X-Git-Url: http://sjero.net/git/?a=blobdiff_plain;f=src%2Futils.c;h=74e4552fd4eeb5c67e58085a3e741a24ea42267f;hb=9228f0bf53d3b42459daeb28372196a007de3014;hp=984b1f80a2bffac35f792e3b48f2599e46b8cc47;hpb=2ffb47eabf9fe89d513dc79bdc535e4092e1d6ee;p=wget diff --git a/src/utils.c b/src/utils.c index 984b1f80..74e4552f 100644 --- a/src/utils.c +++ b/src/utils.c @@ -1,21 +1,32 @@ /* Various functions of utilitarian nature. - Copyright (C) 1995, 1996, 1997, 1998, 2000 Free Software Foundation, Inc. + Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001 + Free Software Foundation, Inc. -This file is part of Wget. +This file is part of GNU Wget. -This program is free software; you can redistribute it and/or modify +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 (at your option) any later version. -This program is distributed in the hope that it will be useful, +GNU Wget is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of 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 this program; if not, write to the Free Software -Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ +along with Wget; if not, write to the Free Software +Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + +In addition, as a special exception, the Free Software Foundation +gives permission to link the code of its release of Wget with the +OpenSSL project's "OpenSSL" library (or with modified versions of it +that use the same license as the "OpenSSL" library), and distribute +the linked executables. You must obey the GNU General Public License +in all respects for all of the code used other than "OpenSSL". If you +modify this file, you may extend this exception to your version of the +file, but you are not obligated to do so. If you do not wish to do +so, delete this exception statement from your version. */ #include @@ -26,7 +37,6 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #else /* not HAVE_STRING_H */ # include #endif /* not HAVE_STRING_H */ -#include #include #ifdef HAVE_UNISTD_H # include @@ -51,6 +61,35 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #include #include +/* For TIOCGWINSZ and friends: */ +#ifdef HAVE_SYS_IOCTL_H +# include +#endif +#ifdef HAVE_TERMIOS_H +# include +#endif + +/* Needed for run_with_timeout. */ +#undef USE_SIGNAL_TIMEOUT +#ifdef HAVE_SIGNAL_H +# include +#endif +#ifdef HAVE_SETJMP_H +# include +#endif +/* If sigsetjmp is a macro, configure won't pick it up. */ +#ifdef sigsetjmp +# define HAVE_SIGSETJMP +#endif +#ifdef HAVE_SIGNAL +# ifdef HAVE_SIGSETJMP +# define USE_SIGNAL_TIMEOUT +# endif +# ifdef HAVE_SIGBLOCK +# define USE_SIGNAL_TIMEOUT +# endif +#endif + #include "wget.h" #include "utils.h" #include "fnmatch.h" @@ -60,81 +99,263 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ extern int errno; #endif +/* This section implements several wrappers around the basic + allocation routines. This is done for two reasons: first, so that + the callers of these functions need not consistently check for + errors. If there is not enough virtual memory for running Wget, + something is seriously wrong, and Wget exits with an appropriate + error message. + + The second reason why these are useful is that, if DEBUG_MALLOC is + defined, they also provide a handy (if crude) malloc debugging + interface that checks memory leaks. */ /* Croak the fatal memory error and bail out with non-zero exit status. */ static void -memfatal (const char *s) +memfatal (const char *what) { - /* HACK: expose save_log_p from log.c, so we can turn it off in - order to prevent saving the log. Saving the log is dangerous - because logprintf() and logputs() can call malloc(), so this - could infloop. When logging is turned off, infloop can no longer - happen. */ - extern int save_log_p; - - save_log_p = 0; - logprintf (LOG_ALWAYS, _("%s: %s: Not enough memory.\n"), exec_name, s); + /* 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); } -/* xmalloc, xrealloc and xstrdup exit the program if there is not - enough memory. xstrdup also implements strdup on systems that do - not have it. xfree is provided to make leak-tracking easier. - Currently it's a no-op. */ +/* These functions end with _real because they need to be + distinguished from the debugging functions, and from the macros. + Explanation follows: -void * -xmalloc (size_t size) -{ - void *res; + If memory debugging is not turned on, wget.h defines these: - res = malloc (size); - if (!res) - memfatal ("malloc"); - return res; -} + #define xmalloc xmalloc_real + #define xrealloc xrealloc_real + #define xstrdup xstrdup_real + #define xfree free -void -xfree (void *ptr) + In case of memory debugging, the definitions are a bit more + complex, because we want to provide more information, *and* we want + to call the debugging code. (The former is the reason why xmalloc + and friends need to be macros in the first place.) Then it looks + like this: + + #define xmalloc(a) xmalloc_debug (a, __FILE__, __LINE__) + #define xfree(a) xfree_debug (a, __FILE__, __LINE__) + #define xrealloc(a, b) xrealloc_debug (a, b, __FILE__, __LINE__) + #define xstrdup(a) xstrdup_debug (a, __FILE__, __LINE__) + + Each of the *_debug function does its magic and calls the real one. */ + +#ifdef DEBUG_MALLOC +# define STATIC_IF_DEBUG static +#else +# define STATIC_IF_DEBUG +#endif + +STATIC_IF_DEBUG void * +xmalloc_real (size_t size) { - free (ptr); + void *ptr = malloc (size); + if (!ptr) + memfatal ("malloc"); + return ptr; } -void * -xrealloc (void *obj, size_t size) +STATIC_IF_DEBUG void * +xrealloc_real (void *ptr, size_t newsize) { - void *res; + void *newptr; /* Not all Un*xes have the feature of realloc() that calling it with a NULL-pointer is the same as malloc(), but it is easy to simulate. */ - if (obj) - res = realloc (obj, size); + if (ptr) + newptr = realloc (ptr, newsize); else - res = malloc (size); - if (!res) + newptr = malloc (newsize); + if (!newptr) memfatal ("realloc"); - return res; + return newptr; } -char * -xstrdup (const char *s) +STATIC_IF_DEBUG char * +xstrdup_real (const char *s) { + char *copy; + #ifndef HAVE_STRDUP int l = strlen (s); - char *s1 = malloc (l + 1); - if (!s1) + copy = malloc (l + 1); + if (!copy) memfatal ("strdup"); - memcpy (s1, s, l + 1); - return s1; + memcpy (copy, s, l + 1); #else /* HAVE_STRDUP */ - char *s1 = strdup (s); - if (!s1) + copy = strdup (s); + if (!copy) memfatal ("strdup"); - return s1; #endif /* HAVE_STRDUP */ + + return copy; } + +#ifdef DEBUG_MALLOC + +/* Crude home-grown routines for debugging some malloc-related + problems. Featured: + + * Counting the number of malloc and free invocations, and reporting + the "balance", i.e. how many times more malloc was called than it + was the case with free. + + * Making malloc store its entry into a simple array and free remove + stuff from that array. At the end, print the pointers which have + not been freed, along with the source file and the line number. + This also has the side-effect of detecting freeing memory that + was never allocated. + + Note that this kind of memory leak checking strongly depends on + every malloc() being followed by a free(), even if the program is + about to finish. Wget is careful to free the data structure it + allocated in init.c. */ + +static int malloc_count, free_count; + +static struct { + char *ptr; + const char *file; + int line; +} malloc_debug[100000]; + +/* Both register_ptr and unregister_ptr take O(n) operations to run, + which can be a real problem. It would be nice to use a hash table + for malloc_debug, but the functions in hash.c are not suitable + because they can call malloc() themselves. Maybe it would work if + the hash table were preallocated to a huge size, and if we set the + rehash threshold to 1.0. */ + +/* Register PTR in malloc_debug. Abort if this is not possible + (presumably due to the number of current allocations exceeding the + size of malloc_debug.) */ + +static void +register_ptr (void *ptr, const char *file, int line) +{ + int i; + for (i = 0; i < 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 (); +} + +/* Unregister PTR from malloc_debug. Abort if PTR is not present in + malloc_debug. (This catches calling free() with a bogus pointer.) */ + +static void +unregister_ptr (void *ptr) +{ + int i; + for (i = 0; i < ARRAY_SIZE (malloc_debug); i++) + if (malloc_debug[i].ptr == ptr) + { + malloc_debug[i].ptr = NULL; + return; + } + abort (); +} + +/* Print the malloc debug stats that can be gathered from the above + information. Currently this is the count of mallocs, frees, the + difference between the two, and the dump of the contents of + malloc_debug. The last part are the memory leaks. */ + +void +print_malloc_debug_stats (void) +{ + int i; + printf ("\nMalloc: %d\nFree: %d\nBalance: %d\n\n", + malloc_count, free_count, malloc_count - free_count); + for (i = 0; i < 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); +} + +void * +xmalloc_debug (size_t size, const char *source_file, int source_line) +{ + void *ptr = xmalloc_real (size); + ++malloc_count; + register_ptr (ptr, source_file, source_line); + return ptr; +} + +void +xfree_debug (void *ptr, const char *source_file, int source_line) +{ + assert (ptr != NULL); + ++free_count; + unregister_ptr (ptr); + free (ptr); +} + +void * +xrealloc_debug (void *ptr, size_t newsize, const char *source_file, int source_line) +{ + void *newptr = xrealloc_real (ptr, newsize); + if (!ptr) + { + ++malloc_count; + register_ptr (newptr, source_file, source_line); + } + else if (newptr != ptr) + { + unregister_ptr (ptr); + register_ptr (newptr, source_file, source_line); + } + return newptr; +} + +char * +xstrdup_debug (const char *s, const char *source_file, int source_line) +{ + char *copy = xstrdup_real (s); + ++malloc_count; + register_ptr (copy, source_file, source_line); + return copy; +} + +#endif /* DEBUG_MALLOC */ +/* Utility function: like xstrdup(), but also lowercases S. */ + +char * +xstrdup_lower (const char *s) +{ + char *copy = xstrdup (s); + char *p = copy; + for (; *p; p++) + *p = 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. */ @@ -184,79 +405,62 @@ sepstring (const char *s) } /* Return pointer to a static char[] buffer in which zero-terminated - string-representation of TM (in form hh:mm:ss) is printed. It is - shamelessly non-reentrant, but it doesn't matter, really. + string-representation of TM (in form hh:mm:ss) is printed. + + If TM is non-NULL, the current time-in-seconds will be stored + there. + + (#### 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.) */ - If TM is non-NULL, the time_t of the current time will be stored - there. */ char * time_str (time_t *tm) { - static char tms[15]; + static char output[15]; struct tm *ptm; - time_t tim; - - *tms = '\0'; - tim = time (tm); - if (tim == -1) - return tms; - ptm = localtime (&tim); - sprintf (tms, "%02d:%02d:%02d", ptm->tm_hour, ptm->tm_min, ptm->tm_sec); - return tms; -} + time_t secs = time (tm); -/* Returns an error message for ERRNUM. #### This requires more work. - This function, as well as the whole error system, is very - ill-conceived. */ -const char * -uerrmsg (uerr_t errnum) -{ - switch (errnum) + if (secs == -1) { - case URLUNKNOWN: - return _("Unknown/unsupported protocol"); - break; - case URLBADPORT: - return _("Invalid port specification"); - break; - case URLBADHOST: - return _("Invalid host name"); - break; - default: - abort (); - /* $@#@#$ compiler. */ - return NULL; + /* 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; } - -/* The Windows versions of the following two functions are defined in - mswindows.c. */ -/* A cuserid() immitation using getpwuid(), to avoid hassling with - utmp. Besides, not all systems have cuesrid(). Under Windows, it - is defined in mswindows.c. +/* Like the above, but include the date: YYYY-MM-DD hh:mm:ss. */ - If WHERE is non-NULL, the username will be stored there. - Otherwise, it will be returned as a static buffer (as returned by - getpwuid()). In the latter case, the buffer should be copied - before calling getpwuid() or pwd_cuserid() again. */ -#ifndef WINDOWS char * -pwd_cuserid (char *where) +datetime_str (time_t *tm) { - struct passwd *pwd; + static char output[20]; /* "YYYY-MM-DD hh:mm:ss" + \0 */ + struct tm *ptm; + time_t secs = time (tm); - if (!(pwd = getpwuid (getuid ())) || !pwd->pw_name) - return NULL; - if (where) + if (secs == -1) { - strcpy (where, pwd->pw_name); - return where; + /* In case of error, return the empty string. Maybe we should + just abort if this happens? */ + *output = '\0'; + return output; } - else - return pwd->pw_name; + 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; } + +/* The Windows versions of the following two functions are defined in + mswindows.c. */ +#ifndef WINDOWS void fork_to_background (void) { @@ -279,137 +483,19 @@ fork_to_background (void) else if (pid != 0) { /* parent, no error */ - printf (_("Continuing in background.\n")); + printf (_("Continuing in background, pid %d.\n"), (int)pid); if (changedp) printf (_("Output will be written to `%s'.\n"), opt.lfilename); - exit (0); - } - /* child: keep running */ -} -#endif /* not WINDOWS */ - -/* Canonicalize PATH, and return a new path. The new path differs from PATH - in that: - Multple `/'s are collapsed to a single `/'. - Leading `./'s and trailing `/.'s are removed. - Trailing `/'s are removed. - Non-leading `../'s and trailing `..'s are handled by removing - portions of the path. - - E.g. "a/b/c/./../d/.." will yield "a/b". This function originates - from GNU Bash. - - Changes for Wget: - Always use '/' as stub_char. - Don't check for local things using canon_stat. - Change the original string instead of strdup-ing. - React correctly when beginning with `./' and `../'. */ -void -path_simplify (char *path) -{ - register int i, start, ddot; - char stub_char; - - if (!*path) - return; - - /*stub_char = (*path == '/') ? '/' : '.';*/ - stub_char = '/'; - - /* Addition: Remove all `./'-s preceding the string. If `../'-s - precede, put `/' in front and remove them too. */ - i = 0; - ddot = 0; - while (1) - { - if (path[i] == '.' && path[i + 1] == '/') - i += 2; - else if (path[i] == '.' && path[i + 1] == '.' && path[i + 2] == '/') - { - i += 3; - ddot = 1; - } - else - break; + exit (0); /* #### should we use _exit()? */ } - if (i) - strcpy (path, path + i - ddot); - - /* Replace single `.' or `..' with `/'. */ - if ((path[0] == '.' && path[1] == '\0') - || (path[0] == '.' && path[1] == '.' && path[2] == '\0')) - { - path[0] = stub_char; - path[1] = '\0'; - return; - } - /* Walk along PATH looking for things to compact. */ - i = 0; - while (1) - { - if (!path[i]) - break; - - while (path[i] && path[i] != '/') - i++; - - start = i++; - - /* If we didn't find any slashes, then there is nothing left to do. */ - if (!path[start]) - break; - - /* Handle multiple `/'s in a row. */ - while (path[i] == '/') - i++; - if ((start + 1) != i) - { - strcpy (path + start + 1, path + i); - i = start + 1; - } - - /* Check for trailing `/'. */ - if (start && !path[i]) - { - zero_last: - path[--i] = '\0'; - break; - } - - /* Check for `../', `./' or trailing `.' by itself. */ - if (path[i] == '.') - { - /* Handle trailing `.' by itself. */ - if (!path[i + 1]) - goto zero_last; - - /* Handle `./'. */ - if (path[i + 1] == '/') - { - strcpy (path + i, path + i + 1); - i = (start < 0) ? 0 : start; - continue; - } - - /* Handle `../' or trailing `..' by itself. */ - if (path[i + 1] == '.' && - (path[i + 2] == '/' || !path[i + 2])) - { - while (--start > -1 && path[start] != '/'); - strcpy (path + start + 1, path + i + 2); - i = (start < 0) ? 0 : start; - continue; - } - } /* path == '.' */ - } /* while */ - - if (!*path) - { - *path = stub_char; - path[1] = '\0'; - } + /* child: give up the privileges and keep running. */ + setsid (); + freopen ("/dev/null", "r", stdin); + freopen ("/dev/null", "w", stdout); + freopen ("/dev/null", "w", stderr); } +#endif /* not WINDOWS */ /* "Touch" FILE, i.e. make its atime and mtime equal to the time specified with TM. */ @@ -478,6 +564,22 @@ file_non_directory_p (const char *path) return S_ISDIR (buf.st_mode) ? 0 : 1; } +/* Return the size of file named by FILENAME, or -1 if it cannot be + opened or seeked into. */ +long +file_size (const char *filename) +{ + long size; + /* We use fseek rather than stat to determine the file size because + that way we can also verify whether the file is readable. + Inspired by the POST patch by Arnaud Wylie. */ + FILE *fp = fopen (filename, "rb"); + fseek (fp, 0, SEEK_END); + size = ftell (fp); + fclose (fp); + return size; +} + /* Return a unique filename, given a prefix and count */ static char * unique_name_1 (const char *fileprefix, int count) @@ -524,6 +626,7 @@ make_directory (const char *directory) { int quit = 0; int i; + int ret = 0; char *dir; /* Make a copy of dir, to be able to write to it. Otherwise, the @@ -539,18 +642,45 @@ make_directory (const char *directory) if (!dir[i]) 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; else dir[i] = '/'; } - return 0; + return ret; +} + +/* Merge BASE with FILE. BASE can be a directory or a file name, FILE + should be a file name. + + file_merge("/foo/bar", "baz") => "/foo/baz" + 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. */ + +char * +file_merge (const char *base, const char *file) +{ + char *result; + const char *cut = (const char *)strrchr (base, '/'); + + if (!cut) + return xstrdup (file); + + result = (char *)xmalloc (cut - base + 1 + strlen (file) + 1); + memcpy (result, base, cut - base); + result[cut - base] = '/'; + strcpy (result + (cut - base) + 1, file); + + return result; } static int in_acclist PARAMS ((const char *const *, const char *, int)); @@ -635,20 +765,37 @@ accdir (const char *directory, enum accd flags) return 1; } -/* Match the end of STRING against PATTERN. For instance: +/* Return non-zero if STRING ends with TAIL. For instance: - match_backwards ("abc", "bc") -> 1 - match_backwards ("abc", "ab") -> 0 - match_backwards ("abc", "abc") -> 1 */ -static int -match_backwards (const char *string, const char *pattern) + match_tail ("abc", "bc", 0) -> 1 + match_tail ("abc", "ab", 0) -> 0 + match_tail ("abc", "abc", 0) -> 1 + + If FOLD_CASE_P is non-zero, the comparison will be + case-insensitive. */ + +int +match_tail (const char *string, const char *tail, int fold_case_p) { int i, j; - 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. */ + /* We want this to be fast, so we code two loops, one with + case-folding, one without. */ + + if (!fold_case_p) + { + for (i = strlen (string), j = strlen (tail); i >= 0 && j >= 0; i--, j--) + if (string[i] != tail[j]) + break; + } + else + { + for (i = strlen (string), j = strlen (tail); i >= 0 && j >= 0; i--, j--) + if (TOLOWER (string[i]) != TOLOWER (tail[j])) + break; + } + + /* If the tail was exhausted, the match was succesful. */ if (j == -1) return 1; else @@ -656,7 +803,7 @@ match_backwards (const char *string, const char *pattern) } /* Checks whether string S matches each element of ACCEPTS. A list - element are matched either with fnmatch() or match_backwards(), + 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 @@ -677,7 +824,7 @@ in_acclist (const char *const *accepts, const char *s, int backward) { if (backward) { - if (match_backwards (s, *accepts)) + if (match_tail (s, *accepts, 0)) return 1; } else @@ -690,7 +837,7 @@ in_acclist (const char *const *accepts, const char *s, int backward) return 0; } -/* Return the malloc-ed suffix of STR. For instance: +/* Return the location of STR's suffix (file extension). Examples: suffix ("foo.bar") -> "bar" suffix ("foo.bar.baz") -> "baz" suffix ("/foo/bar") -> NULL @@ -700,18 +847,47 @@ suffix (const char *str) { int i; - for (i = strlen (str); i && str[i] != '/' && str[i] != '.'; i--); + for (i = strlen (str); i && str[i] != '/' && str[i] != '.'; i--) + ; + if (str[i++] == '.') - return xstrdup (str + i); + return (char *)str + i; else return NULL; } -/* Read a line from FP. The function reallocs the storage as needed - to accomodate for any length of the line. Reallocs are done - storage exponentially, doubling the storage after each overflow to - minimize the number of calls to realloc() and fgets(). The newline - character at the end of line is retained. +/* Return non-zero if FNAME ends with a typical HTML suffix. The + following (case-insensitive) suffixes are presumed to be HTML files: + + html + htm + ?html (`?' matches one character) + + #### CAVEAT. This is not necessarily a good indication that FNAME + refers to a file that contains HTML! */ +int +has_html_suffix_p (const char *fname) +{ + char *suf; + + if ((suf = suffix (fname)) == NULL) + return 0; + if (!strcasecmp (suf, "html")) + return 1; + if (!strcasecmp (suf, "htm")) + return 1; + if (suf[0] && !strcasecmp (suf + 1, "html")) + return 1; + return 0; +} + +/* 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. + + 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. After end-of-file is encountered without anything being read, NULL is returned. NULL is also returned on error. To distinguish @@ -721,15 +897,20 @@ char * read_whole_line (FILE *fp) { int length = 0; - int bufsize = 81; + int bufsize = 82; char *line = (char *)xmalloc (bufsize); while (fgets (line + length, bufsize - length, fp)) { length += strlen (line + length); - assert (length > 0); + 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. */ @@ -801,7 +982,7 @@ read_file (const char *file) efficiency, but at some cost to generality. */ fm->content = mmap (NULL, fm->length, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0); - if (fm->content == MAP_FAILED) + if (fm->content == (char *)MAP_FAILED) goto mmap_lose; if (!inhibit_close) close (fd); @@ -1031,7 +1212,7 @@ string_set_add (struct hash_table *ht, const char *s) /* First check whether the set element already exists. If it does, do nothing so that we don't have to free() the old element and then strdup() a new one. */ - if (hash_table_exists (ht, s)) + if (hash_table_contains (ht, s)) return; /* We use "1" as value. It provides us a useful and clear arbitrary @@ -1041,12 +1222,12 @@ string_set_add (struct hash_table *ht, const char *s) hash_table_put (ht, xstrdup (s), "1"); } -/* Synonym for hash_table_exists... */ +/* Synonym for hash_table_contains... */ int -string_set_exists (struct hash_table *ht, const char *s) +string_set_contains (struct hash_table *ht, const char *s) { - return hash_table_exists (ht, s); + return hash_table_contains (ht, s); } static int @@ -1080,7 +1261,7 @@ free_keys_and_values (struct hash_table *ht) } -/* Engine for legible and legible_long_long; this function works on +/* Engine for legible and legible_very_long; this function works on strings. */ static char * @@ -1124,80 +1305,416 @@ legible (long l) { char inbuf[24]; /* Print the number into the buffer. */ - long_to_string (inbuf, l); + number_to_string (inbuf, l); return legible_1 (inbuf); } +/* 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. + + 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. + + 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. + + 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.) */ + +static void +very_long_to_string (char *buffer, VERY_LONG_TYPE number) +{ + int i = 0; + int j; + + /* Print the number backwards... */ + do + { + buffer[i++] = '0' + number % 10; + number /= 10; + } + while (number); + + /* ...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'; +} + /* The same as legible(), but works on VERY_LONG_TYPE. See sysdep.h. */ char * legible_very_long (VERY_LONG_TYPE l) { char inbuf[128]; /* Print the number into the buffer. */ - sprintf (inbuf, VERY_LONG_FORMAT, l); + very_long_to_string (inbuf, l); return legible_1 (inbuf); } /* Count the digits in a (long) integer. */ int -numdigit (long a) +numdigit (long number) { - int res = 1; - while ((a /= 10) != 0) - ++res; - return res; + int cnt = 1; + if (number < 0) + { + number = -number; + ++cnt; + } + while ((number /= 10) > 0) + ++cnt; + return cnt; } -/* Print NUMBER to BUFFER. This is equivalent to sprintf(buffer, - "%ld", number), only much faster. +/* A half-assed implementation of INT_MAX on machines that don't + bother to define one. */ +#ifndef INT_MAX +# define INT_MAX ((int) ~((unsigned)1 << 8 * sizeof (int) - 1)) +#endif - BUFFER should accept 24 bytes. This should suffice for the longest - numbers on 64-bit machines, including the `-' sign and the trailing - \0. */ -void -long_to_string (char *buffer, long number) +#define ONE_DIGIT(figure) *p++ = n / (figure) + '0' +#define ONE_DIGIT_ADVANCE(figure) (ONE_DIGIT (figure), n %= (figure)) + +#define DIGITS_1(figure) ONE_DIGIT (figure) +#define DIGITS_2(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_1 ((figure) / 10) +#define DIGITS_3(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_2 ((figure) / 10) +#define DIGITS_4(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_3 ((figure) / 10) +#define DIGITS_5(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_4 ((figure) / 10) +#define DIGITS_6(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_5 ((figure) / 10) +#define DIGITS_7(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_6 ((figure) / 10) +#define DIGITS_8(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_7 ((figure) / 10) +#define DIGITS_9(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_8 ((figure) / 10) +#define DIGITS_10(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_9 ((figure) / 10) + +/* DIGITS_<11-20> are only used on machines with 64-bit longs. */ + +#define DIGITS_11(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_10 ((figure) / 10) +#define DIGITS_12(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_11 ((figure) / 10) +#define DIGITS_13(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_12 ((figure) / 10) +#define DIGITS_14(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_13 ((figure) / 10) +#define DIGITS_15(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_14 ((figure) / 10) +#define DIGITS_16(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_15 ((figure) / 10) +#define DIGITS_17(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_16 ((figure) / 10) +#define DIGITS_18(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_17 ((figure) / 10) +#define DIGITS_19(figure) ONE_DIGIT_ADVANCE (figure); DIGITS_18 ((figure) / 10) + +/* Print NUMBER to BUFFER in base 10. This should be completely + equivalent to `sprintf(buffer, "%ld", number)', only much faster. + + 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. + + 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 + 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) { -#if (SIZEOF_LONG != 4) && (SIZEOF_LONG != 8) - /* Huh? */ - sprintf (buffer, "%ld", number); -#else /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */ char *p = buffer; - int force = 0; + long n = number; - if (number < 0) +#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) */ + + if (n < 0) { + if (n < -INT_MAX) + { + /* We cannot print a '-' and assign -n to n because -n would + overflow. Let sprintf deal with this border case. */ + sprintf (buffer, "%ld", n); + p += strlen (buffer); + return p; + } + *p++ = '-'; - number = -number; + n = -n; } -#define FROB(figure) do { \ - if (force || number >= figure) \ - *p++ = number / figure + '0', number %= figure, force = 1; \ - } while (0) -#if SIZEOF_LONG == 8 - FROB (1000000000000000000L); - FROB (100000000000000000L); - FROB (10000000000000000L); - FROB (1000000000000000L); - FROB (100000000000000L); - FROB (10000000000000L); - FROB (1000000000000L); - FROB (100000000000L); - FROB (10000000000L); -#endif /* SIZEOF_LONG == 8 */ - FROB (1000000000); - FROB (100000000); - FROB (10000000); - FROB (1000000); - FROB (100000); - FROB (10000); - FROB (1000); - FROB (100); - FROB (10); -#undef FROB - *p++ = number + '0'; + 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 */ + *p = '\0'; #endif /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */ + + return p; +} + +#undef ONE_DIGIT +#undef ONE_DIGIT_ADVANCE + +#undef DIGITS_1 +#undef DIGITS_2 +#undef DIGITS_3 +#undef DIGITS_4 +#undef DIGITS_5 +#undef DIGITS_6 +#undef DIGITS_7 +#undef DIGITS_8 +#undef DIGITS_9 +#undef DIGITS_10 +#undef DIGITS_11 +#undef DIGITS_12 +#undef DIGITS_13 +#undef DIGITS_14 +#undef DIGITS_15 +#undef DIGITS_16 +#undef DIGITS_17 +#undef DIGITS_18 +#undef DIGITS_19 + +/* Support for timers. */ + +#undef TIMER_WINDOWS +#undef TIMER_GETTIMEOFDAY +#undef TIMER_TIME + +/* Depending on the OS and availability of gettimeofday(), one and + only one of the above constants will be defined. Virtually all + modern Unix systems will define TIMER_GETTIMEOFDAY; Windows will + use TIMER_WINDOWS. TIMER_TIME is a catch-all method for + non-Windows systems without gettimeofday. + + #### Perhaps we should also support ftime(), which exists on old + BSD 4.2-influenced systems? (It also existed under MS DOS Borland + C, if memory serves me.) */ + +#ifdef WINDOWS +# define TIMER_WINDOWS +#else /* not WINDOWS */ +# ifdef HAVE_GETTIMEOFDAY +# define TIMER_GETTIMEOFDAY +# else +# define TIMER_TIME +# endif +#endif /* not WINDOWS */ + +#ifdef TIMER_GETTIMEOFDAY +typedef struct timeval wget_sys_time; +#endif + +#ifdef TIMER_TIME +typedef time_t wget_sys_time; +#endif + +#ifdef TIMER_WINDOWS +typedef ULARGE_INTEGER wget_sys_time; +#endif + +struct wget_timer { + /* The starting point in time which, subtracted from the current + time, yields elapsed time. */ + wget_sys_time start; + + /* The most recent elapsed time, calculated by wtimer_elapsed(). + Measured in milliseconds. */ + double elapsed_last; + + /* Approximately, the time elapsed between the true start of the + measurement and the time represented by START. */ + double elapsed_pre_start; +}; + +/* Allocate a timer. It is not legal to do anything with a freshly + allocated timer, except call wtimer_reset() or wtimer_delete(). */ + +struct wget_timer * +wtimer_allocate (void) +{ + struct wget_timer *wt = + (struct wget_timer *)xmalloc (sizeof (struct wget_timer)); + return wt; +} + +/* Allocate a new timer and reset it. Return the new timer. */ + +struct wget_timer * +wtimer_new (void) +{ + struct wget_timer *wt = wtimer_allocate (); + wtimer_reset (wt); + return wt; +} + +/* Free the resources associated with the timer. Its further use is + prohibited. */ + +void +wtimer_delete (struct wget_timer *wt) +{ + xfree (wt); +} + +/* Store system time to WST. */ + +static void +wtimer_sys_set (wget_sys_time *wst) +{ +#ifdef TIMER_GETTIMEOFDAY + gettimeofday (wst, NULL); +#endif + +#ifdef TIMER_TIME + time (wst); +#endif + +#ifdef TIMER_WINDOWS + /* We use GetSystemTime to get the elapsed time. MSDN warns that + system clock adjustments can skew the output of GetSystemTime + when used as a timer and gives preference to GetTickCount and + high-resolution timers. But GetTickCount can overflow, and hires + timers are typically used for profiling, not for regular time + measurement. Since we handle clock skew anyway, we just use + GetSystemTime. */ + FILETIME ft; + SYSTEMTIME st; + GetSystemTime (&st); + + /* As recommended by MSDN, we convert SYSTEMTIME to FILETIME, copy + FILETIME to ULARGE_INTEGER, and use regular 64-bit integer + arithmetic on that. */ + SystemTimeToFileTime (&st, &ft); + wst->HighPart = ft.dwHighDateTime; + wst->LowPart = ft.dwLowDateTime; +#endif +} + +/* Reset timer WT. This establishes the starting point from which + wtimer_elapsed() will return the number of elapsed + milliseconds. It is allowed to reset a previously used timer. */ + +void +wtimer_reset (struct wget_timer *wt) +{ + /* Set the start time to the current time. */ + wtimer_sys_set (&wt->start); + wt->elapsed_last = 0; + wt->elapsed_pre_start = 0; +} + +static double +wtimer_sys_diff (wget_sys_time *wst1, wget_sys_time *wst2) +{ +#ifdef TIMER_GETTIMEOFDAY + return ((double)(wst1->tv_sec - wst2->tv_sec) * 1000 + + (double)(wst1->tv_usec - wst2->tv_usec) / 1000); +#endif + +#ifdef TIMER_TIME + return 1000 * (*wst1 - *wst2); +#endif + +#ifdef WINDOWS + return (double)(wst1->QuadPart - wst2->QuadPart) / 10000; +#endif +} + +/* Return the number of milliseconds elapsed since the timer was last + reset. It is allowed to call this function more than once to get + increasingly higher elapsed values. These timers handle clock + skew. */ + +double +wtimer_elapsed (struct wget_timer *wt) +{ + wget_sys_time now; + double elapsed; + + wtimer_sys_set (&now); + elapsed = wt->elapsed_pre_start + wtimer_sys_diff (&now, &wt->start); + + /* Ideally we'd just return the difference between NOW and + wt->start. However, the system timer can be set back, and we + could return a value smaller than when we were last called, even + a negative value. Both of these would confuse the callers, which + expect us to return monotonically nondecreasing values. + + Therefore: if ELAPSED is smaller than its previous known value, + we reset wt->start to the current time and effectively start + measuring from this point. But since we don't want the elapsed + value to start from zero, we set elapsed_pre_start to the last + elapsed time and increment all future calculations by that + amount. */ + + if (elapsed < wt->elapsed_last) + { + wt->start = now; + wt->elapsed_pre_start = wt->elapsed_last; + elapsed = wt->elapsed_last; + } + + wt->elapsed_last = elapsed; + return elapsed; +} + +/* Return the assessed granularity of the timer implementation, in + milliseconds. This is used by code that tries to substitute a + better value for timers that have returned zero. */ + +double +wtimer_granularity (void) +{ +#ifdef TIMER_GETTIMEOFDAY + /* Granularity of gettimeofday varies wildly between architectures. + However, it appears that on modern machines it tends to be better + than 1ms. Assume 100 usecs. (Perhaps the configure process + could actually measure this?) */ + return 0.1; +#endif + +#ifdef TIMER_TIME + /* This is clear. */ + return 1000; +#endif + +#ifdef TIMER_WINDOWS + /* According to MSDN, GetSystemTime returns a broken-down time + structure the smallest member of which are milliseconds. */ + return 1; +#endif } /* This should probably be at a better place, but it doesn't really @@ -1274,3 +1791,177 @@ html_quote_string (const char *s) *p = '\0'; return res; } + +/* Determine the width of the terminal we're running on. If that's + not possible, return 0. */ + +int +determine_screen_width (void) +{ + /* 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; + + if (opt.lfilename != NULL) + return 0; + + fd = fileno (stderr); + if (ioctl (fd, TIOCGWINSZ, &wsz) < 0) + return 0; /* most likely ENOTTY */ + + return wsz.ws_col; +#endif /* TIOCGWINSZ */ +} + +/* 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. + + 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. */ + +int +random_number (int max) +{ + static int seeded; + double bounded; + int rnd; + + if (!seeded) + { + srand (time (NULL)); + 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 + + /* 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; +} + +#if 0 +/* A debugging function for checking whether an MD5 library works. */ + +#include "gen-md5.h" + +char * +debug_test_md5 (char *buf) +{ + unsigned char raw[16]; + static char res[33]; + unsigned char *p1; + char *p2; + int cnt; + ALLOCA_MD5_CONTEXT (ctx); + + gen_md5_init (ctx); + gen_md5_update ((unsigned char *)buf, strlen (buf), ctx); + gen_md5_finish (ctx, raw); + + p1 = raw; + p2 = res; + cnt = 16; + while (cnt--) + { + *p2++ = XDIGIT_TO_xchar (*p1 >> 4); + *p2++ = XDIGIT_TO_xchar (*p1 & 0xf); + ++p1; + } + *p2 = '\0'; + + return res; +} +#endif + +/* Implementation of run_with_timeout, a generic timeout handler for + systems with Unix-like signal handling. */ +#ifdef USE_SIGNAL_TIMEOUT +# ifdef HAVE_SIGSETJMP +# define SETJMP(env) sigsetjmp (env, 1) + +static sigjmp_buf run_with_timeout_env; + +static RETSIGTYPE +abort_run_with_timeout (int sig) +{ + assert (sig == SIGALRM); + siglongjmp (run_with_timeout_env, -1); +} +# else /* not HAVE_SIGSETJMP */ +# define SETJMP(env) setjmp (env) + +static jmp_buf run_with_timeout_env; + +static RETSIGTYPE +abort_run_with_timeout (int sig) +{ + 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. */ + int mask = siggetmask (); + mask &= ~sigmask(SIGALRM); + sigsetmask (mask); + + /* Now it's safe to longjump. */ + longjmp (run_with_timeout_env, -1); +} +# endif /* not HAVE_SIGSETJMP */ +#endif /* USE_SIGNAL_TIMEOUT */ + +int +run_with_timeout (long timeout, void (*fun) (void *), void *arg) +{ +#ifndef USE_SIGNAL_TIMEOUT + fun (arg); + return 0; +#else + int saved_errno; + + if (timeout == 0) + { + fun (arg); + return 0; + } + + 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 1; + } + alarm (timeout); + fun (arg); + + /* Preserve errno in case alarm() or signal() modifies it. */ + saved_errno = errno; + alarm (0); + signal (SIGALRM, SIG_DFL); + errno = saved_errno; + + return 0; +#endif +} +