#else /* not HAVE_STRING_H */
# include <strings.h>
#endif /* not HAVE_STRING_H */
-#include <ctype.h>
#include <sys/types.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
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. */
+ happen.
+
+ #### This is no longer really necessary because the new routines
+ in log.c cons only if the line exceeds eighty characters. But
+ this can come at the end of a line, so it's OK to be careful.
+
+ On a more serious note, it would be good to have a
+ log_forced_shutdown() routine that exposes this cleanly. */
extern int save_log_p;
save_log_p = 0;
- logprintf (LOG_ALWAYS, _("%s: %s: Not enough memory.\n"), exec_name, s);
+ 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. */
-void *
-xmalloc (size_t size)
-{
- void *res;
+/* These functions end with _real because they need to be
+ distinguished from the debugging functions, and from the macros.
+ Explanation follows:
+
+ If memory debugging is not turned on, wget.h defines these:
+
+ #define xmalloc xmalloc_real
+ #define xrealloc xrealloc_real
+ #define xstrdup xstrdup_real
+ #define xfree free
+
+ In case of memory debugging, the definitions are a bit more
+ complex, because we want to provide more information, *and* we want
+ to call the debugging code. (The former is the reason why xmalloc
+ and friends need to be macros in the first place.) Then it looks
+ like this:
- res = malloc (size);
- if (!res)
+ #define xmalloc(a) xmalloc_debug (a, __FILE__, __LINE__)
+ #define xfree(a) xfree_debug (a, __FILE__, __LINE__)
+ #define xrealloc(a, b) xrealloc_debug (a, b, __FILE__, __LINE__)
+ #define xstrdup(a) xstrdup_debug (a, __FILE__, __LINE__)
+
+ Each of the *_debug function does its magic and calls the real one. */
+
+#ifdef DEBUG_MALLOC
+# define STATIC_IF_DEBUG static
+#else
+# define STATIC_IF_DEBUG
+#endif
+
+STATIC_IF_DEBUG void *
+xmalloc_real (size_t size)
+{
+ void *ptr = malloc (size);
+ if (!ptr)
memfatal ("malloc");
- return res;
+ 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 */
\f
/* 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.
}
\f
/* 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 secs = time (tm);
+
+ 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;
+}
+
+/* Like the above, but include the date: YYYY-MM-DD hh:mm:ss. */
+
+char *
+datetime_str (time_t *tm)
+{
+ static char output[20]; /* "YYYY-MM-DD hh:mm:ss" + \0 */
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);
+
+ 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;
}
/* Returns an error message for ERRNUM. #### This requires more work.
return filename;
else
{
- free (filename);
+ xfree (filename);
return NULL;
}
}
}
if (length == 0 || ferror (fp))
{
- free (line);
+ xfree (line);
return NULL;
}
if (length + 1 < bufsize)
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);
lose:
if (!inhibit_close)
close (fd);
- free (fm->content);
- free (fm);
+ xfree (fm->content);
+ xfree (fm);
return NULL;
}
/* Release the resources held by FM. Specifically, this calls
- munmap() or free() on fm->content, depending whether mmap or
+ munmap() or xfree() on fm->content, depending whether mmap or
malloc/read were used to read in the file. It also frees the
memory needed to hold the FM structure itself. */
else
#endif
{
- free (fm->content);
+ xfree (fm->content);
}
- free (fm);
+ xfree (fm);
}
\f
/* Free the pointers in a NULL-terminated vector of pointers, then
{
char **p = vec;
while (*p)
- free (*p++);
- free (vec);
+ xfree (*p++);
+ xfree (vec);
}
}
if (!*v2)
{
/* To avoid j == 0 */
- free (v2);
+ xfree (v2);
return v1;
}
/* Count v1. */
/* Reallocate v1. */
v1 = (char **)xrealloc (v1, (i + j + 1) * sizeof (char **));
memcpy (v1 + i, v2, (j + 1) * sizeof (char *));
- free (v2);
+ xfree (v2);
return v1;
}
while (l)
{
slist *n = l->next;
- free (l->string);
- free (l);
+ xfree (l->string);
+ xfree (l);
l = n;
}
}
static int
string_set_free_mapper (void *key, void *value_ignored, void *arg_ignored)
{
- free (key);
+ xfree (key);
return 0;
}
static int
free_keys_and_values_mapper (void *key, void *value, void *arg_ignored)
{
- free (key);
- free (value);
+ xfree (key);
+ xfree (value);
return 0;
}
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 need 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);
}
*p = '\0';
#endif /* (SIZEOF_LONG == 4) || (SIZEOF_LONG == 8) */
}
+\f
+/* This should probably be at a better place, but it doesn't really
+ fit into html-parse.c. */
+
+/* The function returns the pointer to the malloc-ed quoted version of
+ string s. It will recognize and quote numeric and special graphic
+ entities, as per RFC1866:
+
+ `&' -> `&'
+ `<' -> `<'
+ `>' -> `>'
+ `"' -> `"'
+ SP -> ` '
+
+ No other entities are recognized or replaced. */
+char *
+html_quote_string (const char *s)
+{
+ 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;
+}
projects / wget / blobdiff