-/* Various functions of utilitarian nature.
- Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001
- Free Software Foundation, Inc.
+/* Various utility functions.
+ Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
+ 2005, 2006, 2007, 2008, 2009, 2010, 2011 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/>.
-#include <config.h>
+Additional permission under GNU GPL version 3 section 7
+
+If you modify this program, or any covered work, by linking or
+combining it with the OpenSSL project's OpenSSL library (or a
+modified version of that library), containing parts covered by the
+terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
+grants you additional permission to convey the resulting work.
+Corresponding Source for a non-source form of such a combination
+shall include the source code for the parts of OpenSSL used as well
+as that of the covered work. */
+
+#include "wget.h"
#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>
-#ifdef HAVE_UNISTD_H
-# include <unistd.h>
-#endif
+#include <string.h>
+#include <time.h>
+#include <unistd.h>
#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
-#ifdef HAVE_SYS_UTIME_H
-# include <sys/utime.h>
+#ifdef HAVE_PROCESS_H
+# include <process.h> /* getpid() */
#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 <sys/ioctl.h>
-#endif
-#ifdef HAVE_TERMIOS_H
-# include <termios.h>
-#endif
-
-#include "wget.h"
-#include "utils.h"
-#include "fnmatch.h"
-#include "hash.h"
+#if HAVE_UTIME
+# include <sys/types.h>
+# ifdef HAVE_UTIME_H
+# include <utime.h>
+# endif
-#ifndef errno
-extern int errno;
+# ifdef HAVE_SYS_UTIME_H
+# include <sys/utime.h>
+# endif
#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.
+#include <sys/time.h>
- 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. */
+#include <sys/stat.h>
-/* 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);
-}
+/* For TIOCGWINSZ and friends: */
+#include <sys/ioctl.h>
+#ifdef HAVE_TERMIOS_H
+# include <termios.h>
+#endif
-/* These functions end with _real because they need to be
- distinguished from the debugging functions, and from the macros.
- Explanation follows:
+/* Needed for Unix version of run_with_timeout. */
+#include <signal.h>
+#include <setjmp.h>
- If memory debugging is not turned on, wget.h defines these:
+#include <regex.h>
+#ifdef HAVE_LIBPCRE
+# include <pcre.h>
+#endif
- #define xmalloc xmalloc_real
- #define xrealloc xrealloc_real
- #define xstrdup xstrdup_real
- #define xfree free
+#ifndef HAVE_SIGSETJMP
+/* If sigsetjmp is a macro, configure won't pick it up. */
+# ifdef sigsetjmp
+# define HAVE_SIGSETJMP
+# endif
+#endif
- 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:
+#if defined HAVE_SIGSETJMP || defined HAVE_SIGBLOCK
+# define USE_SIGNAL_TIMEOUT
+#endif
- #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__)
+#include "utils.h"
+#include "hash.h"
- Each of the *_debug function does its magic and calls the real one. */
+#ifdef __VMS
+#include "vms.h"
+#endif /* def __VMS */
-#ifdef DEBUG_MALLOC
-# define STATIC_IF_DEBUG static
-#else
-# define STATIC_IF_DEBUG
+#ifdef TESTING
+#include "test.h"
#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)
+static void
+memfatal (const char *context, long attempted_size)
{
- void *newptr;
+ /* Make sure we don't try to store part of the log line, and thus
+ call malloc. */
+ log_set_save_context (false);
- /* 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);
+ /* We have different log outputs in different situations:
+ 1) output without bytes information
+ 2) output with bytes information */
+ if (attempted_size == UNKNOWN_ATTEMPTED_SIZE)
+ {
+ logprintf (LOG_ALWAYS,
+ _("%s: %s: Failed to allocate enough memory; memory exhausted.\n"),
+ exec_name, context);
+ }
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 */
+ {
+ logprintf (LOG_ALWAYS,
+ _("%s: %s: Failed to allocate %ld bytes; memory exhausted.\n"),
+ exec_name, context, attempted_size);
+ }
- return copy;
+ exit (1);
}
-#ifdef DEBUG_MALLOC
+/* Character property table for (re-)escaping VMS ODS5 extended file
+ names. Note that this table ignores Unicode.
-/* Crude home-grown routines for debugging some malloc-related
- problems. Featured:
+ ODS2 valid characters: 0-9 A-Z a-z $ - _ ~
- * 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.
+ ODS5 Invalid characters:
+ C0 control codes (0x00 to 0x1F inclusive)
+ Asterisk (*)
+ Question mark (?)
- * 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.
+ ODS5 Invalid characters only in VMS V7.2 (which no one runs, right?):
+ Double quotation marks (")
+ Backslash (\)
+ Colon (:)
+ Left angle bracket (<)
+ Right angle bracket (>)
+ Slash (/)
+ Vertical bar (|)
- 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. */
+ Characters escaped by "^":
+ SP ! " # % & ' ( ) + , . : ; =
+ @ [ \ ] ^ ` { | } ~
-static int malloc_count, free_count;
+ Either "^_" or "^ " is accepted as a space. Period (.) is a special
+ case. Note that un-escaped < and > can also confuse a directory
+ spec.
-static struct {
- char *ptr;
- const char *file;
- int line;
-} malloc_debug[100000];
+ Characters put out as ^xx:
+ 7F (DEL)
+ 80-9F (C1 control characters)
+ A0 (nonbreaking space)
+ FF (Latin small letter y diaeresis)
-/* 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. */
+ Other cases:
+ Unicode: "^Uxxxx", where "xxxx" is four hex digits.
-/* 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.) */
+ Property table values:
+ Normal escape: 1
+ Space: 2
+ Dot: 4
+ Hex-hex escape: 8
+ ODS2 normal: 16
+ ODS2 lower case: 32
+ Hex digit: 64
+*/
-static void
-register_ptr (void *ptr, const char *file, int line)
-{
- int i;
- for (i = 0; i < ARRAY_SIZE (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 ();
-}
+unsigned char char_prop[ 256] = {
-/* Unregister PTR from malloc_debug. Abort if PTR is not present in
- malloc_debug. (This catches calling free() with a bogus pointer.) */
+/* NUL SOH STX ETX EOT ENQ ACK BEL BS HT LF VT FF CR SO SI */
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-static void
-unregister_ptr (void *ptr)
-{
- int i;
- for (i = 0; i < ARRAY_SIZE (malloc_debug); i++)
- if (malloc_debug[i].ptr == ptr)
- {
- malloc_debug[i].ptr = NULL;
- return;
- }
- abort ();
-}
+/* DLE DC1 DC2 DC3 DC4 NAK SYN ETB CAN EM SUB ESC FS GS RS US */
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-/* 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. */
+/* SP ! " # $ % & ' ( ) * + , - . / */
+ 2, 1, 1, 1, 16, 1, 1, 1, 1, 1, 0, 1, 1, 16, 4, 0,
-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 < ARRAY_SIZE (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);
-}
+/* 0 1 2 3 4 5 6 7 8 9 : ; < = > ? */
+ 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 1, 1, 1, 1, 1, 1,
-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;
-}
+/* @ A B C D E F G H I J K L M N O */
+ 1, 80, 80, 80, 80, 80, 80, 16, 16, 16, 16, 16, 16, 16, 16, 16,
-void
-xfree_debug (void *ptr, const char *source_file, int source_line)
-{
- assert (ptr != NULL);
- ++free_count;
- unregister_ptr (ptr);
- free (ptr);
-}
+/* P Q R S T U V W X Y Z [ \ ] ^ _ */
+ 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 1, 1, 1, 1, 16,
-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;
-}
+/* ` a b c d e f g h i j k l m n o */
+ 1, 96, 96, 96, 96, 96, 96, 32, 32, 32, 32, 32, 32, 32, 32, 32,
-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;
-}
+/* p q r s t u v w x y z { | } ~ DEL */
+ 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 1, 1, 1, 17, 8,
+
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+ 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8
+};
-#endif /* DEBUG_MALLOC */
-\f
/* Utility function: like xstrdup(), but also lowercases S. */
char *
char *copy = xstrdup (s);
char *p = copy;
for (; *p; p++)
- *p = TOLOWER (*p);
+ *p = c_tolower (*p);
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;
while (*s)
{
if (*s == ',')
- {
- res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
- res[i] = strdupdelim (p, s);
- res[++i] = NULL;
- ++s;
- /* Skip the blanks following the ','. */
- while (ISSPACE (*s))
- ++s;
- p = s;
- }
+ {
+ res = xrealloc (res, (i + 2) * sizeof (char *));
+ res[i] = strdupdelim (p, s);
+ res[++i] = NULL;
+ ++s;
+ /* Skip the blanks following the ','. */
+ while (c_isspace (*s))
+ ++s;
+ p = s;
+ }
else
- ++s;
+ ++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.
+/* 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.
- If TM is non-NULL, the current time-in-seconds will be stored
- there.
+ 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. */
- (#### 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.) */
+/* Constant is using for limits memory allocation for text buffer.
+ Applicable in situation when: vasprintf is not available in the system
+ and vsnprintf return -1 when long line is truncated (in old versions of
+ glibc and in other system where C99 doesn`t support) */
+
+#define FMT_MAX_LENGTH 1048576
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)
+ memfatal ("aprintf", UNKNOWN_ATTEMPTED_SIZE); /* 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);
- if (secs == -1)
+ /* #### This code will infloop and eventually abort in xrealloc if
+ passed a FMT that causes snprintf to consistently return -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 if (size >= FMT_MAX_LENGTH) /* We have a huge buffer, */
+ { /* maybe we have some wrong
+ format string? */
+ logprintf (LOG_ALWAYS,
+ _("%s: aprintf: text buffer is too big (%ld bytes), "
+ "aborting.\n"),
+ exec_name, size); /* printout a log message */
+ abort (); /* and abort... */
+ }
+ else
+ {
+ /* else, we continue to grow our
+ * buffer: Twice the old size. */
+ size <<= 1;
+ }
+ 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;
+ size_t argcount;
+
+ /* 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);
- if (secs == -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. */
+ mswindows.c. On MSDOS this function should never be called. */
-#ifndef WINDOWS
+#ifdef __VMS
+
+void
+fork_to_background (void)
+{
+ return;
+}
+
+#else /* def __VMS */
+
+#if !defined(WINDOWS) && !defined(MSDOS)
void
fork_to_background (void)
{
pid_t pid;
/* Whether we arrange our own version of opt.lfilename here. */
- int changedp = 0;
+ bool logfile_changed = false;
- if (!opt.lfilename)
+ if (!opt.lfilename && (!opt.quiet || opt.server_response))
{
- opt.lfilename = unique_name (DEFAULT_LOGFILE);
- 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 (_("Output will be written to `%s'.\n"), opt.lfilename);
- exit (0); /* #### should we use _exit()? */
+ printf (_("Continuing in background, pid %d.\n"), (int) pid);
+ if (logfile_changed)
+ printf (_("Output will be written to %s.\n"), quote (opt.lfilename));
+ exit (0); /* #### should we use _exit()? */
}
/* child: give up the privileges and keep running. */
setsid ();
- freopen ("/dev/null", "r", stdin);
- freopen ("/dev/null", "w", stdout);
- freopen ("/dev/null", "w", stderr);
+ if (freopen ("/dev/null", "r", stdin) == NULL)
+ DEBUGP (("Failed to redirect stdin to /dev/null.\n"));
+ if (freopen ("/dev/null", "w", stdout) == NULL)
+ DEBUGP (("Failed to redirect stdout to /dev/null.\n"));
+ if (freopen ("/dev/null", "w", stderr) == NULL)
+ DEBUGP (("Failed to redirect stderr to /dev/null.\n"));
}
-#endif /* not WINDOWS */
-\f
-/* Resolve "." and ".." elements of PATH by destructively modifying
- PATH. "." is resolved by removing that path element, and ".." is
- resolved by removing the preceding path element. Leading and
- trailing slashes are preserved.
-
- Return non-zero if any changes have been made.
-
- For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
- test examples are provided below. If you change anything in this
- function, run test_path_simplify to make sure you haven't broken a
- test case.
+#endif /* !WINDOWS && !MSDOS */
- A previous version of this function was based on path_simplify()
- from GNU Bash, but it has been rewritten for Wget 1.8.1. */
-
-int
-path_simplify (char *path)
-{
- int change = 0;
- char *p, *end;
-
- if (path[0] == '/')
- ++path; /* preserve the leading '/'. */
-
- p = path;
- end = p + strlen (p) + 1; /* position past the terminating zero. */
-
- while (1)
- {
- again:
- /* P should point to the beginning of a path element. */
-
- if (*p == '.' && (*(p + 1) == '/' || *(p + 1) == '\0'))
- {
- /* Handle "./foo" by moving "foo" two characters to the
- left. */
- if (*(p + 1) == '/')
- {
- change = 1;
- memmove (p, p + 2, end - p);
- end -= 2;
- goto again;
- }
- else
- {
- change = 1;
- *p = '\0';
- break;
- }
- }
- else if (*p == '.' && *(p + 1) == '.'
- && (*(p + 2) == '/' || *(p + 2) == '\0'))
- {
- /* Handle "../foo" by moving "foo" one path element to the
- left. */
- char *b = p; /* not p-1 because P can equal PATH */
-
- /* Backtrack by one path element, but not past the beginning
- of PATH. */
-
- /* foo/bar/../baz */
- /* ^ p */
- /* ^ b */
-
- if (b > path)
- {
- /* Move backwards until B hits the beginning of the
- previous path element or the beginning of path. */
- for (--b; b > path && *(b - 1) != '/'; b--)
- ;
- }
-
- change = 1;
- if (*(p + 2) == '/')
- {
- memmove (b, p + 3, end - (p + 3));
- end -= (p + 3) - b;
- p = b;
- }
- else
- {
- *b = '\0';
- break;
- }
-
- goto again;
- }
- else if (*p == '/')
- {
- /* Remove empty path elements. Not mandated by rfc1808 et
- al, but empty path elements are not all that useful, and
- the rest of Wget might not deal with them well. */
- char *q = p;
- while (*q == '/')
- ++q;
- change = 1;
- if (*q == '\0')
- {
- *p = '\0';
- break;
- }
- memmove (p, q, end - q);
- end -= q - p;
- goto again;
- }
-
- /* Skip to the next path element. */
- while (*p && *p != '/')
- ++p;
- if (*p == '\0')
- break;
-
- /* Make sure P points to the beginning of the next path element,
- which is location after the slash. */
- ++p;
- }
+#endif /* def __VMS [else] */
- return change;
-}
\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
+#if HAVE_UTIME
+# ifdef HAVE_STRUCT_UTIMBUF
struct utimbuf times;
- times.actime = times.modtime = tm;
-#else
- time_t times[2];
- times[0] = times[1] = tm;
-#endif
-
+# else
+ 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));
+#else
+ struct timespec timespecs[2];
+ int fd;
+
+ fd = open (file, O_WRONLY);
+ if (fd < 0)
+ {
+ logprintf (LOG_NOTQUIET, "open(%s): %s\n", file, strerror (errno));
+ return;
+ }
+
+ timespecs[0].tv_sec = time (NULL);
+ timespecs[0].tv_nsec = 0L;
+ timespecs[1].tv_sec = tm;
+ timespecs[1].tv_nsec = 0L;
+
+ if (futimens (fd, timespecs) == -1)
+ logprintf (LOG_NOTQUIET, "futimens(%s): %s\n", file, strerror (errno));
+
+ close (fd);
+#endif
}
/* Checks if FILE is a symbolic link, and removes it if it is. Does
remove_link (const char *file)
{
int err = 0;
- struct stat st;
+ struct_stat st;
if (lstat (file, &st) == 0 && S_ISLNK (st.st_mode))
{
DEBUGP (("Unlinking %s (symlink).\n", file));
err = unlink (file);
if (err != 0)
- logprintf (LOG_VERBOSE, _("Failed to unlink symlink `%s': %s\n"),
- file, strerror (errno));
+ logprintf (LOG_VERBOSE, _("Failed to unlink symlink %s: %s\n"),
+ quote (file), strerror (errno));
}
return err;
}
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. */
+wgint
+file_size (const char *filename)
+{
+#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 that the file is readable without
+ explicitly checking for permissions. Inspired by the POST patch
+ by Arnaud Wylie. */
+ FILE *fp = fopen (filename, "rb");
+ 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
}
-/* Return a unique filename, given a prefix and count */
+/* 2005-02-19 SMS.
+ If no UNIQ_SEP is defined (as on VMS), have unique_name() return the
+ original name. With the VMS file systems' versioning, everything
+ should be fine, and appending ".NN" just causes trouble.
+*/
+
+#ifdef UNIQ_SEP
+
+/* stat file names named PREFIX.1, PREFIX.2, etc., until one that
+ doesn't exist is found. Return a freshly allocated copy of the
+ unused file name. */
+
static char *
-unique_name_1 (const char *fileprefix, int count)
+unique_name_1 (const char *prefix)
+{
+ int count = 1;
+ int plen = strlen (prefix);
+ char *template = (char *)alloca (plen + 1 + 24);
+ char *template_tail = template + plen;
+
+ memcpy (template, prefix, plen);
+ *template_tail++ = UNIQ_SEP;
+
+ do
+ number_to_string (template_tail, count++);
+ while (file_exists_p (template));
+
+ return xstrdup (template);
+}
+
+/* Return a unique file name, based on FILE.
+
+ More precisely, if FILE doesn't exist, it is returned unmodified.
+ If not, FILE.1 is tried, then FILE.2, etc. The first FILE.<number>
+ file name that doesn't exist is returned.
+
+ 2005-02-19 SMS. "." is now UNIQ_SEP, and may be different.
+
+ The resulting file is not created, only verified that it didn't
+ 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
+ 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, bool allow_passthrough)
+{
+ /* If the FILE itself doesn't exist, return it without
+ modification. */
+ if (!file_exists_p (file))
+ return allow_passthrough ? (char *)file : xstrdup (file);
+
+ /* Otherwise, find a numeric suffix that results in unused file name
+ and return it. */
+ return unique_name_1 (file);
+}
+
+#else /* def UNIQ_SEP */
+
+/* Dummy unique_name() for VMS. Return the original name as easily as
+ possible.
+*/
+char *
+unique_name (const char *file, bool allow_passthrough)
{
- char *filename;
+ /* Return the FILE itself, without modification, irregardful. */
+ return allow_passthrough ? (char *)file : xstrdup (file);
+}
+
+#endif /* def UNIQ_SEP [else] */
+
+/* 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. */
- if (count)
+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)
{
- filename = (char *)xmalloc (strlen (fileprefix) + numdigit (count) + 2);
- sprintf (filename, "%s.%d", fileprefix, count);
+ xfree (uname);
+ uname = unique_name (name, false);
}
- else
- filename = xstrdup (fileprefix);
-
- if (!file_exists_p (filename))
- return filename;
- else
+ if (opened_name)
{
- xfree (filename);
- return NULL;
+ if (fp)
+ *opened_name = uname;
+ else
+ {
+ *opened_name = NULL;
+ xfree (uname);
+ }
}
+ else
+ xfree (uname);
+ return fp;
}
-/* Return a unique file name, based on PREFIX. */
-char *
-unique_name (const char *prefix)
+/* 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, int binary)
{
- char *file = NULL;
- int count = 0;
+ int fd;
+#ifdef O_EXCL
+
+/* 2005-04-14 SMS.
+ VMS lacks O_BINARY, but makes up for it in weird and wonderful ways.
+ It also has file versions which obviate all the O_EXCL effort.
+ O_TRUNC (something of a misnomer) requests a new version.
+*/
+# ifdef __VMS
+/* Common open() optional arguments:
+ sequential access only, access callback function.
+*/
+# define OPEN_OPT_ARGS "fop=sqo", "acc", acc_cb, &open_id
+
+ int open_id;
+ int flags = O_WRONLY | O_CREAT | O_TRUNC;
+
+ if (binary > 1)
+ {
+ open_id = 11;
+ fd = open( fname, /* File name. */
+ flags, /* Flags. */
+ 0777, /* Mode for default protection. */
+ "ctx=bin,stm", /* Binary, stream access. */
+ "rfm=stmlf", /* Stream_LF. */
+ OPEN_OPT_ARGS); /* Access callback. */
+ }
+ else if (binary)
+ {
+ open_id = 12;
+ fd = open( fname, /* File name. */
+ flags, /* Flags. */
+ 0777, /* Mode for default protection. */
+ "ctx=bin,stm", /* Binary, stream access. */
+ "rfm=fix", /* Fixed-length, */
+ "mrs=512", /* 512-byte records. */
+ OPEN_OPT_ARGS); /* Access callback. */
+ }
+ else
+ {
+ open_id = 13;
+ fd = open( fname, /* File name. */
+ flags, /* Flags. */
+ 0777, /* Mode for default protection. */
+ "rfm=stmlf", /* Stream_LF. */
+ OPEN_OPT_ARGS); /* Access callback. */
+ }
+# else /* def __VMS */
+ int flags = O_WRONLY | O_CREAT | O_EXCL;
+# ifdef O_BINARY
+ if (binary)
+ flags |= O_BINARY;
+# endif
+ fd = open (fname, flags, 0666);
+# endif /* def __VMS [else] */
- while (!file)
- file = unique_name_1 (prefix, count++);
- return file;
+ 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
int
make_directory (const char *directory)
{
- int quit = 0;
- int i;
+ int i, ret, quit = 0;
char *dir;
/* Make a copy of dir, to be able to write to it. Otherwise, the
for (i = (*dir == '/'); 1; ++i)
{
for (; dir[i] && dir[i] != '/'; i++)
- ;
+ ;
if (!dir[i])
- quit = 1;
+ quit = 1;
dir[i] = '\0';
- /* Check whether the directory already exists. */
+ /* Check whether the directory already exists. Allow creation of
+ of intermediate directories to fail, as the initial path components
+ are not necessarily directories! */
if (!file_exists_p (dir))
- {
- if (mkdir (dir, 0777) < 0)
- return -1;
- }
+ ret = mkdir (dir, 0777);
+ else
+ ret = 0;
if (quit)
- break;
+ break;
else
- dir[i] = '/';
+ dir[i] = '/';
}
- return 0;
+ return ret;
}
/* Merge BASE with FILE. BASE can be a directory or a file name, FILE
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 = c_tolower (*pattern);
+ *p = '\0';
+ for (p = strcopy; *string; string++, p++)
+ *p = c_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);
+ const char *p;
+
+ if (opt.output_document && strcmp (s, opt.output_document) == 0)
+ return true;
+
+ if ((p = strrchr (s, '/')))
+ s = p + 1;
- while (l && s[l] != '/')
- --l;
- if (s[l] == '/')
- s += (l + 1);
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)
+/* Determine whether an URL is acceptable to be followed, according to
+ regex patterns to accept/reject. */
+bool
+accept_url (const char *s)
{
- for (; *s1 && *s2 && (*s1 == *s2); ++s1, ++s2);
- return !*s1;
+ if (opt.acceptregex && !opt.regex_match_fun (opt.acceptregex, s))
+ return false;
+ if (opt.rejectregex && opt.regex_match_fun (opt.rejectregex, s))
+ return false;
+
+ return true;
}
-/* 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)
+/* 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)
+{
+ if (*d1 == '\0')
+ return true;
+ if (!opt.ignore_case)
+ for (; *d1 && *d2 && (*d1 == *d2); ++d1, ++d2)
+ ;
+ else
+ for (; *d1 && *d2 && (c_tolower (*d1) == c_tolower (*d2)); ++d1, ++d2)
+ ;
+
+ return *d1 == '\0' && (*d2 == '\0' || *d2 == '/');
+}
+
+/* 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 (const char **dirlist, const char *dir)
{
- char **x;
+ const 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 '/' */
+ const 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;
}
-/* Match the end of STRING against PATTERN. For instance:
+/* Return true if STRING ends with TAIL. For instance:
- match_backwards ("abc", "bc") -> 1
- match_backwards ("abc", "ab") -> 0
- match_backwards ("abc", "abc") -> 1 */
-int
-match_tail (const char *string, const char *pattern)
+ match_tail ("abc", "bc", false) -> 1
+ match_tail ("abc", "ab", false) -> 0
+ match_tail ("abc", "abc", false) -> 1
+
+ If FOLD_CASE is true, the comparison will be case-insensitive. */
+
+bool
+match_tail (const char *string, const char *tail, bool fold_case)
{
- int i, j;
+ int pos = strlen (string) - strlen (tail);
- for (i = strlen (string), j = strlen (pattern); i >= 0 && j >= 0; i--, j--)
- if (string[i] != pattern[j])
- break;
- /* If the pattern was exhausted, the match was succesful. */
- if (j == -1)
- return 1;
+ if (pos < 0)
+ return false; /* tail is longer than string. */
+
+ if (!fold_case)
+ return !strcmp (string + pos, tail);
else
- return 0;
+ return !strcasecmp (string + pos, tail);
}
/* 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))
- return 1;
- }
- else
- {
- if (!strcmp (s, *accepts))
- return 1;
- }
- }
+ {
+ if (backward)
+ {
+ if (match_tail (s, *accepts, opt.ignore_case))
+ return true;
+ }
+ else
+ {
+ 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:
char *
suffix (const char *str)
{
- int i;
+ char *p;
- for (i = strlen (str); i && str[i] != '/' && str[i] != '.'; i--)
- ;
+ if ((p = strrchr (str, '.')) && !strchr (p + 1, '/'))
+ return p + 1;
- if (str[i++] == '.')
- return (char *)str + i;
- else
- return NULL;
+ return NULL;
}
-/* 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.
+/* Return true if S contains globbing wildcards (`*', `?', `[' or
+ `]'). */
- 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
- terminated with a zero character.
+bool
+has_wildcards_p (const char *s)
+{
+ return !!strpbrk (s, "*?[]");
+}
- After end-of-file is encountered without anything being read, NULL
- is returned. NULL is also returned on error. To distinguish
- between these two cases, use the stdio function ferror(). */
+/* Return true if FNAME ends with a typical HTML suffix. The
+ following (case-insensitive) suffixes are presumed to be HTML
+ files:
-char *
-read_whole_line (FILE *fp)
+ html
+ htm
+ ?html (`?' matches one character)
+
+ #### CAVEAT. This is not necessarily a good indication that FNAME
+ refers to a file that contains HTML! */
+bool
+has_html_suffix_p (const char *fname)
{
- int length = 0;
- int bufsize = 82;
- char *line = (char *)xmalloc (bufsize);
+ char *suf;
- while (fgets (line + length, bufsize - length, fp))
- {
- length += strlen (line + length);
- if (length == 0)
- /* Possible for example when reading from a binary file where
- a line begins with \0. */
- continue;
-
- if (line[length - 1] == '\n')
- break;
-
- /* fgets() guarantees to read the whole line, or to use up the
- space we've given it. We can double the buffer
- unconditionally. */
- bufsize <<= 1;
- line = xrealloc (line, bufsize);
- }
- if (length == 0 || ferror (fp))
- {
- xfree (line);
- return NULL;
- }
- if (length + 1 < bufsize)
- /* Relieve the memory from our exponential greediness. We say
- `length + 1' because the terminating \0 is not included in
- LENGTH. We don't need to zero-terminate the string ourselves,
- though, because fgets() does that. */
- line = xrealloc (line, length + 1);
- return line;
+ if ((suf = suffix (fname)) == NULL)
+ return false;
+ if (!strcasecmp (suf, "html"))
+ return true;
+ if (!strcasecmp (suf, "htm"))
+ return true;
+ if (suf[0] && !strcasecmp (suf + 1, "html"))
+ return true;
+ return false;
}
-\f
+
/* Read FILE into memory. A pointer to `struct file_memory' are
returned; use struct element `content' to access file contents, and
the element `length' to know the file length. `content' is *not*
zero-terminated, and you should *not* read or write beyond the [0,
length) range of characters.
- After you are done with the file contents, call read_file_free to
+ After you are done with the file contents, call wget_read_file_free to
release the memory.
Depending on the operating system and the type of file that is
- being read, read_file() either mmap's the file into memory, or
+ being read, wget_read_file() either mmap's the file into memory, or
reads the file into the core using read().
If file is named "-", fileno(stdin) is used for reading instead.
If you want to read from a real file named "-", use "./-" instead. */
struct file_memory *
-read_file (const char *file)
+wget_read_file (const char *file)
{
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;
specify PROT_READ and MAP_SHARED for a marginal gain in
efficiency, but at some cost to generality. */
fm->content = mmap (NULL, fm->length, PROT_READ | PROT_WRITE,
- MAP_PRIVATE, fd, 0);
+ MAP_PRIVATE, fd, 0);
if (fm->content == (char *)MAP_FAILED)
goto mmap_lose;
if (!inhibit_close)
#endif /* HAVE_MMAP */
fm->length = 0;
- size = 512; /* number of bytes fm->contents can
+ size = 512; /* number of bytes fm->contents can
hold at any given time. */
fm->content = xmalloc (size);
while (1)
{
- long nread;
+ wgint nread;
if (fm->length > size / 2)
- {
- /* #### I'm not sure whether the whole exponential-growth
+ {
+ /* #### I'm not sure whether the whole exponential-growth
thing makes sense with kernel read. On Linux at least,
read() refuses to read more than 4K from a file at a
single chunk anyway. But other Unixes might optimize it
better, and it doesn't *hurt* anything, so I'm leaving
it. */
- /* Normally, we grow SIZE exponentially to make the number
+ /* 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. */
- size <<= 1;
- fm->content = xrealloc (fm->content, size);
- }
+ size <<= 1;
+ fm->content = xrealloc (fm->content, size);
+ }
nread = read (fd, fm->content + fm->length, size - fm->length);
if (nread > 0)
- /* Successful read. */
- fm->length += nread;
+ /* Successful read. */
+ fm->length += nread;
else if (nread < 0)
- /* Error. */
- goto lose;
+ /* Error. */
+ goto lose;
else
- /* EOF */
- break;
+ /* EOF */
+ break;
}
if (!inhibit_close)
close (fd);
memory needed to hold the FM structure itself. */
void
-read_file_free (struct file_memory *fm)
+wget_read_file_free (struct file_memory *fm)
{
#ifdef HAVE_MMAP
if (fm->mmap_p)
{
char **p = vec;
while (*p)
- xfree (*p++);
+ xfree (*p++);
xfree (vec);
}
}
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 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. */
+/* 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. */
-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 ! USE_NLS_PROGRESS_BAR
+ /* We can't count column widths, so ensure that the separator
+ * is single-byte only (let check below determine what byte). */
+ if (strlen(cached_sep) > 1)
+ cached_sep = "";
+#endif
+ 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, const int acc, const int decimals)
{
- 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];
+ size_t 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 the accuracy level (acc) with (decimal)
+ * decimal digits, and others without any decimals. */
+ snprintf (buf, sizeof (buf), "%.*f%c",
+ val < acc ? decimals : 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;
}
-#define ONE_DIGIT(figure) *p++ = n / (figure) + '0'
-#define ONE_DIGIT_ADVANCE(figure) (ONE_DIGIT (figure), n %= (figure))
+#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)
-#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 wgints. */
-/* DIGITS_<11-20> are only used on machines with 64-bit longs. */
+#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)
-#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)
+/* Shorthand for casting to wgint. */
+#define W wgint
-/* Print NUMBER to BUFFER in base 10. This should be completely
- equivalent to `sprintf(buffer, "%ld", number)', only much faster.
+/* 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 < -WGINT_MAX)
+ {
+ /* 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
+
+#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.) */
+
+char *
+number_to_static_string (wgint number)
+{
+ static char ring[RING_SIZE][24];
+ static int ringpos;
+ char *buf = ring[ringpos];
+ number_to_string (buf, number);
+ ringpos = (ringpos + 1) % RING_SIZE;
+ return buf;
+}
+
+/* Converts the byte to bits format if --report-bps option is enabled
+ */
+wgint
+convert_to_bits (wgint num)
+{
+ if (opt.report_bps)
+ return num * 8;
+ return num;
+}
+
\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 */
+/* Determine the width of the terminal we're running on. If that's
+ not possible, return 0. */
-struct wget_timer {
-#ifdef TIMER_GETTIMEOFDAY
- long secs;
- long usecs;
-#endif
+int
+determine_screen_width (void)
+{
+ /* If there's a way to get the terminal size using POSIX
+ tcgetattr(), somebody please tell me. */
+#ifdef TIOCGWINSZ
+ int fd;
+ struct winsize wsz;
-#ifdef TIMER_TIME
- time_t secs;
-#endif
+ if (opt.lfilename != NULL)
+ return 0;
-#ifdef TIMER_WINDOWS
- ULARGE_INTEGER wintime;
-#endif
-};
+ fd = fileno (stderr);
+ if (ioctl (fd, TIOCGWINSZ, &wsz) < 0)
+ return 0; /* most likely ENOTTY */
-/* Allocate a timer. It is not legal to do anything with a freshly
- allocated timer, except call wtimer_reset() or wtimer_delete(). */
+ return wsz.ws_col;
+#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;
-struct wget_timer *
-wtimer_allocate (void)
+/* Return a random number between 0 and MAX-1, inclusive.
+
+ 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 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)
{
- struct wget_timer *wt =
- (struct wget_timer *)xmalloc (sizeof (struct wget_timer));
- return wt;
+#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 (!rnd_seeded)
+ {
+ srand ((unsigned) time (NULL) ^ (unsigned) getpid ());
+ rnd_seeded = 1;
+ }
+ rnd = rand ();
+
+ /* Like rand() % max, but uses the high-order bits for better
+ randomness on architectures where rand() is implemented using a
+ simple congruential generator. */
+
+ bounded = (double) max * rnd / (RAND_MAX + 1.0);
+ return (int) bounded;
+
+#endif /* not HAVE_DRAND48 */
}
-/* Allocate a new timer and reset it. Return the new timer. */
+/* Return a random uniformly distributed floating point number in the
+ [0, 1) range. Uses drand48 where available, and a really lame
+ kludge elsewhere. */
-struct wget_timer *
-wtimer_new (void)
+double
+random_float (void)
{
- struct wget_timer *wt = wtimer_allocate ();
- wtimer_reset (wt);
- return wt;
+#ifdef HAVE_DRAND48
+ if (!rnd_seeded)
+ {
+ srand48 ((long) time (NULL) ^ (long) getpid ());
+ rnd_seeded = 1;
+ }
+ 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 */
}
+\f
+/* Implementation of run_with_timeout, a generic timeout-forcing
+ routine for systems with Unix-like signal handling. */
-/* Free the resources associated with the timer. Its further use is
- prohibited. */
+#ifdef USE_SIGNAL_TIMEOUT
+# ifdef HAVE_SIGSETJMP
+# define SETJMP(env) sigsetjmp (env, 1)
-void
-wtimer_delete (struct wget_timer *wt)
+static sigjmp_buf run_with_timeout_env;
+
+static void
+abort_run_with_timeout (int sig)
{
- xfree (wt);
+ assert (sig == SIGALRM);
+ siglongjmp (run_with_timeout_env, -1);
}
+# else /* not HAVE_SIGSETJMP */
+# define SETJMP(env) setjmp (env)
-/* 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. */
+static jmp_buf run_with_timeout_env;
-void
-wtimer_reset (struct wget_timer *wt)
+static void
+abort_run_with_timeout (int sig)
{
-#ifdef TIMER_GETTIMEOFDAY
- struct timeval t;
- gettimeofday (&t, NULL);
- wt->secs = t.tv_sec;
- wt->usecs = t.tv_usec;
-#endif
+ assert (sig == SIGALRM);
+ /* We don't have siglongjmp to preserve the set of blocked signals;
+ if we longjumped out of the handler at this point, SIGALRM would
+ remain blocked. We must unblock it manually. */
+ sigset_t set;
+ sigemptyset (&set);
+ sigaddset (&set, SIGALRM);
+ sigprocmask (SIG_BLOCK, &set, NULL);
-#ifdef TIMER_TIME
- wt->secs = time (NULL);
-#endif
+ /* Now it's safe to longjump. */
+ longjmp (run_with_timeout_env, -1);
+}
+# endif /* not HAVE_SIGSETJMP */
-#ifdef TIMER_WINDOWS
- FILETIME ft;
- SYSTEMTIME st;
- GetSystemTime (&st);
- SystemTimeToFileTime (&st, &ft);
- wt->wintime.HighPart = ft.dwHighDateTime;
- wt->wintime.LowPart = ft.dwLowDateTime;
-#endif
+/* Arrange for SIGALRM to be delivered in TIMEOUT seconds. This uses
+ setitimer where available, alarm otherwise.
+
+ TIMEOUT should be non-zero. If the timeout value is so small that
+ it would be rounded to zero, it is rounded to the least legal value
+ instead (1us for setitimer, 1s for alarm). That ensures that
+ SIGALRM will be delivered in all cases. */
+
+static void
+alarm_set (double timeout)
+{
+#ifdef ITIMER_REAL
+ /* Use the modern itimer interface. */
+ struct itimerval itv;
+ xzero (itv);
+ itv.it_value.tv_sec = (long) timeout;
+ itv.it_value.tv_usec = 1000000 * (timeout - (long)timeout);
+ if (itv.it_value.tv_sec == 0 && itv.it_value.tv_usec == 0)
+ /* Ensure that we wait for at least the minimum interval.
+ Specifying zero would mean "wait forever". */
+ itv.it_value.tv_usec = 1;
+ setitimer (ITIMER_REAL, &itv, NULL);
+#else /* not ITIMER_REAL */
+ /* Use the old alarm() interface. */
+ int secs = (int) timeout;
+ if (secs == 0)
+ /* Round TIMEOUTs smaller than 1 to 1, not to zero. This is
+ because alarm(0) means "never deliver the alarm", i.e. "wait
+ forever", which is not what someone who specifies a 0.5s
+ timeout would expect. */
+ secs = 1;
+ alarm (secs);
+#endif /* not ITIMER_REAL */
+}
+
+/* Cancel the alarm set with alarm_set. */
+
+static void
+alarm_cancel (void)
+{
+#ifdef ITIMER_REAL
+ struct itimerval disable;
+ xzero (disable);
+ setitimer (ITIMER_REAL, &disable, NULL);
+#else /* not ITIMER_REAL */
+ alarm (0);
+#endif /* not ITIMER_REAL */
}
-/* 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. */
+/* 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.
-long
-wtimer_elapsed (struct wget_timer *wt)
+ This works by setting up SIGALRM to be delivered in TIMEOUT seconds
+ using setitimer() or alarm(). The timeout is enforced by
+ longjumping out of the SIGALRM handler. This has several
+ advantages compared to the traditional approach of relying on
+ signals causing system calls to exit with EINTR:
+
+ * The callback function is *forcibly* interrupted after the
+ timeout expires, (almost) regardless of what it was doing and
+ whether it was in a syscall. For example, a calculation that
+ takes a long time is interrupted as reliably as an IO
+ operation.
+
+ * It works with both SYSV and BSD signals because it doesn't
+ depend on the default setting of SA_RESTART.
+
+ * It doesn't require special handler setup beyond a simple call
+ to signal(). (It does use sigsetjmp/siglongjmp, but they're
+ optional.)
+
+ The only downside is that, if FUN allocates internal resources that
+ are normally freed prior to exit from the functions, they will be
+ lost in case of timeout. */
+
+bool
+run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
-#ifdef TIMER_GETTIMEOFDAY
- struct timeval t;
- gettimeofday (&t, NULL);
- return (t.tv_sec - wt->secs) * 1000 + (t.tv_usec - wt->usecs) / 1000;
-#endif
+ int saved_errno;
-#ifdef TIMER_TIME
- time_t now = time (NULL);
- return 1000 * (now - wt->secs);
-#endif
+ if (timeout == 0)
+ {
+ fun (arg);
+ return false;
+ }
-#ifdef WINDOWS
- FILETIME ft;
- SYSTEMTIME st;
- ULARGE_INTEGER uli;
- GetSystemTime (&st);
- SystemTimeToFileTime (&st, &ft);
- uli.HighPart = ft.dwHighDateTime;
- uli.LowPart = ft.dwLowDateTime;
- return (long)((uli.QuadPart - wt->wintime.QuadPart) / 10000);
-#endif
+ signal (SIGALRM, abort_run_with_timeout);
+ if (SETJMP (run_with_timeout_env) != 0)
+ {
+ /* Longjumped out of FUN with a timeout. */
+ signal (SIGALRM, SIG_DFL);
+ return true;
+ }
+ alarm_set (timeout);
+ fun (arg);
+
+ /* Preserve errno in case alarm() or signal() modifies it. */
+ saved_errno = errno;
+ alarm_cancel ();
+ signal (SIGALRM, SIG_DFL);
+ errno = saved_errno;
+
+ return false;
}
-/* Return the assessed granularity of the timer implementation. This
- is important for certain code that tries to deal with "zero" time
- intervals. */
+#else /* not USE_SIGNAL_TIMEOUT */
-long
-wtimer_granularity (void)
+#ifndef WINDOWS
+/* A stub version of run_with_timeout that just calls FUN(ARG). Don't
+ define it under Windows, because Windows has its own version of
+ run_with_timeout that uses threads. */
+
+bool
+run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
-#ifdef TIMER_GETTIMEOFDAY
- /* Granularity of gettimeofday is hugely architecture-dependent.
- However, it appears that on modern machines it is better than
- 1ms. */
- return 1;
-#endif
+ fun (arg);
+ return false;
+}
+#endif /* not WINDOWS */
+#endif /* not USE_SIGNAL_TIMEOUT */
+\f
+#ifndef WINDOWS
-#ifdef TIMER_TIME
- /* This is clear. */
- return 1000;
-#endif
+/* Sleep the specified amount of seconds. On machines without
+ nanosleep(), this may sleep shorter if interrupted by signals. */
-#ifdef TIMER_WINDOWS
- /* ? */
- return 1;
+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
}
-\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:
+#endif /* not WINDOWS */
- `&' -> `&'
- `<' -> `<'
- `>' -> `>'
- `"' -> `"'
- SP -> ` '
+/* 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.
- No other entities are recognized or replaced. */
-char *
-html_quote_string (const char *s)
-{
- const char *b = s;
- char *p, *res;
- int i;
+ This implementation does not emit newlines after 76 characters of
+ base64 data. */
- /* Pass through the string, and count the new size. */
- for (i = 0; *s; s++, i++)
+size_t
+base64_encode (const void *data, size_t 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)
{
- 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; */
+ *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];
}
- res = (char *)xmalloc (i + 1);
- s = b;
- for (p = res; *s; s++)
+
+ /* Pad the result if necessary... */
+ switch (length % 3)
{
- 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;
- }
+ 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 res;
+
+ return p - dest;
}
-/* Determine the width of the terminal we're running on. If that's
- not possible, return 0. */
+/* 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 (c_isspace (c))
-int
-determine_screen_width (void)
+#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. */
+
+ssize_t
+base64_decode (const char *base64, void *dest)
{
- /* If there's a way to get the terminal size using POSIX
- tcgetattr(), somebody please tell me. */
-#ifndef TIOCGWINSZ
- return 0;
-#else /* TIOCGWINSZ */
- int fd;
- struct winsize wsz;
+ /* 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;
- if (opt.lfilename != NULL)
- return 0;
+ 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
- fd = fileno (stderr);
- if (ioctl (fd, TIOCGWINSZ, &wsz) < 0)
- return 0; /* most likely ENOTTY */
+ return q - (char *) dest;
+}
- return wsz.ws_col;
-#endif /* TIOCGWINSZ */
+#ifdef HAVE_LIBPCRE
+/* Compiles the PCRE regex. */
+void *
+compile_pcre_regex (const char *str)
+{
+ const char *errbuf;
+ int erroffset;
+ pcre *regex = pcre_compile (str, 0, &errbuf, &erroffset, 0);
+ if (! regex)
+ {
+ fprintf (stderr, _("Invalid regular expression %s, %s\n"),
+ quote (str), errbuf);
+ return false;
+ }
+ return regex;
}
+#endif
+
+/* Compiles the POSIX regex. */
+void *
+compile_posix_regex (const char *str)
+{
+ regex_t *regex = xmalloc (sizeof (regex_t));
+ int errcode = regcomp ((regex_t *) regex, str, REG_EXTENDED | REG_NOSUB);
+ if (errcode != 0)
+ {
+ size_t errbuf_size = regerror (errcode, (regex_t *) regex, NULL, 0);
+ char *errbuf = xmalloc (errbuf_size);
+ regerror (errcode, (regex_t *) regex, errbuf, errbuf_size);
+ fprintf (stderr, _("Invalid regular expression %s, %s\n"),
+ quote (str), errbuf);
+ xfree (errbuf);
+ return NULL;
+ }
-#if 0
-/* A debugging function for checking whether an MD5 library works. */
+ return regex;
+}
-#include "gen-md5.h"
+#ifdef HAVE_LIBPCRE
+#define OVECCOUNT 30
+/* Matches a PCRE regex. */
+bool
+match_pcre_regex (const void *regex, const char *str)
+{
+ size_t l = strlen (str);
+ int ovector[OVECCOUNT];
-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--)
+ int rc = pcre_exec ((pcre *) regex, 0, str, (int) l, 0, 0, ovector, OVECCOUNT);
+ if (rc == PCRE_ERROR_NOMATCH)
+ return false;
+ else if (rc < 0)
{
- *p2++ = XDIGIT_TO_xchar (*p1 >> 4);
- *p2++ = XDIGIT_TO_xchar (*p1 & 0xf);
- ++p1;
+ logprintf (LOG_VERBOSE, _("Error while matching %s: %d\n"),
+ quote (str), rc);
+ return false;
}
- *p2 = '\0';
-
- return res;
+ else
+ return true;
}
+#undef OVECCOUNT
#endif
-#if 0
-/* Debugging and testing support for path_simplify. */
-
-/* Debug: run path_simplify on PATH and return the result in a new
- string. Useful for calling from the debugger. */
-static char *
-ps (char *path)
+/* Matches a POSIX regex. */
+bool
+match_posix_regex (const void *regex, const char *str)
{
- char *copy = xstrdup (path);
- path_simplify (copy);
- return copy;
+ int rc = regexec ((regex_t *) regex, str, 0, NULL, 0);
+ if (rc == REG_NOMATCH)
+ return false;
+ else if (rc == 0)
+ return true;
+ else
+ {
+ size_t errbuf_size = regerror (rc, opt.acceptregex, NULL, 0);
+ char *errbuf = xmalloc (errbuf_size);
+ regerror (rc, opt.acceptregex, errbuf, errbuf_size);
+ logprintf (LOG_VERBOSE, _("Error while matching %s: %d\n"),
+ quote (str), rc);
+ xfree (errbuf);
+ return false;
+ }
}
+#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
-run_test (char *test, char *expected_result, int expected_change)
+mergesort_internal (void *base, void *temp, size_t size, size_t from, size_t to,
+ int (*cmpfun) (const void *, const void *))
{
- char *test_copy = xstrdup (test);
- int modified = path_simplify (test_copy);
+#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
+}
- if (0 != strcmp (test_copy, expected_result))
+/* 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)
{
- printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
- test, expected_result, test_copy);
+ void *temp = alloca (nmemb * size * sizeof (void *));
+ mergesort_internal (base, temp, size, 0, nmemb - 1, cmpfun);
}
- if (modified != expected_change)
+}
+\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;
+}
+
+/* Get the maximum name length for the given path. */
+/* Return 0 if length is unknown. */
+long
+get_max_length (const char *path, int length, int name)
+{
+ long ret;
+ char *p, *d;
+
+ /* Make a copy of the path that we can modify. */
+ p = path ? strdupdelim (path, path + length) : strdup ("");
+
+ for (;;)
{
- if (expected_change == 1)
- printf ("Expected no modification with path_simplify(\"%s\").\n",
- test);
+ errno = 0;
+ /* For an empty path query the current directory. */
+#if HAVE_PATHCONF
+ ret = pathconf (*p ? p : ".", name);
+ if (!(ret < 0 && errno == ENOENT))
+ break;
+#else
+ ret = PATH_MAX;
+#endif
+
+ /* The path does not exist yet, but may be created. */
+ /* Already at current or root directory, give up. */
+ if (!*p || strcmp (p, "/") == 0)
+ break;
+
+ /* Remove one directory level and try again. */
+ d = strrchr (p, '/');
+ if (d == p)
+ p[1] = '\0'; /* check root directory */
+ else if (d)
+ *d = '\0'; /* remove last directory part */
else
- printf ("Expected modification with path_simplify(\"%s\").\n",
- test);
+ *p = '\0'; /* check current directory */
}
- xfree (test_copy);
+
+ xfree (p);
+
+ if (ret < 0)
+ {
+ /* pathconf() has a message for us. */
+ if (errno != 0)
+ perror ("pathconf");
+
+ /* If (errno == 0) then there is no max length.
+ Even on error return 0 so the caller can continue. */
+ return 0;
+ }
+
+ return ret;
}
-static void
-test_path_simplify (void)
+#ifdef TESTING
+
+const char *
+test_subdir_p(void)
{
- static struct {
- char *test, *result;
- int should_modify;
- } tests[] = {
- { "", "", 0 },
- { ".", "", 1 },
- { "..", "", 1 },
- { "foo", "foo", 0 },
- { "foo/bar", "foo/bar", 0 },
- { "foo///bar", "foo/bar", 1 },
- { "foo/.", "foo/", 1 },
- { "foo/./", "foo/", 1 },
- { "foo./", "foo./", 0 },
- { "foo/../bar", "bar", 1 },
- { "foo/../bar/", "bar/", 1 },
- { "foo/bar/..", "foo/", 1 },
- { "foo/bar/../x", "foo/x", 1 },
- { "foo/bar/../x/", "foo/x/", 1 },
- { "foo/..", "", 1 },
- { "foo/../..", "", 1 },
- { "a/b/../../c", "c", 1 },
- { "./a/../b", "b", 1 }
+ static const struct {
+ const char *d1;
+ const char *d2;
+ bool result;
+ } test_array[] = {
+ { "/somedir", "/somedir", true },
+ { "/somedir", "/somedir/d2", true },
+ { "/somedir/d1", "/somedir", false },
};
- int i;
+ unsigned i;
- for (i = 0; i < ARRAY_SIZE (tests); i++)
+ for (i = 0; i < countof(test_array); ++i)
{
- char *test = tests[i].test;
- char *expected_result = tests[i].result;
- int expected_change = tests[i].should_modify;
- run_test (test, expected_result, expected_change);
- }
+ bool res = subdir_p (test_array[i].d1, test_array[i].d2);
- /* Now run all the tests with a leading slash before the test case,
- to prove that the slash is being preserved. */
- for (i = 0; i < ARRAY_SIZE (tests); i++)
- {
- char *test, *expected_result;
- int expected_change = tests[i].should_modify;
+ mu_assert ("test_subdir_p: wrong result",
+ res == test_array[i].result);
+ }
- test = xmalloc (1 + strlen (tests[i].test) + 1);
- sprintf (test, "/%s", tests[i].test);
+ return NULL;
+}
- expected_result = xmalloc (1 + strlen (tests[i].result) + 1);
- sprintf (expected_result, "/%s", tests[i].result);
+const char *
+test_dir_matches_p(void)
+{
+ static struct {
+ const char *dirlist[3];
+ const 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 },
+ { { "*/*d1", "/someotherdir", NULL }, "somedir/d1", true },
+ { { "/somedir/d1", "/someotherdir", NULL }, "d1", false },
+ { { "!COMPLETE", NULL, NULL }, "!COMPLETE", true },
+ { { "*COMPLETE", NULL, NULL }, "!COMPLETE", true },
+ { { "*/!COMPLETE", NULL, NULL }, "foo/!COMPLETE", true },
+ { { "*COMPLETE", NULL, NULL }, "foo/!COMPLETE", false },
+ { { "*/*COMPLETE", NULL, NULL }, "foo/!COMPLETE", true },
+ { { "/dir with spaces", NULL, NULL }, "dir with spaces", true },
+ { { "/dir*with*spaces", NULL, NULL }, "dir with spaces", true },
+ { { "/Tmp/has", NULL, NULL }, "/Tmp/has space", false },
+ { { "/Tmp/has", NULL, NULL }, "/Tmp/has,comma", false },
+ };
+ unsigned i;
- run_test (test, expected_result, expected_change);
+ for (i = 0; i < countof(test_array); ++i)
+ {
+ bool res = dir_matches_p (test_array[i].dirlist, test_array[i].dir);
- xfree (test);
- xfree (expected_result);
+ mu_assert ("test_dir_matches_p: wrong result",
+ res == test_array[i].result);
}
+
+ return NULL;
}
-#endif
+
+#endif /* TESTING */
+
projects / wget / blobdiff