X-Git-Url: http://sjero.net/git/?a=blobdiff_plain;f=src%2Futils.c;h=90b50043c59772253924e49fce99b115f8dbe092;hb=4d7c5e087b2bc82c9f503dff003916d1047903ce;hp=2894d8f30a69b62253b3a5a83358e3f3d6322070;hpb=e65ff5729a876e9f0010ea3d00edb41cc0367a3e;p=wget
diff --git a/src/utils.c b/src/utils.c
index 2894d8f3..90b50043 100644
--- a/src/utils.c
+++ b/src/utils.c
@@ -1,11 +1,11 @@
/* Various utility functions.
- Copyright (C) 2005 Free Software Foundation, Inc.
+ Copyright (C) 1996-2006 Free Software Foundation, Inc.
This file is part of GNU Wget.
GNU Wget is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
GNU Wget is distributed in the hope that it will be useful,
@@ -14,8 +14,7 @@ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
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 .
In addition, as a special exception, the Free Software Foundation
gives permission to link the code of its release of Wget with the
@@ -31,24 +30,17 @@ so, delete this exception statement from your version. */
#include
#include
-#ifdef HAVE_STRING_H
-# include
-#else /* not HAVE_STRING_H */
-# include
-#endif /* not HAVE_STRING_H */
-#include
+#include
+#include
+#ifdef HAVE_SYS_TIME_H
+# include
+#endif
#ifdef HAVE_UNISTD_H
# include
#endif
#ifdef HAVE_MMAP
# include
#endif
-#ifdef HAVE_PWD_H
-# include
-#endif
-#ifdef HAVE_LIMITS_H
-# include
-#endif
#ifdef HAVE_UTIME_H
# include
#endif
@@ -56,16 +48,10 @@ so, delete this exception statement from your version. */
# include
#endif
#include
-#ifdef NeXT
-# include /* for access() */
-#endif
#include
#include
-#ifdef WGET_USE_STDARG
-# include
-#else
-# include
-#endif
+#include
+#include
/* For TIOCGWINSZ and friends: */
#ifdef HAVE_SYS_IOCTL_H
@@ -75,14 +61,9 @@ so, delete this exception statement from your version. */
# include
#endif
-/* Needed for run_with_timeout. */
-#undef USE_SIGNAL_TIMEOUT
-#ifdef HAVE_SIGNAL_H
-# include
-#endif
-#ifdef HAVE_SETJMP_H
-# include
-#endif
+/* Needed for Unix version of run_with_timeout. */
+#include
+#include
#ifndef HAVE_SIGSETJMP
/* If sigsetjmp is a macro, configure won't pick it up. */
@@ -91,22 +72,17 @@ so, delete this exception statement from your version. */
# endif
#endif
-#ifdef HAVE_SIGNAL
-# ifdef HAVE_SIGSETJMP
-# define USE_SIGNAL_TIMEOUT
-# endif
-# ifdef HAVE_SIGBLOCK
-# define USE_SIGNAL_TIMEOUT
-# endif
+#if defined HAVE_SIGSETJMP || defined HAVE_SIGBLOCK
+# define USE_SIGNAL_TIMEOUT
#endif
#include "wget.h"
#include "utils.h"
#include "hash.h"
-#ifndef errno
-extern int errno;
-#endif
+#ifdef TESTING
+#include "test.h"
+#endif
/* Utility function: like xstrdup(), but also lowercases S. */
@@ -126,7 +102,7 @@ xstrdup_lower (const char *s)
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;
@@ -150,7 +126,7 @@ sepstring (const char *s)
{
if (*s == ',')
{
- res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
+ res = xrealloc (res, (i + 2) * sizeof (char *));
res[i] = strdupdelim (p, s);
res[++i] = NULL;
++s;
@@ -162,41 +138,55 @@ sepstring (const char *s)
else
++s;
}
- res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
+ res = xrealloc (res, (i + 2) * sizeof (char *));
res[i] = strdupdelim (p, s);
res[i + 1] = NULL;
return res;
}
-#ifdef WGET_USE_STDARG
-# define VA_START(args, arg1) va_start (args, arg1)
-#else
-# define VA_START(args, ignored) va_start (args)
-#endif
+/* Like sprintf, but 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.
-/* Like sprintf, but allocates a string of sufficient size with malloc
- and returns it. GNU libc has a similar function named asprintf,
- which requires the pointer to the string to be passed. */
+ Internally the function either calls vasprintf or loops around
+ vsnprintf until the correct size is found. Since Wget also ships a
+ fallback implementation of vsnprintf, this should be portable. */
char *
aprintf (const char *fmt, ...)
{
- /* This function is implemented using vsnprintf, which we provide
- for the systems that don't have it. Therefore, it should be 100%
- portable. */
+#if defined HAVE_VASPRINTF && !defined DEBUG_MALLOC
+ /* Use vasprintf. */
+ int ret;
+ va_list args;
+ char *str;
+ va_start (args, fmt);
+ ret = vasprintf (&str, fmt, args);
+ va_end (args);
+ if (ret < 0 && errno == ENOMEM)
+ abort (); /* for consistency with xmalloc/xrealloc */
+ else if (ret < 0)
+ return NULL;
+ return str;
+#else /* not HAVE_VASPRINTF */
+ /* vasprintf is unavailable. snprintf into a small buffer and
+ resize it as necessary. */
int size = 32;
char *str = xmalloc (size);
+ /* #### This code will infloop and eventually abort in xrealloc if
+ passed a FMT that causes snprintf to consistently return -1. */
+
while (1)
{
int n;
va_list args;
- /* See log_vprintf_internal for explanation why it's OK to rely
- on the return value of vsnprintf. */
-
- VA_START (args, fmt);
+ va_start (args, fmt);
n = vsnprintf (str, size, fmt, args);
va_end (args);
@@ -211,7 +201,7 @@ aprintf (const char *fmt, ...)
size <<= 1; /* twice the old size */
str = xrealloc (str, size);
}
- return NULL; /* unreached */
+#endif /* not HAVE_VASPRINTF */
}
/* Concatenate the NULL-terminated list of string arguments into
@@ -231,7 +221,7 @@ concat_strings (const char *str0, ...)
/* Calculate the length of and allocate the resulting string. */
argcount = 0;
- VA_START (args, str0);
+ va_start (args, str0);
for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
{
int len = strlen (next_str);
@@ -245,7 +235,7 @@ concat_strings (const char *str0, ...)
/* Copy the strings into the allocated space. */
argcount = 0;
- VA_START (args, str0);
+ va_start (args, str0);
for (next_str = str0; next_str != NULL; next_str = va_arg (args, char *))
{
int len;
@@ -262,51 +252,38 @@ concat_strings (const char *str0, ...)
return ret;
}
+/* 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 *tm)
+time_str (time_t t)
{
- static char output[15];
- struct tm *ptm;
- time_t secs = tm ? *tm : time (NULL);
-
- if (secs == -1)
- {
- /* In case of error, return the empty string. Maybe we should
- just abort if this happens? */
- *output = '\0';
- return output;
- }
- ptm = localtime (&secs);
- sprintf (output, "%02d:%02d:%02d", ptm->tm_hour, ptm->tm_min, ptm->tm_sec);
- return output;
+ return fmttime(t, "%H:%M:%S");
}
/* Like the above, but include the date: YYYY-MM-DD hh:mm:ss. */
char *
-datetime_str (time_t *tm)
+datetime_str (time_t t)
{
- static char output[20]; /* "YYYY-MM-DD hh:mm:ss" + \0 */
- struct tm *ptm;
- time_t secs = tm ? *tm : time (NULL);
-
- if (secs == -1)
- {
- /* In case of error, return the empty string. Maybe we should
- just abort if this happens? */
- *output = '\0';
- return output;
- }
- 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);
- return output;
+ return fmttime(t, "%Y-%m-%d %H:%M:%S");
}
/* The Windows versions of the following two functions are defined in
@@ -318,7 +295,7 @@ fork_to_background (void)
{
pid_t pid;
/* Whether we arrange our own version of opt.lfilename here. */
- int logfile_changed = 0;
+ bool logfile_changed = false;
if (!opt.lfilename)
{
@@ -327,10 +304,10 @@ fork_to_background (void)
use fopen_excl) or lying to the user about the log file name
(which arises from using unique_name, printing the name, and
using fopen_excl later on.) */
- FILE *new_log_fp = unique_create (DEFAULT_LOGFILE, 0, &opt.lfilename);
+ FILE *new_log_fp = unique_create (DEFAULT_LOGFILE, false, &opt.lfilename);
if (new_log_fp)
{
- logfile_changed = 1;
+ logfile_changed = true;
fclose (new_log_fp);
}
}
@@ -344,7 +321,7 @@ fork_to_background (void)
else if (pid != 0)
{
/* parent, no error */
- printf (_("Continuing in background, pid %d.\n"), (int)pid);
+ printf (_("Continuing in background, pid %d.\n"), (int) pid);
if (logfile_changed)
printf (_("Output will be written to `%s'.\n"), opt.lfilename);
exit (0); /* #### should we use _exit()? */
@@ -358,19 +335,23 @@ fork_to_background (void)
}
#endif /* not WINDOWS */
-/* "Touch" FILE, i.e. make its atime and mtime equal to the time
- specified with TM. */
+/* "Touch" FILE, i.e. make its mtime ("modified time") equal the time
+ specified with TM. The atime ("access time") is set to the current
+ time. */
+
void
touch (const char *file, time_t tm)
{
#ifdef HAVE_STRUCT_UTIMBUF
struct utimbuf times;
- times.actime = times.modtime = tm;
#else
- time_t times[2];
- times[0] = times[1] = tm;
+ struct {
+ time_t actime;
+ time_t modtime;
+ } times;
#endif
-
+ times.modtime = tm;
+ times.actime = time (NULL);
if (utime (file, ×) == -1)
logprintf (LOG_NOTQUIET, "utime(%s): %s\n", file, strerror (errno));
}
@@ -401,7 +382,7 @@ remove_link (const char *file)
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
@@ -414,15 +395,15 @@ file_exists_p (const char *filename)
/* Returns 0 if PATH is a directory, 1 otherwise (any kind of file).
Returns 0 on error. */
-int
+bool
file_non_directory_p (const char *path)
{
struct_stat buf;
/* Use lstat() rather than stat() so that symbolic links pointing to
directories can be identified correctly. */
if (lstat (path, &buf) != 0)
- return 0;
- return S_ISDIR (buf.st_mode) ? 0 : 1;
+ return false;
+ return S_ISDIR (buf.st_mode) ? false : true;
}
/* Return the size of file named by FILENAME, or -1 if it cannot be
@@ -490,7 +471,7 @@ unique_name_1 (const char *prefix)
(and therefore doesn't need changing). */
char *
-unique_name (const char *file, int allow_passthrough)
+unique_name (const char *file, bool allow_passthrough)
{
/* If the FILE itself doesn't exist, return it without
modification. */
@@ -508,15 +489,15 @@ unique_name (const char *file, int allow_passthrough)
opening the file returned by unique_name. */
FILE *
-unique_create (const char *name, int binary, char **opened_name)
+unique_create (const char *name, bool binary, char **opened_name)
{
/* unique file name, based on NAME */
- char *uname = unique_name (name, 0);
+ char *uname = unique_name (name, false);
FILE *fp;
while ((fp = fopen_excl (uname, binary)) == NULL && errno == EEXIST)
{
xfree (uname);
- uname = unique_name (name, 0);
+ uname = unique_name (name, false);
}
if (opened_name && fp != NULL)
{
@@ -544,7 +525,7 @@ unique_create (const char *name, int binary, char **opened_name)
appropriately. */
FILE *
-fopen_excl (const char *fname, int binary)
+fopen_excl (const char *fname, bool binary)
{
int fd;
#ifdef O_EXCL
@@ -617,7 +598,7 @@ make_directory (const char *directory)
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)
@@ -628,7 +609,7 @@ 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);
@@ -636,11 +617,35 @@ file_merge (const char *base, const char *file)
return result;
}
-static int in_acclist PARAMS ((const char *const *, const char *, int));
+/* Like fnmatch, but performs a case-insensitive match. */
+
+int
+fnmatch_nocase (const char *pattern, const char *string, int flags)
+{
+#ifdef FNM_CASEFOLD
+ /* The FNM_CASEFOLD flag started as a GNU extension, but it is now
+ also present on *BSD platforms, and possibly elsewhere. */
+ return fnmatch (pattern, string, flags | FNM_CASEFOLD);
+#else
+ /* Turn PATTERN and STRING to lower case and call fnmatch on them. */
+ char *patcopy = (char *) alloca (strlen (pattern) + 1);
+ char *strcopy = (char *) alloca (strlen (string) + 1);
+ char *p;
+ for (p = patcopy; *pattern; pattern++, p++)
+ *p = TOLOWER (*pattern);
+ *p = '\0';
+ for (p = strcopy; *string; string++, p++)
+ *p = TOLOWER (*string);
+ *p = '\0';
+ return fnmatch (patcopy, strcopy, flags);
+#endif
+}
+
+static bool in_acclist (const char *const *, const char *, bool);
/* Determine whether a file is acceptable to be followed, according to
lists of patterns to accept/reject. */
-int
+bool
acceptable (const char *s)
{
int l = strlen (s);
@@ -652,90 +657,103 @@ acceptable (const char *s)
if (opt.accepts)
{
if (opt.rejects)
- return (in_acclist ((const char *const *)opt.accepts, s, 1)
- && !in_acclist ((const char *const *)opt.rejects, s, 1));
+ return (in_acclist ((const char *const *)opt.accepts, s, true)
+ && !in_acclist ((const char *const *)opt.rejects, s, true));
else
- return in_acclist ((const char *const *)opt.accepts, s, 1);
+ return in_acclist ((const char *const *)opt.accepts, s, true);
}
else if (opt.rejects)
- return !in_acclist ((const char *const *)opt.rejects, s, 1);
- return 1;
+ return !in_acclist ((const char *const *)opt.rejects, s, true);
+ return true;
}
-/* Compare S1 and S2 frontally; S2 must begin with S1. E.g. if S1 is
- `/something', frontcmp() will return 1 only if S2 begins with
- `/something'. Otherwise, 0 is returned. */
-int
-frontcmp (const char *s1, const char *s2)
+/* Check if D2 is a subdirectory of D1. E.g. if D1 is `/something', subdir_p()
+ will return true if and only if D2 begins with `/something/' or is exactly
+ '/something'. */
+bool
+subdir_p (const char *d1, const char *d2)
{
- for (; *s1 && *s2 && (*s1 == *s2); ++s1, ++s2);
- return !*s1;
+ if (!opt.ignore_case)
+ for (; *d1 && *d2 && (*d1 == *d2); ++d1, ++d2)
+ ;
+ else
+ for (; *d1 && *d2 && (TOLOWER (*d1) == TOLOWER (*d2)); ++d1, ++d2)
+ ;
+
+ return *d1 == '\0' && (*d2 == '\0' || *d2 == '/');
}
-/* Iterate through STRLIST, and return the first element that matches
- S, through wildcards or front comparison (as appropriate). */
-static char *
-proclist (char **strlist, const char *s, enum accd flags)
+/* Iterate through DIRLIST (which must be NULL-terminated), and return the
+ first element that matches DIR, through wildcards or front comparison (as
+ appropriate). */
+static bool
+dir_matches_p (char **dirlist, const char *dir)
{
char **x;
+ int (*matcher) (const char *, const char *, int)
+ = opt.ignore_case ? fnmatch_nocase : fnmatch;
- for (x = strlist; *x; x++)
- if (has_wildcards_p (*x))
- {
- if (fnmatch (*x, s, FNM_PATHNAME) == 0)
- break;
- }
- else
- {
- char *p = *x + ((flags & ALLABS) && (**x == '/')); /* Remove '/' */
- if (frontcmp (p, s))
- break;
- }
- return *x;
+ for (x = dirlist; *x; x++)
+ {
+ /* Remove leading '/' */
+ char *p = *x + (**x == '/');
+ if (has_wildcards_p (p))
+ {
+ if (matcher (p, dir, FNM_PATHNAME) == 0)
+ break;
+ }
+ else
+ {
+ if (subdir_p (p, dir))
+ break;
+ }
+ }
+
+ return *x ? true : false;
}
/* Returns whether DIRECTORY is acceptable for download, wrt the
include/exclude lists.
- If FLAGS is ALLABS, the leading `/' is ignored in paths; relative
- and absolute paths may be freely intermixed. */
-int
-accdir (const char *directory, enum accd flags)
+ The leading `/' is ignored in paths; relative and absolute paths
+ may be freely intermixed. */
+
+bool
+accdir (const char *directory)
{
/* Remove starting '/'. */
- if (flags & ALLABS && *directory == '/')
+ if (*directory == '/')
++directory;
if (opt.includes)
{
- if (!proclist (opt.includes, directory, flags))
- return 0;
+ if (!dir_matches_p (opt.includes, directory))
+ return false;
}
if (opt.excludes)
{
- if (proclist (opt.excludes, directory, flags))
- return 0;
+ if (dir_matches_p (opt.excludes, directory))
+ return false;
}
- return 1;
+ return true;
}
-/* Return non-zero if STRING ends with TAIL. For instance:
+/* Return true if STRING ends with TAIL. For instance:
- match_tail ("abc", "bc", 0) -> 1
- match_tail ("abc", "ab", 0) -> 0
- match_tail ("abc", "abc", 0) -> 1
+ match_tail ("abc", "bc", false) -> 1
+ match_tail ("abc", "ab", false) -> 0
+ match_tail ("abc", "abc", false) -> 1
- If FOLD_CASE_P is non-zero, the comparison will be
- case-insensitive. */
+ If FOLD_CASE is true, the comparison will be case-insensitive. */
-int
-match_tail (const char *string, const char *tail, int fold_case_p)
+bool
+match_tail (const char *string, const char *tail, bool fold_case)
{
int i, j;
/* We want this to be fast, so we code two loops, one with
case-folding, one without. */
- if (!fold_case_p)
+ if (!fold_case)
{
for (i = strlen (string), j = strlen (tail); i >= 0 && j >= 0; i--, j--)
if (string[i] != tail[j])
@@ -750,44 +768,47 @@ match_tail (const char *string, const char *tail, int fold_case_p)
/* If the tail was exhausted, the match was succesful. */
if (j == -1)
- return 1;
+ return true;
else
- return 0;
+ return false;
}
/* Checks whether string S matches each element of ACCEPTS. A list
element are matched either with fnmatch() or match_tail(),
according to whether the element contains wildcards or not.
- If the BACKWARD is 0, don't do backward comparison -- just compare
+ If the BACKWARD is false, don't do backward comparison -- just compare
them normally. */
-static int
-in_acclist (const char *const *accepts, const char *s, int backward)
+static bool
+in_acclist (const char *const *accepts, const char *s, bool backward)
{
for (; *accepts; accepts++)
{
if (has_wildcards_p (*accepts))
{
- /* fnmatch returns 0 if the pattern *does* match the
- string. */
- if (fnmatch (*accepts, s, 0) == 0)
- return 1;
+ int res = opt.ignore_case
+ ? fnmatch_nocase (*accepts, s, 0) : fnmatch (*accepts, s, 0);
+ /* fnmatch returns 0 if the pattern *does* match the string. */
+ if (res == 0)
+ return true;
}
else
{
if (backward)
{
- if (match_tail (s, *accepts, 0))
- return 1;
+ if (match_tail (s, *accepts, opt.ignore_case))
+ return true;
}
else
{
- if (!strcmp (s, *accepts))
- return 1;
+ int cmp = opt.ignore_case
+ ? strcasecmp (s, *accepts) : strcmp (s, *accepts);
+ if (cmp == 0)
+ return true;
}
}
}
- return 0;
+ return false;
}
/* Return the location of STR's suffix (file extension). Examples:
@@ -809,20 +830,21 @@ suffix (const char *str)
return NULL;
}
-/* Return non-zero if S contains globbing wildcards (`*', `?', `[' or
+/* Return true if S contains globbing wildcards (`*', `?', `[' or
`]'). */
-int
+bool
has_wildcards_p (const char *s)
{
for (; *s; s++)
if (*s == '*' || *s == '?' || *s == '[' || *s == ']')
- return 1;
- return 0;
+ return true;
+ return false;
}
-/* Return non-zero if FNAME ends with a typical HTML suffix. The
- following (case-insensitive) suffixes are presumed to be HTML files:
+/* Return true if FNAME ends with a typical HTML suffix. The
+ following (case-insensitive) suffixes are presumed to be HTML
+ files:
html
htm
@@ -830,20 +852,20 @@ has_wildcards_p (const char *s)
#### CAVEAT. This is not necessarily a good indication that FNAME
refers to a file that contains HTML! */
-int
+bool
has_html_suffix_p (const char *fname)
{
char *suf;
if ((suf = suffix (fname)) == NULL)
- return 0;
+ return false;
if (!strcasecmp (suf, "html"))
- return 1;
+ return true;
if (!strcasecmp (suf, "htm"))
- return 1;
+ return true;
if (suf[0] && !strcasecmp (suf + 1, "html"))
- return 1;
- return 0;
+ return true;
+ return false;
}
/* Read a line from FP and return the pointer to freshly allocated
@@ -863,7 +885,7 @@ read_whole_line (FILE *fp)
{
int length = 0;
int bufsize = 82;
- char *line = (char *)xmalloc (bufsize);
+ char *line = xmalloc (bufsize);
while (fgets (line + length, bufsize - length, fp))
{
@@ -918,14 +940,14 @@ 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. */
}
@@ -937,7 +959,7 @@ read_file (const char *file)
#ifdef HAVE_MMAP
{
- struct_stat buf;
+ struct_fstat buf;
if (fstat (fd, &buf) < 0)
goto mmap_lose;
fm->length = buf.st_size;
@@ -1071,15 +1093,41 @@ merge_vecs (char **v1, char **v2)
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;
}
+
+/* Append a freshly allocated copy of STR to VEC. If VEC is NULL, it
+ is allocated as needed. Return the new value of the vector. */
+
+char **
+vec_append (char **vec, const char *str)
+{
+ int cnt; /* count of vector elements, including
+ the one we're about to append */
+ if (vec != NULL)
+ {
+ for (cnt = 0; vec[cnt]; cnt++)
+ ;
+ ++cnt;
+ }
+ else
+ cnt = 1;
+ /* Reallocate the array to fit the new element and the NULL. */
+ vec = xrealloc (vec, (cnt + 1) * sizeof (char *));
+ /* Append a copy of STR to the vector. */
+ vec[cnt - 1] = xstrdup (str);
+ vec[cnt] = NULL;
+ return vec;
+}
/* Sometimes it's useful to create "sets" of strings, i.e. special
hash tables where you want to store strings as keys and merely
@@ -1110,128 +1158,143 @@ string_set_contains (struct hash_table *ht, const char *s)
return hash_table_contains (ht, s);
}
-static int
-string_set_to_array_mapper (void *key, void *value_ignored, void *arg)
-{
- char ***arrayptr = (char ***) arg;
- *(*arrayptr)++ = (char *) key;
- return 0;
-}
-
/* 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)
{
- hash_table_map (ht, string_set_to_array_mapper, &array);
+ hash_table_iterator iter;
+ for (hash_table_iterate (ht, &iter); hash_table_iter_next (&iter); )
+ *array++ = iter.key;
}
-static int
-string_set_free_mapper (void *key, void *value_ignored, void *arg_ignored)
-{
- xfree (key);
- return 0;
-}
+/* 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);
+ }
}
-
-/* Add thousand separators to a number already in string form. Used
- by with_thousand_seps and with_thousand_seps_large. */
-
-static char *
-add_thousand_seps (const char *repr)
-{
- static char outbuf[48];
- int i, i1, mod;
- char *outptr;
- const char *inptr;
+/* 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.
- /* Reset the pointers. */
- outptr = outbuf;
- inptr = repr;
+ 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. */
- /* Ignore the sign for the purpose of adding thousand
- separators. */
- if (*inptr == '-')
- {
- *outptr++ = '-';
- ++inptr;
- }
- /* How many digits before the first separator? */
- mod = strlen (inptr) % 3;
- /* Insert them. */
- for (i = 0; i < mod; i++)
- *outptr++ = inptr[i];
- /* Now insert the rest of them, putting separator before every
- third digit. */
- for (i1 = i, i = 0; inptr[i1]; i++, i1++)
+static void
+get_grouping_data (const char **sep, const char **grouping)
+{
+ static const char *cached_sep;
+ static const char *cached_grouping;
+ static bool initialized;
+ if (!initialized)
{
- if (i % 3 == 0 && i1 != 0)
- *outptr++ = ',';
- *outptr++ = inptr[i1];
+ /* Get the grouping info from the locale. */
+ struct lconv *lconv = localeconv ();
+ cached_sep = lconv->thousands_sep;
+ cached_grouping = lconv->grouping;
+ if (!*cached_sep)
+ {
+ /* Many locales (such as "C" or "hr_HR") don't specify
+ grouping, which we still want to use it for legibility.
+ In those locales set the sep char to ',', unless that
+ character is used for decimal point, in which case set it
+ to ".". */
+ if (*lconv->decimal_point != ',')
+ cached_sep = ",";
+ else
+ cached_sep = ".";
+ cached_grouping = "\x03";
+ }
+ initialized = true;
}
- /* Zero-terminate the string. */
- *outptr = '\0';
- return outbuf;
+ *sep = cached_sep;
+ *grouping = cached_grouping;
}
-/* Return a static pointer to the number printed with thousand
- separators inserted at the right places. */
+/* 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.
-char *
-with_thousand_seps (wgint l)
+ 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 add_thousand_seps (inbuf);
-}
+ static char outbuf[48];
+ char *p = outbuf + sizeof outbuf;
-/* Write a string representation of LARGE_INT NUMBER into the provided
- buffer.
+ /* Info received from locale */
+ const char *grouping, *sep;
+ int seplen;
- It would be dangerous to use sprintf, because the code wouldn't
- work on a machine with gcc-provided long long support, but without
- libc support for "%lld". However, such old systems platforms
- typically lack snprintf and will end up using our version, which
- does support "%lld" whereever long longs are available. */
+ /* State information */
+ int i = 0, groupsize;
+ const char *atgroup;
-static void
-large_int_to_string (char *buffer, int bufsize, LARGE_INT number)
-{
- snprintf (buffer, bufsize, LARGE_INT_FMT, number);
-}
+ bool negative = n < 0;
-/* The same as with_thousand_seps, but works on LARGE_INT. */
+ /* Initialize grouping data. */
+ get_grouping_data (&sep, &grouping);
+ seplen = strlen (sep);
+ atgroup = grouping;
+ groupsize = *atgroup++;
-char *
-with_thousand_seps_large (LARGE_INT l)
-{
- char inbuf[48];
- large_int_to_string (inbuf, sizeof (inbuf), l);
- return add_thousand_seps (inbuf);
+ /* This would overflow on WGINT_MIN, but printing negative numbers
+ is not an important goal of this fuinction. */
+ if (negative)
+ n = -n;
+
+ /* Write the number into the buffer, backwards, inserting the
+ separators as necessary. */
+ *--p = '\0';
+ while (1)
+ {
+ *--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++;
+ }
+ }
+ if (negative)
+ *--p = '-';
+
+ return p;
}
/* N, a byte quantity, is converted to a human-readable abberviated
@@ -1245,14 +1308,13 @@ with_thousand_seps_large (LARGE_INT l)
usually improves readability."
This intentionally uses kilobyte (KB), megabyte (MB), etc. in their
- original computer science meaning of "multiples of 1024".
- Multiples of 1000 would be useless since Wget already adds thousand
- separators for legibility. We don't use the "*bibyte" names
- invented in 1998, and seldom used in practice. Wikipedia's entry
- on kilobyte discusses this in some detail. */
+ original computer-related meaning of "powers of 1024". 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 *
-human_readable (wgint n)
+human_readable (HR_NUMTYPE n)
{
/* These suffixes are compatible with those of GNU `ls -lh'. */
static char powers[] =
@@ -1282,19 +1344,16 @@ human_readable (wgint n)
/* At each iteration N is greater than the *subsequent* power.
That way N/1024.0 produces a decimal number in the units of
*this* power. */
- if ((n >> 10) < 1024 || i == countof (powers) - 1)
+ if ((n / 1024) < 1024 || i == countof (powers) - 1)
{
- /* Must cast to long first because MS VC can't directly cast
- __int64 to double. (This is safe because N is known to
- be <2**20.) */
- double val = (double) (long) n / 1024.0;
+ double val = n / 1024.0;
/* Print values smaller than 10 with one decimal digits, and
others without any decimals. */
snprintf (buf, sizeof (buf), "%.*f%c",
val < 10 ? 1 : 0, val, powers[i]);
return buf;
}
- n >>= 10;
+ n /= 1024;
}
return NULL; /* unreached */
}
@@ -1343,24 +1402,6 @@ numdigit (wgint number)
#define DIGITS_18(mask) PR (mask), n %= (mask), DIGITS_17 ((mask) / 10)
#define DIGITS_19(mask) PR (mask), n %= (mask), DIGITS_18 ((mask) / 10)
-/* SPRINTF_WGINT is used by number_to_string to handle pathological
- cases and to portably support strange sizes of wgint. Ideally this
- would just use "%j" and intmax_t, but many systems don't support
- it, so it's used only if nothing else works. */
-#if SIZEOF_LONG >= SIZEOF_WGINT
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%ld", (long) (n))
-#else
-# if SIZEOF_LONG_LONG >= SIZEOF_WGINT
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%lld", (long long) (n))
-# else
-# ifdef WINDOWS
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%I64", (__int64) (n))
-# else
-# define SPRINTF_WGINT(buf, n) sprintf (buf, "%j", (intmax_t) (n))
-# endif
-# endif
-#endif
-
/* Shorthand for casting to wgint. */
#define W wgint
@@ -1370,15 +1411,15 @@ numdigit (wgint number)
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 when converting integers to strings.
+ 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'. */
@@ -1389,21 +1430,29 @@ number_to_string (char *buffer, wgint number)
char *p = buffer;
wgint n = number;
+ int last_digit_char = 0;
+
#if (SIZEOF_WGINT != 4) && (SIZEOF_WGINT != 8)
- /* We are running in a strange or misconfigured environment. Let
- sprintf cope with it. */
- SPRINTF_WGINT (buffer, n);
- p += strlen (buffer);
+ /* 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 would overflow. Have sprintf deal with this. */
- SPRINTF_WGINT (buffer, n);
- p += strlen (buffer);
- return p;
+ /* n = -n would overflow because -n would evaluate to a
+ wgint value larger than WGINT_MAX. Need to make n
+ smaller and handle the last digit separately. */
+ int last_digit = n % 10;
+ /* The sign of n%10 is implementation-defined. */
+ if (last_digit < 0)
+ last_digit_char = '0' - last_digit;
+ else
+ last_digit_char = '0' + last_digit;
+ /* After n is made smaller, -n will not overflow. */
+ n /= 10;
}
*p++ = '-';
@@ -1427,10 +1476,10 @@ number_to_string (char *buffer, wgint number)
/* wgint is 32 bits wide: no number has more than 10 digits. */
else DIGITS_10 (1000000000);
#else
- /* wgint is 64 bits wide: handle numbers with more than 9 decimal
- digits. Constants are constructed by compile-time multiplication
- to avoid dealing with different notations for 64-bit constants
- (nnnL, nnnLL, and nnnI64, depending on the compiler). */
+ /* wgint is 64 bits wide: handle numbers with 9-19 decimal digits.
+ Constants are constructed by compile-time multiplication to avoid
+ dealing with different notations for 64-bit constants
+ (nL/nLL/nI64, depending on the compiler and architecture). */
else if (n < 10*(W)1000000000) DIGITS_10 (1000000000);
else if (n < 100*(W)1000000000) DIGITS_11 (10*(W)1000000000);
else if (n < 1000*(W)1000000000) DIGITS_12 (100*(W)1000000000);
@@ -1443,6 +1492,9 @@ number_to_string (char *buffer, wgint number)
else DIGITS_19 (1000000000*(W)1000000000);
#endif
+ if (last_digit_char)
+ *p++ = last_digit_char;
+
*p = '\0';
#endif /* (SIZEOF_WGINT == 4) || (SIZEOF_WGINT == 8) */
@@ -1451,6 +1503,7 @@ number_to_string (char *buffer, wgint number)
#undef PR
#undef W
+#undef SPRINTF_WGINT
#undef DIGITS_1
#undef DIGITS_2
#undef DIGITS_3
@@ -1515,352 +1568,6 @@ number_to_static_string (wgint number)
return buf;
}
-/* Support for timers. */
-
-#undef TIMER_WINDOWS
-#undef TIMER_GETTIMEOFDAY
-#undef TIMER_TIME
-
-/* Depending on the OS and availability of gettimeofday(), one and
- only one of the above constants will be defined. Virtually all
- modern Unix systems will define TIMER_GETTIMEOFDAY; Windows will
- use TIMER_WINDOWS. TIMER_TIME is a catch-all method for
- non-Windows systems without gettimeofday. */
-
-#ifdef WINDOWS
-# define TIMER_WINDOWS
-#else /* not WINDOWS */
-# ifdef HAVE_GETTIMEOFDAY
-# define TIMER_GETTIMEOFDAY
-# else
-# define TIMER_TIME
-# endif
-#endif /* not WINDOWS */
-
-#ifdef TIMER_GETTIMEOFDAY
-typedef struct timeval wget_sys_time;
-#endif
-
-#ifdef TIMER_TIME
-typedef time_t wget_sys_time;
-#endif
-
-#ifdef TIMER_WINDOWS
-typedef union {
- DWORD lores; /* In case GetTickCount is used */
- LARGE_INTEGER hires; /* In case high-resolution timer is used */
-} wget_sys_time;
-#endif
-
-struct wget_timer {
- /* Whether the start time has been initialized. */
- int initialized;
-
- /* The starting point in time which, subtracted from the current
- time, yields elapsed time. */
- wget_sys_time start;
-
- /* The most recent elapsed time, calculated by wtimer_update().
- Measured in milliseconds. */
- double elapsed_last;
-
- /* Approximately, the time elapsed between the true start of the
- measurement and the time represented by START. */
- double elapsed_pre_start;
-};
-
-#ifdef TIMER_WINDOWS
-
-/* Whether high-resolution timers are used. Set by wtimer_initialize_once
- the first time wtimer_allocate is called. */
-static int using_hires_timers;
-
-/* Frequency of high-resolution timers -- number of updates per
- millisecond. Calculated the first time wtimer_allocate is called
- provided that high-resolution timers are available. */
-static double hires_millisec_freq;
-
-/* The first time a timer is created, determine whether to use
- high-resolution timers. */
-
-static void
-wtimer_initialize_once (void)
-{
- static int init_done;
- if (!init_done)
- {
- LARGE_INTEGER freq;
- init_done = 1;
- freq.QuadPart = 0;
- QueryPerformanceFrequency (&freq);
- if (freq.QuadPart != 0)
- {
- using_hires_timers = 1;
- hires_millisec_freq = (double) freq.QuadPart / 1000.0;
- }
- }
-}
-#endif /* TIMER_WINDOWS */
-
-/* Allocate a timer. Calling wtimer_read on the timer will return
- zero. It is not legal to call wtimer_update with a freshly
- allocated timer -- use wtimer_reset first. */
-
-struct wget_timer *
-wtimer_allocate (void)
-{
- struct wget_timer *wt = xnew (struct wget_timer);
- xzero (*wt);
-
-#ifdef TIMER_WINDOWS
- wtimer_initialize_once ();
-#endif
-
- return wt;
-}
-
-/* Allocate a new timer and reset it. Return the new timer. */
-
-struct wget_timer *
-wtimer_new (void)
-{
- struct wget_timer *wt = wtimer_allocate ();
- wtimer_reset (wt);
- return wt;
-}
-
-/* Free the resources associated with the timer. Its further use is
- prohibited. */
-
-void
-wtimer_delete (struct wget_timer *wt)
-{
- xfree (wt);
-}
-
-/* Store system time to WST. */
-
-static void
-wtimer_sys_set (wget_sys_time *wst)
-{
-#ifdef TIMER_GETTIMEOFDAY
- gettimeofday (wst, NULL);
-#endif
-
-#ifdef TIMER_TIME
- time (wst);
-#endif
-
-#ifdef TIMER_WINDOWS
- if (using_hires_timers)
- {
- QueryPerformanceCounter (&wst->hires);
- }
- else
- {
- /* Where hires counters are not available, use GetTickCount rather
- GetSystemTime, because it is unaffected by clock skew and simpler
- to use. Note that overflows don't affect us because we never use
- absolute values of the ticker, only the differences. */
- wst->lores = GetTickCount ();
- }
-#endif
-}
-
-/* Reset timer WT. This establishes the starting point from which
- wtimer_read() will return the number of elapsed milliseconds.
- It is allowed to reset a previously used timer. */
-
-void
-wtimer_reset (struct wget_timer *wt)
-{
- /* Set the start time to the current time. */
- wtimer_sys_set (&wt->start);
- wt->elapsed_last = 0;
- wt->elapsed_pre_start = 0;
- wt->initialized = 1;
-}
-
-static double
-wtimer_sys_diff (wget_sys_time *wst1, wget_sys_time *wst2)
-{
-#ifdef TIMER_GETTIMEOFDAY
- return ((double)(wst1->tv_sec - wst2->tv_sec) * 1000
- + (double)(wst1->tv_usec - wst2->tv_usec) / 1000);
-#endif
-
-#ifdef TIMER_TIME
- return 1000 * (*wst1 - *wst2);
-#endif
-
-#ifdef WINDOWS
- if (using_hires_timers)
- return (wst1->hires.QuadPart - wst2->hires.QuadPart) / hires_millisec_freq;
- else
- return wst1->lores - wst2->lores;
-#endif
-}
-
-/* Update the timer's elapsed interval. This function causes the
- timer to call gettimeofday (or time(), etc.) to update its idea of
- current time. To get the elapsed interval in milliseconds, use
- wtimer_read.
-
- This function handles clock skew, i.e. time that moves backwards is
- ignored. */
-
-void
-wtimer_update (struct wget_timer *wt)
-{
- wget_sys_time now;
- double elapsed;
-
- assert (wt->initialized != 0);
-
- wtimer_sys_set (&now);
- elapsed = wt->elapsed_pre_start + wtimer_sys_diff (&now, &wt->start);
-
- /* Ideally we'd just return the difference between NOW and
- wt->start. However, the system timer can be set back, and we
- could return a value smaller than when we were last called, even
- a negative value. Both of these would confuse the callers, which
- expect us to return monotonically nondecreasing values.
-
- Therefore: if ELAPSED is smaller than its previous known value,
- we reset wt->start to the current time and effectively start
- measuring from this point. But since we don't want the elapsed
- value to start from zero, we set elapsed_pre_start to the last
- elapsed time and increment all future calculations by that
- amount. */
-
- if (elapsed < wt->elapsed_last)
- {
- wt->start = now;
- wt->elapsed_pre_start = wt->elapsed_last;
- elapsed = wt->elapsed_last;
- }
-
- wt->elapsed_last = elapsed;
-}
-
-/* Return the elapsed time in milliseconds between the last call to
- wtimer_reset and the last call to wtimer_update.
-
- A typical use of the timer interface would be:
-
- struct wtimer *timer = wtimer_new ();
- ... do something that takes a while ...
- wtimer_update ();
- double msecs = wtimer_read (); */
-
-double
-wtimer_read (const struct wget_timer *wt)
-{
- return wt->elapsed_last;
-}
-
-/* Return the assessed granularity of the timer implementation, in
- milliseconds. This is used by code that tries to substitute a
- better value for timers that have returned zero. */
-
-double
-wtimer_granularity (void)
-{
-#ifdef TIMER_GETTIMEOFDAY
- /* Granularity of gettimeofday varies wildly between architectures.
- However, it appears that on modern machines it tends to be better
- than 1ms. Assume 100 usecs. (Perhaps the configure process
- could actually measure this?) */
- return 0.1;
-#endif
-
-#ifdef TIMER_TIME
- return 1000;
-#endif
-
-#ifdef TIMER_WINDOWS
- if (using_hires_timers)
- return 1.0 / hires_millisec_freq;
- else
- return 10; /* according to MSDN */
-#endif
-}
-
-/* This should probably be at a better place, but it doesn't really
- fit into html-parse.c. */
-
-/* The function returns the pointer to the malloc-ed quoted version of
- string s. It will recognize and quote numeric and special graphic
- entities, as per RFC1866:
-
- `&' -> `&'
- `<' -> `<'
- `>' -> `>'
- `"' -> `"'
- SP -> ` '
-
- No other entities are recognized or replaced. */
-char *
-html_quote_string (const char *s)
-{
- const char *b = s;
- char *p, *res;
- int i;
-
- /* Pass through the string, and count the new size. */
- for (i = 0; *s; s++, i++)
- {
- if (*s == '&')
- i += 4; /* `amp;' */
- else if (*s == '<' || *s == '>')
- i += 3; /* `lt;' and `gt;' */
- else if (*s == '\"')
- i += 5; /* `quot;' */
- else if (*s == ' ')
- i += 4; /* #32; */
- }
- res = (char *)xmalloc (i + 1);
- s = b;
- for (p = res; *s; s++)
- {
- switch (*s)
- {
- case '&':
- *p++ = '&';
- *p++ = 'a';
- *p++ = 'm';
- *p++ = 'p';
- *p++ = ';';
- break;
- case '<': case '>':
- *p++ = '&';
- *p++ = (*s == '<' ? 'l' : 'g');
- *p++ = 't';
- *p++ = ';';
- break;
- case '\"':
- *p++ = '&';
- *p++ = 'q';
- *p++ = 'u';
- *p++ = 'o';
- *p++ = 't';
- *p++ = ';';
- break;
- case ' ':
- *p++ = '&';
- *p++ = '#';
- *p++ = '3';
- *p++ = '2';
- *p++ = ';';
- break;
- default:
- *p++ = *s;
- }
- }
- *p = '\0';
- return res;
-}
-
/* Determine the width of the terminal we're running on. If that's
not possible, return 0. */
@@ -1881,79 +1588,83 @@ determine_screen_width (void)
return 0; /* most likely ENOTTY */
return wsz.ws_col;
-#else /* not TIOCGWINSZ */
-# ifdef WINDOWS
+#elif defined(WINDOWS)
CONSOLE_SCREEN_BUFFER_INFO csbi;
if (!GetConsoleScreenBufferInfo (GetStdHandle (STD_ERROR_HANDLE), &csbi))
return 0;
return csbi.dwSize.X;
-# else /* neither WINDOWS nor TIOCGWINSZ */
+#else /* neither TIOCGWINSZ nor WINDOWS */
return 0;
-#endif /* neither WINDOWS nor TIOCGWINSZ */
-#endif /* not TIOCGWINSZ */
+#endif /* neither TIOCGWINSZ nor WINDOWS */
}
+
+/* Whether the rnd system (either rand or [dl]rand48) has been
+ seeded. */
+static int rnd_seeded;
/* Return a random number between 0 and MAX-1, inclusive.
- If MAX is greater than the value of RAND_MAX+1 on the system, the
- returned value will be in the range [0, RAND_MAX]. This may be
- fixed in a future release.
-
+ If the system does not support lrand48 and MAX is greater than the
+ value of RAND_MAX+1 on the system, the returned value will be in
+ the range [0, RAND_MAX]. This may be fixed in a future release.
The random number generator is seeded automatically the first time
it is called.
- This uses rand() for portability. It has been suggested that
- random() offers better randomness, but this is not required for
- Wget, so I chose to go for simplicity and use rand
- unconditionally.
-
- DO NOT use this for cryptographic purposes. It is only meant to be
- used in situations where quality of the random numbers returned
- doesn't really matter. */
+ This uses lrand48 where available, rand elsewhere. DO NOT use it
+ for cryptography. It is only meant to be used in situations where
+ quality of the random numbers returned doesn't really matter. */
int
random_number (int max)
{
- static int seeded;
+#ifdef HAVE_DRAND48
+ if (!rnd_seeded)
+ {
+ srand48 ((long) time (NULL) ^ (long) getpid ());
+ rnd_seeded = 1;
+ }
+ return lrand48 () % max;
+#else /* not HAVE_DRAND48 */
+
double bounded;
int rnd;
-
- if (!seeded)
+ if (!rnd_seeded)
{
- srand (time (NULL));
- seeded = 1;
+ srand ((unsigned) time (NULL) ^ (unsigned) getpid ());
+ rnd_seeded = 1;
}
rnd = rand ();
- /* On systems that don't define RAND_MAX, assume it to be 2**15 - 1,
- and enforce that assumption by masking other bits. */
-#ifndef RAND_MAX
-# define RAND_MAX 32767
- rnd &= RAND_MAX;
-#endif
+ /* Like rand() % max, but uses the high-order bits for better
+ randomness on architectures where rand() is implemented using a
+ simple congruential generator. */
- /* This is equivalent to rand() % max, but uses the high-order bits
- for better randomness on architecture where rand() is implemented
- using a simple congruential generator. */
+ bounded = (double) max * rnd / (RAND_MAX + 1.0);
+ return (int) bounded;
- bounded = (double)max * rnd / (RAND_MAX + 1.0);
- return (int)bounded;
+#endif /* not HAVE_DRAND48 */
}
/* Return a random uniformly distributed floating point number in the
- [0, 1) range. The precision of returned numbers is 9 digits.
-
- Modify this to use erand48() where available! */
+ [0, 1) range. Uses drand48 where available, and a really lame
+ kludge elsewhere. */
double
random_float (void)
{
- /* We can't rely on any specific value of RAND_MAX, but I'm pretty
- sure it's greater than 1000. */
- int rnd1 = random_number (1000);
- int rnd2 = random_number (1000);
- int rnd3 = random_number (1000);
- return rnd1 / 1000.0 + rnd2 / 1000000.0 + rnd3 / 1000000000.0;
+#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 */
}
/* Implementation of run_with_timeout, a generic timeout-forcing
@@ -1965,7 +1676,7 @@ random_float (void)
static sigjmp_buf run_with_timeout_env;
-static RETSIGTYPE
+static void
abort_run_with_timeout (int sig)
{
assert (sig == SIGALRM);
@@ -1976,7 +1687,7 @@ abort_run_with_timeout (int sig)
static jmp_buf run_with_timeout_env;
-static RETSIGTYPE
+static void
abort_run_with_timeout (int sig)
{
assert (sig == SIGALRM);
@@ -2042,8 +1753,8 @@ alarm_cancel (void)
}
/* Call FUN(ARG), but don't allow it to run for more than TIMEOUT
- seconds. Returns non-zero if the function was interrupted with a
- timeout, zero otherwise.
+ seconds. Returns true if the function was interrupted with a
+ timeout, false otherwise.
This works by setting up SIGALRM to be delivered in TIMEOUT seconds
using setitimer() or alarm(). The timeout is enforced by
@@ -2068,7 +1779,7 @@ alarm_cancel (void)
are normally freed prior to exit from the functions, they will be
lost in case of timeout. */
-int
+bool
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
int saved_errno;
@@ -2076,7 +1787,7 @@ run_with_timeout (double timeout, void (*fun) (void *), void *arg)
if (timeout == 0)
{
fun (arg);
- return 0;
+ return false;
}
signal (SIGALRM, abort_run_with_timeout);
@@ -2084,7 +1795,7 @@ run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
/* Longjumped out of FUN with a timeout. */
signal (SIGALRM, SIG_DFL);
- return 1;
+ return true;
}
alarm_set (timeout);
fun (arg);
@@ -2095,7 +1806,7 @@ run_with_timeout (double timeout, void (*fun) (void *), void *arg)
signal (SIGALRM, SIG_DFL);
errno = saved_errno;
- return 0;
+ return false;
}
#else /* not USE_SIGNAL_TIMEOUT */
@@ -2109,7 +1820,7 @@ int
run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
fun (arg);
- return 0;
+ return false;
}
#endif /* not WINDOWS */
#endif /* not USE_SIGNAL_TIMEOUT */
@@ -2135,8 +1846,7 @@ xsleep (double seconds)
/* If nanosleep has been interrupted by a signal, adjust the
sleeping period and return to sleep. */
sleep = remaining;
-#else /* not HAVE_NANOSLEEP */
-#ifdef HAVE_USLEEP
+#elif defined(HAVE_USLEEP)
/* If usleep is available, use it in preference to select. */
if (seconds >= 1)
{
@@ -2147,13 +1857,11 @@ xsleep (double seconds)
seconds -= (long) seconds;
}
usleep (seconds * 1000000);
-#else /* not HAVE_USLEEP */
-#ifdef HAVE_SELECT
- /* Note that, although Windows supports select, this sleeping
- strategy doesn't work there because Winsock's select doesn't
- implement timeout when it is passed NULL pointers for all fd
- sets. (But it does work under Cygwin, which implements its own
- select.) */
+#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);
@@ -2162,11 +1870,312 @@ xsleep (double seconds)
interrupted by a signal. But without knowing how long we've
actually slept, we can't return to sleep. Using gettimeofday to
track sleeps is slow and unreliable due to clock skew. */
-#else /* not HAVE_SELECT */
- sleep (seconds);
-#endif /* not HAVE_SELECT */
-#endif /* not HAVE_USLEEP */
-#endif /* not HAVE_NANOSLEEP */
+#endif
}
#endif /* not WINDOWS */
+
+/* Encode the octets in DATA of length LENGTH to base64 format,
+ storing the result to DEST. The output will be zero-terminated,
+ and must point to a writable buffer of at least
+ 1+BASE64_LENGTH(length) bytes. The function returns the length of
+ the resulting base64 data, not counting the terminating zero.
+
+ This implementation does not emit newlines after 76 characters of
+ base64 data. */
+
+int
+base64_encode (const void *data, int length, char *dest)
+{
+ /* Conversion table. */
+ static const char tbl[64] = {
+ 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P',
+ 'Q','R','S','T','U','V','W','X','Y','Z','a','b','c','d','e','f',
+ 'g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v',
+ 'w','x','y','z','0','1','2','3','4','5','6','7','8','9','+','/'
+ };
+ /* Access bytes in DATA as unsigned char, otherwise the shifts below
+ don't work for data with MSB set. */
+ const unsigned char *s = data;
+ /* Theoretical ANSI violation when length < 3. */
+ const unsigned char *end = (const unsigned char *) data + length - 2;
+ char *p = dest;
+
+ /* Transform the 3x8 bits to 4x6 bits, as required by base64. */
+ for (; s < end; s += 3)
+ {
+ *p++ = tbl[s[0] >> 2];
+ *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
+ *p++ = tbl[((s[1] & 0xf) << 2) + (s[2] >> 6)];
+ *p++ = tbl[s[2] & 0x3f];
+ }
+
+ /* Pad the result if necessary... */
+ switch (length % 3)
+ {
+ case 1:
+ *p++ = tbl[s[0] >> 2];
+ *p++ = tbl[(s[0] & 3) << 4];
+ *p++ = '=';
+ *p++ = '=';
+ break;
+ case 2:
+ *p++ = tbl[s[0] >> 2];
+ *p++ = tbl[((s[0] & 3) << 4) + (s[1] >> 4)];
+ *p++ = tbl[((s[1] & 0xf) << 2)];
+ *p++ = '=';
+ break;
+ }
+ /* ...and zero-terminate it. */
+ *p = '\0';
+
+ return p - dest;
+}
+
+/* Store in C the next non-whitespace character from the string, or \0
+ when end of string is reached. */
+#define NEXT_CHAR(c, p) do { \
+ c = (unsigned char) *p++; \
+} while (ISSPACE (c))
+
+#define IS_ASCII(c) (((c) & 0x80) == 0)
+
+/* Decode data from BASE64 (a null-terminated string) into memory
+ pointed to by DEST. DEST is assumed to be large enough to
+ accomodate the decoded data, which is guaranteed to be no more than
+ 3/4*strlen(base64).
+
+ Since DEST is assumed to contain binary data, it is not
+ NUL-terminated. The function returns the length of the data
+ written to TO. -1 is returned in case of error caused by malformed
+ base64 input.
+
+ This function originates from Free Recode. */
+
+int
+base64_decode (const char *base64, void *dest)
+{
+ /* Table of base64 values for first 128 characters. Note that this
+ assumes ASCII (but so does Wget in other places). */
+ static const signed char base64_char_to_value[128] =
+ {
+ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* 0- 9 */
+ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* 10- 19 */
+ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* 20- 29 */
+ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, /* 30- 39 */
+ -1, -1, -1, 62, -1, -1, -1, 63, 52, 53, /* 40- 49 */
+ 54, 55, 56, 57, 58, 59, 60, 61, -1, -1, /* 50- 59 */
+ -1, -1, -1, -1, -1, 0, 1, 2, 3, 4, /* 60- 69 */
+ 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, /* 70- 79 */
+ 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, /* 80- 89 */
+ 25, -1, -1, -1, -1, -1, -1, 26, 27, 28, /* 90- 99 */
+ 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, /* 100-109 */
+ 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, /* 110-119 */
+ 49, 50, 51, -1, -1, -1, -1, -1 /* 120-127 */
+ };
+#define BASE64_CHAR_TO_VALUE(c) ((int) base64_char_to_value[c])
+#define IS_BASE64(c) ((IS_ASCII (c) && BASE64_CHAR_TO_VALUE (c) >= 0) || c == '=')
+
+ const char *p = base64;
+ char *q = dest;
+
+ while (1)
+ {
+ unsigned char c;
+ unsigned long value;
+
+ /* Process first byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ break;
+ if (c == '=' || !IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+ value = BASE64_CHAR_TO_VALUE (c) << 18;
+
+ /* Process second byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (c == '=' || !IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+ value |= BASE64_CHAR_TO_VALUE (c) << 12;
+ *q++ = value >> 16;
+
+ /* Process third byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (!IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+
+ if (c == '=')
+ {
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (c != '=')
+ return -1; /* padding `=' expected but not found */
+ continue;
+ }
+
+ value |= BASE64_CHAR_TO_VALUE (c) << 6;
+ *q++ = 0xff & value >> 8;
+
+ /* Process fourth byte of a quadruplet. */
+ NEXT_CHAR (c, p);
+ if (!c)
+ return -1; /* premature EOF while decoding base64 */
+ if (c == '=')
+ continue;
+ if (!IS_BASE64 (c))
+ return -1; /* illegal char while decoding base64 */
+
+ value |= BASE64_CHAR_TO_VALUE (c);
+ *q++ = 0xff & value;
+ }
+#undef IS_BASE64
+#undef BASE64_CHAR_TO_VALUE
+
+ return q - (char *) dest;
+}
+
+#undef IS_ASCII
+#undef NEXT_CHAR
+
+/* Simple merge sort for use by stable_sort. Implementation courtesy
+ Zeljko Vrba with additional debugging by Nenad Barbutov. */
+
+static void
+mergesort_internal (void *base, void *temp, size_t size, size_t from, size_t to,
+ int (*cmpfun) (const void *, const void *))
+{
+#define ELT(array, pos) ((char *)(array) + (pos) * size)
+ if (from < to)
+ {
+ size_t i, j, k;
+ size_t mid = (to + from) / 2;
+ mergesort_internal (base, temp, size, from, mid, cmpfun);
+ mergesort_internal (base, temp, size, mid + 1, to, cmpfun);
+ i = from;
+ j = mid + 1;
+ for (k = from; (i <= mid) && (j <= to); k++)
+ if (cmpfun (ELT (base, i), ELT (base, j)) <= 0)
+ memcpy (ELT (temp, k), ELT (base, i++), size);
+ else
+ memcpy (ELT (temp, k), ELT (base, j++), size);
+ while (i <= mid)
+ memcpy (ELT (temp, k++), ELT (base, i++), size);
+ while (j <= to)
+ memcpy (ELT (temp, k++), ELT (base, j++), size);
+ for (k = from; k <= to; k++)
+ memcpy (ELT (base, k), ELT (temp, k), size);
+ }
+#undef ELT
+}
+
+/* Stable sort with interface exactly like standard library's qsort.
+ Uses mergesort internally, allocating temporary storage with
+ alloca. */
+
+void
+stable_sort (void *base, size_t nmemb, size_t size,
+ int (*cmpfun) (const void *, const void *))
+{
+ if (size > 1)
+ {
+ void *temp = alloca (nmemb * size * sizeof (void *));
+ mergesort_internal (base, temp, size, 0, nmemb - 1, cmpfun);
+ }
+}
+
+/* Print a decimal number. If it is equal to or larger than ten, the
+ number is rounded. Otherwise it is printed with one significant
+ digit without trailing zeros and with no more than three fractional
+ digits total. For example, 0.1 is printed as "0.1", 0.035 is
+ printed as "0.04", 0.0091 as "0.009", and 0.0003 as simply "0".
+
+ This is useful for displaying durations because it provides
+ order-of-magnitude information without unnecessary clutter --
+ long-running downloads are shown without the fractional part, and
+ short ones still retain one significant digit. */
+
+const char *
+print_decimal (double number)
+{
+ static char buf[32];
+ double n = number >= 0 ? number : -number;
+
+ if (n >= 9.95)
+ /* Cut off at 9.95 because the below %.1f would round 9.96 to
+ "10.0" instead of "10". OTOH 9.94 will print as "9.9". */
+ snprintf (buf, sizeof buf, "%.0f", number);
+ else if (n >= 0.95)
+ snprintf (buf, sizeof buf, "%.1f", number);
+ else if (n >= 0.001)
+ snprintf (buf, sizeof buf, "%.1g", number);
+ else if (n >= 0.0005)
+ /* round [0.0005, 0.001) to 0.001 */
+ snprintf (buf, sizeof buf, "%.3f", number);
+ else
+ /* print numbers close to 0 as 0, not 0.000 */
+ strcpy (buf, "0");
+
+ return buf;
+}
+
+#ifdef TESTING
+
+const char *
+test_subdir_p()
+{
+ int i;
+ struct {
+ char *d1;
+ char *d2;
+ bool result;
+ } test_array[] = {
+ { "/somedir", "/somedir", true },
+ { "/somedir", "/somedir/d2", true },
+ { "/somedir/d1", "/somedir", false },
+ };
+
+ for (i = 0; i < countof(test_array); ++i)
+ {
+ bool res = subdir_p (test_array[i].d1, test_array[i].d2);
+
+ mu_assert ("test_subdir_p: wrong result",
+ res == test_array[i].result);
+ }
+
+ return NULL;
+}
+
+const char *
+test_dir_matches_p()
+{
+ int i;
+ struct {
+ char *dirlist[3];
+ char *dir;
+ bool result;
+ } test_array[] = {
+ { { "/somedir", "/someotherdir", NULL }, "somedir", true },
+ { { "/somedir", "/someotherdir", NULL }, "anotherdir", false },
+ { { "/somedir", "/*otherdir", NULL }, "anotherdir", true },
+ { { "/somedir/d1", "/someotherdir", NULL }, "somedir/d1", true },
+ { { "/somedir/d1", "/someotherdir", NULL }, "d1", false },
+ };
+
+ for (i = 0; i < countof(test_array); ++i)
+ {
+ bool res = dir_matches_p (test_array[i].dirlist, test_array[i].dir);
+
+ mu_assert ("test_dir_matches_p: wrong result",
+ res == test_array[i].result);
+ }
+
+ return NULL;
+}
+
+#endif /* TESTING */
+