1 /* Various functions of utilitarian nature.
2 Copyright (C) 1995, 1996, 1997, 1998, 2000 Free Software Foundation, Inc.
4 This file is part of Wget.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
26 #else /* not HAVE_STRING_H */
28 #endif /* not HAVE_STRING_H */
29 #include <sys/types.h>
34 # include <sys/mman.h>
43 #ifdef HAVE_SYS_UTIME_H
44 # include <sys/utime.h>
48 # include <libc.h> /* for access() */
62 /* This section implements several wrappers around the basic
63 allocation routines. This is done for two reasons: first, so that
64 the callers of these functions need not consistently check for
65 errors. If there is not enough virtual memory for running Wget,
66 something is seriously wrong, and Wget exits with an appropriate
69 The second reason why these are useful is that, if DEBUG_MALLOC is
70 defined, they also provide a handy (if crude) malloc debugging
71 interface that checks memory leaks. */
73 /* Croak the fatal memory error and bail out with non-zero exit
76 memfatal (const char *what)
78 /* HACK: expose save_log_p from log.c, so we can turn it off in
79 order to prevent saving the log. Saving the log is dangerous
80 because logprintf() and logputs() can call malloc(), so this
81 could infloop. When logging is turned off, infloop can no longer
84 #### This is no longer really necessary because the new routines
85 in log.c cons only if the line exceeds eighty characters. But
86 this can come at the end of a line, so it's OK to be careful.
88 On a more serious note, it would be good to have a
89 log_forced_shutdown() routine that exposes this cleanly. */
90 extern int save_log_p;
93 logprintf (LOG_ALWAYS, _("%s: %s: Not enough memory.\n"), exec_name, what);
97 /* These functions end with _real because they need to be
98 distinguished from the debugging functions, and from the macros.
101 If memory debugging is not turned on, wget.h defines these:
103 #define xmalloc xmalloc_real
104 #define xrealloc xrealloc_real
105 #define xstrdup xstrdup_real
108 In case of memory debugging, the definitions are a bit more
109 complex, because we want to provide more information, *and* we want
110 to call the debugging code. (The former is the reason why xmalloc
111 and friends need to be macros in the first place.) Then it looks
114 #define xmalloc(a) xmalloc_debug (a, __FILE__, __LINE__)
115 #define xfree(a) xfree_debug (a, __FILE__, __LINE__)
116 #define xrealloc(a, b) xrealloc_debug (a, b, __FILE__, __LINE__)
117 #define xstrdup(a) xstrdup_debug (a, __FILE__, __LINE__)
119 Each of the *_debug function does its magic and calls the real one. */
122 # define STATIC_IF_DEBUG static
124 # define STATIC_IF_DEBUG
127 STATIC_IF_DEBUG void *
128 xmalloc_real (size_t size)
130 void *ptr = malloc (size);
136 STATIC_IF_DEBUG void *
137 xrealloc_real (void *ptr, size_t newsize)
141 /* Not all Un*xes have the feature of realloc() that calling it with
142 a NULL-pointer is the same as malloc(), but it is easy to
145 newptr = realloc (ptr, newsize);
147 newptr = malloc (newsize);
149 memfatal ("realloc");
153 STATIC_IF_DEBUG char *
154 xstrdup_real (const char *s)
160 copy = malloc (l + 1);
163 memcpy (copy, s, l + 1);
164 #else /* HAVE_STRDUP */
168 #endif /* HAVE_STRDUP */
175 /* Crude home-grown routines for debugging some malloc-related
178 * Counting the number of malloc and free invocations, and reporting
179 the "balance", i.e. how many times more malloc was called than it
180 was the case with free.
182 * Making malloc store its entry into a simple array and free remove
183 stuff from that array. At the end, print the pointers which have
184 not been freed, along with the source file and the line number.
185 This also has the side-effect of detecting freeing memory that
188 Note that this kind of memory leak checking strongly depends on
189 every malloc() being followed by a free(), even if the program is
190 about to finish. Wget is careful to free the data structure it
191 allocated in init.c. */
193 static int malloc_count, free_count;
199 } malloc_debug[100000];
201 /* Both register_ptr and unregister_ptr take O(n) operations to run,
202 which can be a real problem. It would be nice to use a hash table
203 for malloc_debug, but the functions in hash.c are not suitable
204 because they can call malloc() themselves. Maybe it would work if
205 the hash table were preallocated to a huge size, and if we set the
206 rehash threshold to 1.0. */
208 /* Register PTR in malloc_debug. Abort if this is not possible
209 (presumably due to the number of current allocations exceeding the
210 size of malloc_debug.) */
213 register_ptr (void *ptr, const char *file, int line)
216 for (i = 0; i < ARRAY_SIZE (malloc_debug); i++)
217 if (malloc_debug[i].ptr == NULL)
219 malloc_debug[i].ptr = ptr;
220 malloc_debug[i].file = file;
221 malloc_debug[i].line = line;
227 /* Unregister PTR from malloc_debug. Abort if PTR is not present in
228 malloc_debug. (This catches calling free() with a bogus pointer.) */
231 unregister_ptr (void *ptr)
234 for (i = 0; i < ARRAY_SIZE (malloc_debug); i++)
235 if (malloc_debug[i].ptr == ptr)
237 malloc_debug[i].ptr = NULL;
243 /* Print the malloc debug stats that can be gathered from the above
244 information. Currently this is the count of mallocs, frees, the
245 difference between the two, and the dump of the contents of
246 malloc_debug. The last part are the memory leaks. */
249 print_malloc_debug_stats (void)
252 printf ("\nMalloc: %d\nFree: %d\nBalance: %d\n\n",
253 malloc_count, free_count, malloc_count - free_count);
254 for (i = 0; i < ARRAY_SIZE (malloc_debug); i++)
255 if (malloc_debug[i].ptr != NULL)
256 printf ("0x%08ld: %s:%d\n", (long)malloc_debug[i].ptr,
257 malloc_debug[i].file, malloc_debug[i].line);
261 xmalloc_debug (size_t size, const char *source_file, int source_line)
263 void *ptr = xmalloc_real (size);
265 register_ptr (ptr, source_file, source_line);
270 xfree_debug (void *ptr, const char *source_file, int source_line)
272 assert (ptr != NULL);
274 unregister_ptr (ptr);
279 xrealloc_debug (void *ptr, size_t newsize, const char *source_file, int source_line)
281 void *newptr = xrealloc_real (ptr, newsize);
285 register_ptr (newptr, source_file, source_line);
287 else if (newptr != ptr)
289 unregister_ptr (ptr);
290 register_ptr (newptr, source_file, source_line);
296 xstrdup_debug (const char *s, const char *source_file, int source_line)
298 char *copy = xstrdup_real (s);
300 register_ptr (copy, source_file, source_line);
304 #endif /* DEBUG_MALLOC */
306 /* Copy the string formed by two pointers (one on the beginning, other
307 on the char after the last char) to a new, malloc-ed location.
310 strdupdelim (const char *beg, const char *end)
312 char *res = (char *)xmalloc (end - beg + 1);
313 memcpy (res, beg, end - beg);
314 res[end - beg] = '\0';
318 /* Parse a string containing comma-separated elements, and return a
319 vector of char pointers with the elements. Spaces following the
320 commas are ignored. */
322 sepstring (const char *s)
336 res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
337 res[i] = strdupdelim (p, s);
340 /* Skip the blanks following the ','. */
348 res = (char **)xrealloc (res, (i + 2) * sizeof (char *));
349 res[i] = strdupdelim (p, s);
354 /* Return pointer to a static char[] buffer in which zero-terminated
355 string-representation of TM (in form hh:mm:ss) is printed. It is
356 shamelessly non-reentrant, but it doesn't matter, really.
358 If TM is non-NULL, the time_t of the current time will be stored
361 time_str (time_t *tm)
371 ptm = localtime (&tim);
372 sprintf (tms, "%02d:%02d:%02d", ptm->tm_hour, ptm->tm_min, ptm->tm_sec);
376 /* Returns an error message for ERRNUM. #### This requires more work.
377 This function, as well as the whole error system, is very
380 uerrmsg (uerr_t errnum)
385 return _("Unknown/unsupported protocol");
388 return _("Invalid port specification");
391 return _("Invalid host name");
395 /* $@#@#$ compiler. */
400 /* The Windows versions of the following two functions are defined in
403 /* A cuserid() immitation using getpwuid(), to avoid hassling with
404 utmp. Besides, not all systems have cuesrid(). Under Windows, it
405 is defined in mswindows.c.
407 If WHERE is non-NULL, the username will be stored there.
408 Otherwise, it will be returned as a static buffer (as returned by
409 getpwuid()). In the latter case, the buffer should be copied
410 before calling getpwuid() or pwd_cuserid() again. */
413 pwd_cuserid (char *where)
417 if (!(pwd = getpwuid (getuid ())) || !pwd->pw_name)
421 strcpy (where, pwd->pw_name);
429 fork_to_background (void)
432 /* Whether we arrange our own version of opt.lfilename here. */
437 opt.lfilename = unique_name (DEFAULT_LOGFILE);
449 /* parent, no error */
450 printf (_("Continuing in background.\n"));
452 printf (_("Output will be written to `%s'.\n"), opt.lfilename);
455 /* child: keep running */
457 #endif /* not WINDOWS */
459 /* Canonicalize PATH, and return a new path. The new path differs from PATH
461 Multple `/'s are collapsed to a single `/'.
462 Leading `./'s and trailing `/.'s are removed.
463 Trailing `/'s are removed.
464 Non-leading `../'s and trailing `..'s are handled by removing
465 portions of the path.
467 E.g. "a/b/c/./../d/.." will yield "a/b". This function originates
471 Always use '/' as stub_char.
472 Don't check for local things using canon_stat.
473 Change the original string instead of strdup-ing.
474 React correctly when beginning with `./' and `../'. */
476 path_simplify (char *path)
478 register int i, start, ddot;
484 /*stub_char = (*path == '/') ? '/' : '.';*/
487 /* Addition: Remove all `./'-s preceding the string. If `../'-s
488 precede, put `/' in front and remove them too. */
493 if (path[i] == '.' && path[i + 1] == '/')
495 else if (path[i] == '.' && path[i + 1] == '.' && path[i + 2] == '/')
504 strcpy (path, path + i - ddot);
506 /* Replace single `.' or `..' with `/'. */
507 if ((path[0] == '.' && path[1] == '\0')
508 || (path[0] == '.' && path[1] == '.' && path[2] == '\0'))
514 /* Walk along PATH looking for things to compact. */
521 while (path[i] && path[i] != '/')
526 /* If we didn't find any slashes, then there is nothing left to do. */
530 /* Handle multiple `/'s in a row. */
531 while (path[i] == '/')
534 if ((start + 1) != i)
536 strcpy (path + start + 1, path + i);
540 /* Check for trailing `/'. */
541 if (start && !path[i])
548 /* Check for `../', `./' or trailing `.' by itself. */
551 /* Handle trailing `.' by itself. */
556 if (path[i + 1] == '/')
558 strcpy (path + i, path + i + 1);
559 i = (start < 0) ? 0 : start;
563 /* Handle `../' or trailing `..' by itself. */
564 if (path[i + 1] == '.' &&
565 (path[i + 2] == '/' || !path[i + 2]))
567 while (--start > -1 && path[start] != '/');
568 strcpy (path + start + 1, path + i + 2);
569 i = (start < 0) ? 0 : start;
582 /* "Touch" FILE, i.e. make its atime and mtime equal to the time
583 specified with TM. */
585 touch (const char *file, time_t tm)
587 #ifdef HAVE_STRUCT_UTIMBUF
588 struct utimbuf times;
589 times.actime = times.modtime = tm;
592 times[0] = times[1] = tm;
595 if (utime (file, ×) == -1)
596 logprintf (LOG_NOTQUIET, "utime(%s): %s\n", file, strerror (errno));
599 /* Checks if FILE is a symbolic link, and removes it if it is. Does
600 nothing under MS-Windows. */
602 remove_link (const char *file)
607 if (lstat (file, &st) == 0 && S_ISLNK (st.st_mode))
609 DEBUGP (("Unlinking %s (symlink).\n", file));
612 logprintf (LOG_VERBOSE, _("Failed to unlink symlink `%s': %s\n"),
613 file, strerror (errno));
618 /* Does FILENAME exist? This is quite a lousy implementation, since
619 it supplies no error codes -- only a yes-or-no answer. Thus it
620 will return that a file does not exist if, e.g., the directory is
621 unreadable. I don't mind it too much currently, though. The
622 proper way should, of course, be to have a third, error state,
623 other than true/false, but that would introduce uncalled-for
624 additional complexity to the callers. */
626 file_exists_p (const char *filename)
629 return access (filename, F_OK) >= 0;
632 return stat (filename, &buf) >= 0;
636 /* Returns 0 if PATH is a directory, 1 otherwise (any kind of file).
637 Returns 0 on error. */
639 file_non_directory_p (const char *path)
642 /* Use lstat() rather than stat() so that symbolic links pointing to
643 directories can be identified correctly. */
644 if (lstat (path, &buf) != 0)
646 return S_ISDIR (buf.st_mode) ? 0 : 1;
649 /* Return a unique filename, given a prefix and count */
651 unique_name_1 (const char *fileprefix, int count)
657 filename = (char *)xmalloc (strlen (fileprefix) + numdigit (count) + 2);
658 sprintf (filename, "%s.%d", fileprefix, count);
661 filename = xstrdup (fileprefix);
663 if (!file_exists_p (filename))
672 /* Return a unique file name, based on PREFIX. */
674 unique_name (const char *prefix)
680 file = unique_name_1 (prefix, count++);
684 /* Create DIRECTORY. If some of the pathname components of DIRECTORY
685 are missing, create them first. In case any mkdir() call fails,
686 return its error status. Returns 0 on successful completion.
688 The behaviour of this function should be identical to the behaviour
689 of `mkdir -p' on systems where mkdir supports the `-p' option. */
691 make_directory (const char *directory)
697 /* Make a copy of dir, to be able to write to it. Otherwise, the
698 function is unsafe if called with a read-only char *argument. */
699 STRDUP_ALLOCA (dir, directory);
701 /* If the first character of dir is '/', skip it (and thus enable
702 creation of absolute-pathname directories. */
703 for (i = (*dir == '/'); 1; ++i)
705 for (; dir[i] && dir[i] != '/'; i++)
710 /* Check whether the directory already exists. */
711 if (!file_exists_p (dir))
713 if (mkdir (dir, 0777) < 0)
724 static int in_acclist PARAMS ((const char *const *, const char *, int));
726 /* Determine whether a file is acceptable to be followed, according to
727 lists of patterns to accept/reject. */
729 acceptable (const char *s)
733 while (l && s[l] != '/')
740 return (in_acclist ((const char *const *)opt.accepts, s, 1)
741 && !in_acclist ((const char *const *)opt.rejects, s, 1));
743 return in_acclist ((const char *const *)opt.accepts, s, 1);
745 else if (opt.rejects)
746 return !in_acclist ((const char *const *)opt.rejects, s, 1);
750 /* Compare S1 and S2 frontally; S2 must begin with S1. E.g. if S1 is
751 `/something', frontcmp() will return 1 only if S2 begins with
752 `/something'. Otherwise, 0 is returned. */
754 frontcmp (const char *s1, const char *s2)
756 for (; *s1 && *s2 && (*s1 == *s2); ++s1, ++s2);
760 /* Iterate through STRLIST, and return the first element that matches
761 S, through wildcards or front comparison (as appropriate). */
763 proclist (char **strlist, const char *s, enum accd flags)
767 for (x = strlist; *x; x++)
768 if (has_wildcards_p (*x))
770 if (fnmatch (*x, s, FNM_PATHNAME) == 0)
775 char *p = *x + ((flags & ALLABS) && (**x == '/')); /* Remove '/' */
782 /* Returns whether DIRECTORY is acceptable for download, wrt the
783 include/exclude lists.
785 If FLAGS is ALLABS, the leading `/' is ignored in paths; relative
786 and absolute paths may be freely intermixed. */
788 accdir (const char *directory, enum accd flags)
790 /* Remove starting '/'. */
791 if (flags & ALLABS && *directory == '/')
795 if (!proclist (opt.includes, directory, flags))
800 if (proclist (opt.excludes, directory, flags))
806 /* Match the end of STRING against PATTERN. For instance:
808 match_backwards ("abc", "bc") -> 1
809 match_backwards ("abc", "ab") -> 0
810 match_backwards ("abc", "abc") -> 1 */
812 match_backwards (const char *string, const char *pattern)
816 for (i = strlen (string), j = strlen (pattern); i >= 0 && j >= 0; i--, j--)
817 if (string[i] != pattern[j])
819 /* If the pattern was exhausted, the match was succesful. */
826 /* Checks whether string S matches each element of ACCEPTS. A list
827 element are matched either with fnmatch() or match_backwards(),
828 according to whether the element contains wildcards or not.
830 If the BACKWARD is 0, don't do backward comparison -- just compare
833 in_acclist (const char *const *accepts, const char *s, int backward)
835 for (; *accepts; accepts++)
837 if (has_wildcards_p (*accepts))
839 /* fnmatch returns 0 if the pattern *does* match the
841 if (fnmatch (*accepts, s, 0) == 0)
848 if (match_backwards (s, *accepts))
853 if (!strcmp (s, *accepts))
861 /* Return the malloc-ed suffix of STR. For instance:
862 suffix ("foo.bar") -> "bar"
863 suffix ("foo.bar.baz") -> "baz"
864 suffix ("/foo/bar") -> NULL
865 suffix ("/foo.bar/baz") -> NULL */
867 suffix (const char *str)
871 for (i = strlen (str); i && str[i] != '/' && str[i] != '.'; i--);
873 return xstrdup (str + i);
878 /* Read a line from FP. The function reallocs the storage as needed
879 to accomodate for any length of the line. Reallocs are done
880 storage exponentially, doubling the storage after each overflow to
881 minimize the number of calls to realloc() and fgets(). The newline
882 character at the end of line is retained.
884 After end-of-file is encountered without anything being read, NULL
885 is returned. NULL is also returned on error. To distinguish
886 between these two cases, use the stdio function ferror(). */
889 read_whole_line (FILE *fp)
893 char *line = (char *)xmalloc (bufsize);
895 while (fgets (line + length, bufsize - length, fp))
897 length += strlen (line + length);
899 if (line[length - 1] == '\n')
901 /* fgets() guarantees to read the whole line, or to use up the
902 space we've given it. We can double the buffer
905 line = xrealloc (line, bufsize);
907 if (length == 0 || ferror (fp))
912 if (length + 1 < bufsize)
913 /* Relieve the memory from our exponential greediness. We say
914 `length + 1' because the terminating \0 is not included in
915 LENGTH. We don't need to zero-terminate the string ourselves,
916 though, because fgets() does that. */
917 line = xrealloc (line, length + 1);
921 /* Read FILE into memory. A pointer to `struct file_memory' are
922 returned; use struct element `content' to access file contents, and
923 the element `length' to know the file length. `content' is *not*
924 zero-terminated, and you should *not* read or write beyond the [0,
925 length) range of characters.
927 After you are done with the file contents, call read_file_free to
930 Depending on the operating system and the type of file that is
931 being read, read_file() either mmap's the file into memory, or
932 reads the file into the core using read().
934 If file is named "-", fileno(stdin) is used for reading instead.
935 If you want to read from a real file named "-", use "./-" instead. */
938 read_file (const char *file)
941 struct file_memory *fm;
943 int inhibit_close = 0;
945 /* Some magic in the finest tradition of Perl and its kin: if FILE
946 is "-", just use stdin. */
951 /* Note that we don't inhibit mmap() in this case. If stdin is
952 redirected from a regular file, mmap() will still work. */
955 fd = open (file, O_RDONLY);
958 fm = xmalloc (sizeof (struct file_memory));
963 if (fstat (fd, &buf) < 0)
965 fm->length = buf.st_size;
966 /* NOTE: As far as I know, the callers of this function never
967 modify the file text. Relying on this would enable us to
968 specify PROT_READ and MAP_SHARED for a marginal gain in
969 efficiency, but at some cost to generality. */
970 fm->content = mmap (NULL, fm->length, PROT_READ | PROT_WRITE,
972 if (fm->content == MAP_FAILED)
982 /* The most common reason why mmap() fails is that FD does not point
983 to a plain file. However, it's also possible that mmap() doesn't
984 work for a particular type of file. Therefore, whenever mmap()
985 fails, we just fall back to the regular method. */
986 #endif /* HAVE_MMAP */
989 size = 512; /* number of bytes fm->contents can
990 hold at any given time. */
991 fm->content = xmalloc (size);
995 if (fm->length > size / 2)
997 /* #### I'm not sure whether the whole exponential-growth
998 thing makes sense with kernel read. On Linux at least,
999 read() refuses to read more than 4K from a file at a
1000 single chunk anyway. But other Unixes might optimize it
1001 better, and it doesn't *hurt* anything, so I'm leaving
1004 /* Normally, we grow SIZE exponentially to make the number
1005 of calls to read() and realloc() logarithmic in relation
1006 to file size. However, read() can read an amount of data
1007 smaller than requested, and it would be unreasonably to
1008 double SIZE every time *something* was read. Therefore,
1009 we double SIZE only when the length exceeds half of the
1010 entire allocated size. */
1012 fm->content = xrealloc (fm->content, size);
1014 nread = read (fd, fm->content + fm->length, size - fm->length);
1016 /* Successful read. */
1017 fm->length += nread;
1027 if (size > fm->length && fm->length != 0)
1028 /* Due to exponential growth of fm->content, the allocated region
1029 might be much larger than what is actually needed. */
1030 fm->content = xrealloc (fm->content, fm->length);
1037 xfree (fm->content);
1042 /* Release the resources held by FM. Specifically, this calls
1043 munmap() or xfree() on fm->content, depending whether mmap or
1044 malloc/read were used to read in the file. It also frees the
1045 memory needed to hold the FM structure itself. */
1048 read_file_free (struct file_memory *fm)
1053 munmap (fm->content, fm->length);
1058 xfree (fm->content);
1063 /* Free the pointers in a NULL-terminated vector of pointers, then
1064 free the pointer itself. */
1066 free_vec (char **vec)
1077 /* Append vector V2 to vector V1. The function frees V2 and
1078 reallocates V1 (thus you may not use the contents of neither
1079 pointer after the call). If V1 is NULL, V2 is returned. */
1081 merge_vecs (char **v1, char **v2)
1091 /* To avoid j == 0 */
1096 for (i = 0; v1[i]; i++);
1098 for (j = 0; v2[j]; j++);
1099 /* Reallocate v1. */
1100 v1 = (char **)xrealloc (v1, (i + j + 1) * sizeof (char **));
1101 memcpy (v1 + i, v2, (j + 1) * sizeof (char *));
1106 /* A set of simple-minded routines to store strings in a linked list.
1107 This used to also be used for searching, but now we have hash
1110 /* It's a shame that these simple things like linked lists and hash
1111 tables (see hash.c) need to be implemented over and over again. It
1112 would be nice to be able to use the routines from glib -- see
1113 www.gtk.org for details. However, that would make Wget depend on
1114 glib, and I want to avoid dependencies to external libraries for
1115 reasons of convenience and portability (I suspect Wget is more
1116 portable than anything ever written for Gnome). */
1118 /* Append an element to the list. If the list has a huge number of
1119 elements, this can get slow because it has to find the list's
1120 ending. If you think you have to call slist_append in a loop,
1121 think about calling slist_prepend() followed by slist_nreverse(). */
1124 slist_append (slist *l, const char *s)
1126 slist *newel = (slist *)xmalloc (sizeof (slist));
1129 newel->string = xstrdup (s);
1134 /* Find the last element. */
1141 /* Prepend S to the list. Unlike slist_append(), this is O(1). */
1144 slist_prepend (slist *l, const char *s)
1146 slist *newel = (slist *)xmalloc (sizeof (slist));
1147 newel->string = xstrdup (s);
1152 /* Destructively reverse L. */
1155 slist_nreverse (slist *l)
1160 slist *next = l->next;
1168 /* Is there a specific entry in the list? */
1170 slist_contains (slist *l, const char *s)
1172 for (; l; l = l->next)
1173 if (!strcmp (l->string, s))
1178 /* Free the whole slist. */
1180 slist_free (slist *l)
1191 /* Sometimes it's useful to create "sets" of strings, i.e. special
1192 hash tables where you want to store strings as keys and merely
1193 query for their existence. Here is a set of utility routines that
1194 makes that transparent. */
1197 string_set_add (struct hash_table *ht, const char *s)
1199 /* First check whether the set element already exists. If it does,
1200 do nothing so that we don't have to free() the old element and
1201 then strdup() a new one. */
1202 if (hash_table_exists (ht, s))
1205 /* We use "1" as value. It provides us a useful and clear arbitrary
1206 value, and it consumes no memory -- the pointers to the same
1207 string "1" will be shared by all the key-value pairs in all `set'
1209 hash_table_put (ht, xstrdup (s), "1");
1212 /* Synonym for hash_table_exists... */
1215 string_set_exists (struct hash_table *ht, const char *s)
1217 return hash_table_exists (ht, s);
1221 string_set_free_mapper (void *key, void *value_ignored, void *arg_ignored)
1228 string_set_free (struct hash_table *ht)
1230 hash_table_map (ht, string_set_free_mapper, NULL);
1231 hash_table_destroy (ht);
1235 free_keys_and_values_mapper (void *key, void *value, void *arg_ignored)
1242 /* Another utility function: call free() on all keys and values of HT. */
1245 free_keys_and_values (struct hash_table *ht)
1247 hash_table_map (ht, free_keys_and_values_mapper, NULL);
1251 /* Engine for legible and legible_long_long; this function works on
1255 legible_1 (const char *repr)
1257 static char outbuf[128];
1262 /* Reset the pointers. */
1265 /* If the number is negative, shift the pointers. */
1271 /* How many digits before the first separator? */
1272 mod = strlen (inptr) % 3;
1274 for (i = 0; i < mod; i++)
1275 *outptr++ = inptr[i];
1276 /* Now insert the rest of them, putting separator before every
1278 for (i1 = i, i = 0; inptr[i1]; i++, i1++)
1280 if (i % 3 == 0 && i1 != 0)
1282 *outptr++ = inptr[i1];
1284 /* Zero-terminate the string. */
1289 /* Legible -- return a static pointer to the legibly printed long. */
1294 /* Print the number into the buffer. */
1295 long_to_string (inbuf, l);
1296 return legible_1 (inbuf);
1299 /* Write a string representation of NUMBER into the provided buffer.
1300 We cannot use sprintf() because we cannot be sure whether the
1301 platform supports printing of what we chose for VERY_LONG_TYPE.
1303 Example: Gcc supports `long long' under many platforms, but on many
1304 of those the native libc knows nothing of it and therefore cannot
1307 How long BUFFER needs to be depends on the platform and the content
1308 of NUMBER. For 64-bit VERY_LONG_TYPE (the most common case), 24
1309 bytes are sufficient. Using more might be a good idea.
1311 This function does not go through the hoops that long_to_string
1312 goes to because it doesn't need to be fast. (It's called perhaps
1313 once in a Wget run.) */
1316 very_long_to_string (char *buffer, VERY_LONG_TYPE number)
1321 /* Print the number backwards... */
1324 buffer[i++] = '0' + number % 10;
1329 /* ...and reverse the order of the digits. */
1330 for (j = 0; j < i / 2; j++)
1333 buffer[j] = buffer[i - 1 - j];
1334 buffer[i - 1 - j] = c;
1339 /* The same as legible(), but works on VERY_LONG_TYPE. See sysdep.h. */
1341 legible_very_long (VERY_LONG_TYPE l)
1344 /* Print the number into the buffer. */
1345 very_long_to_string (inbuf, l);
1346 return legible_1 (inbuf);
1349 /* Count the digits in a (long) integer. */
1354 while ((a /= 10) != 0)
1359 /* Print NUMBER to BUFFER. This is equivalent to sprintf(buffer,
1360 "%ld", number), only much faster.
1362 BUFFER should accept 24 bytes. This should suffice for the longest
1363 numbers on 64-bit machines, including the `-' sign and the trailing
1366 long_to_string (char *buffer, long number)
1368 #if (SIZEOF_LONG != 4) && (SIZEOF_LONG != 8)
1370 sprintf (buffer, "%ld", number);
1371 #else /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */
1381 #define FROB(figure) do { \
1382 if (force || number >= figure) \
1383 *p++ = number / figure + '0', number %= figure, force = 1; \
1385 #if SIZEOF_LONG == 8
1386 FROB (1000000000000000000L);
1387 FROB (100000000000000000L);
1388 FROB (10000000000000000L);
1389 FROB (1000000000000000L);
1390 FROB (100000000000000L);
1391 FROB (10000000000000L);
1392 FROB (1000000000000L);
1393 FROB (100000000000L);
1394 FROB (10000000000L);
1395 #endif /* SIZEOF_LONG == 8 */
1406 *p++ = number + '0';
1408 #endif /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */
1411 /* This should probably be at a better place, but it doesn't really
1412 fit into html-parse.c. */
1414 /* The function returns the pointer to the malloc-ed quoted version of
1415 string s. It will recognize and quote numeric and special graphic
1416 entities, as per RFC1866:
1424 No other entities are recognized or replaced. */
1426 html_quote_string (const char *s)
1432 /* Pass through the string, and count the new size. */
1433 for (i = 0; *s; s++, i++)
1436 i += 4; /* `amp;' */
1437 else if (*s == '<' || *s == '>')
1438 i += 3; /* `lt;' and `gt;' */
1439 else if (*s == '\"')
1440 i += 5; /* `quot;' */
1444 res = (char *)xmalloc (i + 1);
1446 for (p = res; *s; s++)
1459 *p++ = (*s == '<' ? 'l' : 'g');