for systems with Unix-like signal handling. */
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);
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);
+ mask &= ~sigmask (SIGALRM);
sigsetmask (mask);
/* Now it's safe to longjump. */
#ifdef ITIMER_REAL
/* Use the modern itimer interface. */
struct itimerval itv;
- memset (&itv, 0, sizeof (itv));
+ xzero (itv);
itv.it_value.tv_sec = (long) timeout;
- itv.it_value.tv_usec = 1000000L * (timeout - (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". */
{
#ifdef ITIMER_REAL
struct itimerval disable;
- memset (&disable, 0, sizeof (disable));
+ xzero (disable);
setitimer (ITIMER_REAL, &disable, NULL);
#else /* not ITIMER_REAL */
alarm (0);
#endif /* not ITIMER_REAL */
}
-/* Run FUN(ARG) for not more than TIMEOUT seconds. Returns non-zero
- if the function was interrupted with a timeout, zero otherwise.
+/* Call FUN(ARG), but don't allow it to run for more than TIMEOUT
+ 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
* It works with both SYSV and BSD signals because it doesn't
depend on the default setting of SA_RESTART.
- * It doesn't special handler setup beyond a simple call to
- signal(). (It does use sigsetjmp/siglongjmp, but they're
+ * 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
+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 */
+\f
+#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;
+#elif defined(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 /* 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);
+ 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. */
+#endif
+}
+
+#endif /* not WINDOWS */
+
+/* Encode the octets in DATA of length LENGTH to base64 format,
+ storing the result to DEST. The output will be zero-terminated,
+ and must point to a writable buffer of at least
+ 1+BASE64_LENGTH(length) bytes. The function returns the length of
+ the resulting base64 data, not counting the terminating zero.
+
+ This implementation does not emit newlines after 76 characters of
+ base64 data. */
+
+int
+base64_encode (const void *data, int length, char *dest)
+{
+ /* Conversion table. */
+ static const 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','+','/'
+ };
+ /* Access bytes in DATA as unsigned char, otherwise the shifts below
+ don't work for data with MSB set. */
+ const unsigned char *s = data;
+ /* Theoretical ANSI violation when length < 3. */
+ const unsigned char *end = (const unsigned char *) data + length - 2;
+ char *p = dest;
+
+ /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
+ for (; s < end; s += 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];
+ }
+
+ /* Pad the result if necessary... */
+ switch (length % 3)
+ {
+ case 1:
+ *p++ = tbl[s[0] >> 2];
+ *p++ = tbl[(s[0] & 3) << 4];
+ *p++ = '=';
+ *p++ = '=';
+ break;
+ case 2:
+ *p++ = tbl[s[0] >> 2];
+ *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
+ *p++ = tbl[((s[1] & 0xf) << 2)];
+ *p++ = '=';
+ break;
+ }
+ /* ...and zero-terminate it. */
+ *p = '\0';
+
+ return p - dest;
+}
+
+/* Store in C the next non-whitespace character from the string, or \0
+ when end of string is reached. */
+#define NEXT_CHAR(c, p) do { \
+ c = (unsigned char) *p++; \
+} while (ISSPACE (c))
+
+#define IS_ASCII(c) (((c) & 0x80) == 0)
+
+/* Decode data from BASE64 (a null-terminated string) into memory
+ pointed to by DEST. DEST is assumed to be large enough to
+ accomodate the decoded data, which is guaranteed to be no more than
+ 3/4*strlen(base64).
+
+ Since DEST 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.
+
+ This function originates from Free Recode. */
+
+int
+base64_decode (const char *base64, void *dest)
+{
+ /* Table of base64 values for first 128 characters. Note that this
+ assumes ASCII (but so does Wget in other places). */
+ static const signed char 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 */
+ };
+#define BASE64_CHAR_TO_VALUE(c) ((int) base64_char_to_value[c])
+#define IS_BASE64(c) ((IS_ASCII (c) && BASE64_CHAR_TO_VALUE (c) >= 0) || c == '=')
+
+ const char *p = base64;
+ char *q = dest;
+
+ while (1)
+ {
+ unsigned char c;
+ unsigned long value;
+
+ /* Process first byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ break;
+ if (c == '=' || !IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+ value = BASE64_CHAR_TO_VALUE (c) << 18;
+
+ /* Process second byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (c == '=' || !IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+ value |= BASE64_CHAR_TO_VALUE (c) << 12;
+ *q++ = value >> 16;
+
+ /* Process third byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (!IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+
+ if (c == '=')
+ {
+ NEXT_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_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (c == '=')
+ continue;
+ if (!IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+
+ value |= BASE64_CHAR_TO_VALUE (c);
+ *q++ = 0xff & value;
+ }
+#undef IS_BASE64
+#undef BASE64_CHAR_TO_VALUE
+
+ return q - (char *) dest;
+}
+
+#undef IS_ASCII
+#undef NEXT_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);
+ }
+}
+\f
+/* Print a decimal number. If it is equal to or larger than ten, the
+ number is rounded. Otherwise it is printed with one significant
+ digit without trailing zeros and with no more than three fractional
+ digits total. For example, 0.1 is printed as "0.1", 0.035 is
+ printed as "0.04", 0.0091 as "0.009", and 0.0003 as simply "0".
+
+ This is useful for displaying durations because it provides
+ order-of-magnitude information without unnecessary clutter --
+ long-running downloads are shown without the fractional part, and
+ short ones still retain one significant digit. */
+
+const char *
+print_decimal (double number)
+{
+ static char buf[32];
+ double n = number >= 0 ? number : -number;
+
+ if (n >= 9.95)
+ /* Cut off at 9.95 because the below %.1f would round 9.96 to
+ "10.0" instead of "10". OTOH 9.94 will print as "9.9". */
+ snprintf (buf, sizeof buf, "%.0f", number);
+ else if (n >= 0.95)
+ snprintf (buf, sizeof buf, "%.1f", number);
+ else if (n >= 0.001)
+ snprintf (buf, sizeof buf, "%.1g", number);
+ else if (n >= 0.0005)
+ /* round [0.0005, 0.001) to 0.001 */
+ snprintf (buf, sizeof buf, "%.3f", number);
+ else
+ /* print numbers close to 0 as 0, not 0.000 */
+ strcpy (buf, "0");
+
+ return buf;
+}
+
+#ifdef TESTING
+
+const char *
+test_subdir_p()
+{
+ int i;
+ struct {
+ char *d1;
+ char *d2;
+ bool result;
+ } test_array[] = {
+ { "/somedir", "/somedir", true },
+ { "/somedir", "/somedir/d2", true },
+ { "/somedir/d1", "/somedir", false },
+ };
+
+ for (i = 0; i < countof(test_array); ++i)
+ {
+ bool res = subdir_p (test_array[i].d1, test_array[i].d2);
+
+ mu_assert ("test_subdir_p: wrong result",
+ res == test_array[i].result);
+ }
+
+ return NULL;
+}
+
+const char *
+test_dir_matches_p()
+{
+ int i;
+ struct {
+ char *dirlist[3];
+ char *dir;
+ bool result;
+ } test_array[] = {
+ { { "/somedir", "/someotherdir", NULL }, "somedir", true },
+ { { "/somedir", "/someotherdir", NULL }, "anotherdir", false },
+ { { "/somedir", "/*otherdir", NULL }, "anotherdir", true },
+ { { "/somedir/d1", "/someotherdir", NULL }, "somedir/d1", true },
+ { { "/somedir/d1", "/someotherdir", NULL }, "d1", false },
+ };
+
+ for (i = 0; i < countof(test_array); ++i)
+ {
+ bool res = dir_matches_p (test_array[i].dirlist, test_array[i].dir);
+
+ mu_assert ("test_dir_matches_p: wrong result",
+ res == test_array[i].result);
+ }
+
+ return NULL;
+}
+
+#endif /* TESTING */
+
projects / wget / blobdiff