-/* Various functions of utilitarian nature.
- Copyright (C) 1995, 1996, 1997, 1998, 2000, 2001
- Free Software Foundation, Inc.
+/* Various utility functions.
+ Copyright (C) 2003 Free Software Foundation, Inc.
This file is part of GNU Wget.
#ifdef HAVE_SETJMP_H
# include <setjmp.h>
#endif
+
+#ifndef HAVE_SIGSETJMP
/* If sigsetjmp is a macro, configure won't pick it up. */
-#ifdef sigsetjmp
-# define HAVE_SIGSETJMP
+# ifdef sigsetjmp
+# define HAVE_SIGSETJMP
+# endif
#endif
+
#ifdef HAVE_SIGNAL
# ifdef HAVE_SIGSETJMP
# define USE_SIGNAL_TIMEOUT
#include "wget.h"
#include "utils.h"
-#include "fnmatch.h"
#include "hash.h"
#ifndef errno
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 *what)
-{
- /* Make sure we don't try to store part of the log line, and thus
- call malloc. */
- log_set_save_context (0);
- logprintf (LOG_ALWAYS, _("%s: %s: Not enough memory.\n"), exec_name, what);
- exit (1);
-}
-
-/* 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:
-
- #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 ptr;
-}
-
-STATIC_IF_DEBUG void *
-xrealloc_real (void *ptr, size_t newsize)
-{
- 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 (ptr)
- newptr = realloc (ptr, newsize);
- else
- newptr = malloc (newsize);
- if (!newptr)
- memfatal ("realloc");
- return newptr;
-}
-
-STATIC_IF_DEBUG char *
-xstrdup_real (const char *s)
-{
- char *copy;
-
-#ifndef HAVE_STRDUP
- int l = strlen (s);
- copy = malloc (l + 1);
- if (!copy)
- memfatal ("strdup");
- memcpy (copy, s, l + 1);
-#else /* HAVE_STRDUP */
- copy = strdup (s);
- if (!copy)
- memfatal ("strdup");
-#endif /* HAVE_STRDUP */
-
- 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
/* Utility function: like xstrdup(), but also lowercases S. */
char *
that way we can also verify whether the file is readable.
Inspired by the POST patch by Arnaud Wylie. */
FILE *fp = fopen (filename, "rb");
+ if (!fp)
+ return -1;
fseek (fp, 0, SEEK_END);
size = ftell (fp);
fclose (fp);
return NULL;
}
+/* Return non-zero if S contains globbing wildcards (`*', `?', `[' or
+ `]'). */
+
+int
+has_wildcards_p (const char *s)
+{
+ for (; *s; s++)
+ if (*s == '*' || *s == '?' || *s == '[' || *s == ']')
+ return 1;
+ return 0;
+}
+
/* Return non-zero if FNAME ends with a typical HTML suffix. The
following (case-insensitive) suffixes are presumed to be HTML files:
}
/* 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.
+ storage. The storage 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
fd = open (file, O_RDONLY);
if (fd < 0)
return NULL;
- fm = xmalloc (sizeof (struct file_memory));
+ fm = xnew (struct file_memory);
#ifdef HAVE_MMAP
{
/* Normally, we grow SIZE exponentially to make the number
of calls to read() and realloc() logarithmic in relation
to file size. However, read() can read an amount of data
- smaller than requested, and it would be unreasonably to
+ smaller than requested, and it would be unreasonable to
double SIZE every time *something* was read. Therefore,
we double SIZE only when the length exceeds half of the
entire allocated size. */
slist *
slist_append (slist *l, const char *s)
{
- slist *newel = (slist *)xmalloc (sizeof (slist));
+ slist *newel = xnew (slist);
slist *beg = l;
newel->string = xstrdup (s);
slist *
slist_prepend (slist *l, const char *s)
{
- slist *newel = (slist *)xmalloc (sizeof (slist));
+ slist *newel = xnew (slist);
newel->string = xstrdup (s);
newel->next = l;
return newel;
}
\f
-/* Engine for legible and legible_very_long; this function works on
- strings. */
+/* Engine for legible and legible_large_int; add thousand separators
+ to numbers printed in strings. */
static char *
legible_1 (const char *repr)
{
- static char outbuf[128];
+ static char outbuf[48];
int i, i1, mod;
char *outptr;
const char *inptr;
/* Reset the pointers. */
outptr = outbuf;
inptr = repr;
- /* If the number is negative, shift the pointers. */
+
+ /* Ignore the sign for the purpose of adding thousand
+ separators. */
if (*inptr == '-')
{
*outptr++ = '-';
}
/* Legible -- return a static pointer to the legibly printed long. */
+
char *
legible (long 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.
+/* Write a string representation of LARGE_INT NUMBER into the provided
+ buffer. The buffer should be able to accept 24 characters,
+ including the terminating zero.
- 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.) */
+ 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 platforms will typically
+ not have snprintf and will use our version, which does support
+ "%lld" where long longs are available. */
static void
-very_long_to_string (char *buffer, VERY_LONG_TYPE number)
+large_int_to_string (char *buffer, LARGE_INT 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';
+ snprintf (buffer, 24, LARGE_INT_FMT, number);
}
-/* The same as legible(), but works on VERY_LONG_TYPE. See sysdep.h. */
+/* The same as legible(), but works on LARGE_INT. */
+
char *
-legible_very_long (VERY_LONG_TYPE l)
+legible_large_int (LARGE_INT l)
{
- char inbuf[128];
- /* Print the number into the buffer. */
- very_long_to_string (inbuf, l);
+ char inbuf[48];
+ large_int_to_string (inbuf, l);
return legible_1 (inbuf);
}
struct wget_timer *
wtimer_allocate (void)
{
- struct wget_timer *wt =
- (struct wget_timer *)xmalloc (sizeof (struct wget_timer));
+ struct wget_timer *wt = xnew (struct wget_timer);
return wt;
}
#endif
#ifdef TIMER_TIME
- /* This is clear. */
return 1000;
#endif
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. */
+ 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. */
int
random_number (int max)
return (int)bounded;
}
+/* 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! */
+
+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;
+}
+
#if 0
/* A debugging function for checking whether an MD5 library works. */
cnt = 16;
while (cnt--)
{
- *p2++ = XDIGIT_TO_xchar (*p1 >> 4);
- *p2++ = XDIGIT_TO_xchar (*p1 & 0xf);
+ *p2++ = XNUM_TO_digit (*p1 >> 4);
+ *p2++ = XNUM_TO_digit (*p1 & 0xf);
++p1;
}
*p2 = '\0';
}
#endif
\f
-/* Implementation of run_with_timeout, a generic timeout handler for
- systems with Unix-like signal handling. */
+/* Implementation of run_with_timeout, a generic timeout-forcing
+ routine for systems with Unix-like signal handling. */
+
#ifdef USE_SIGNAL_TIMEOUT
# ifdef HAVE_SIGSETJMP
# define SETJMP(env) sigsetjmp (env, 1)
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);
+ 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 */
+
+/* Arrange for SIGALRM to be delivered in TIMEOUT seconds. This uses
+ setitimer where available, alarm otherwise.
+
+ TIMEOUT should be non-zero. If the timeout value is so small that
+ it would be rounded to zero, it is rounded to the least legal value
+ instead (1us for setitimer, 1s for alarm). That ensures that
+ SIGALRM will be delivered in all cases. */
+
+static void
+alarm_set (double timeout)
+{
+#ifdef ITIMER_REAL
+ /* Use the modern itimer interface. */
+ struct itimerval itv;
+ xzero (itv);
+ itv.it_value.tv_sec = (long) timeout;
+ itv.it_value.tv_usec = 1000000L * (timeout - (long)timeout);
+ if (itv.it_value.tv_sec == 0 && itv.it_value.tv_usec == 0)
+ /* Ensure that we wait for at least the minimum interval.
+ Specifying zero would mean "wait forever". */
+ itv.it_value.tv_usec = 1;
+ setitimer (ITIMER_REAL, &itv, NULL);
+#else /* not ITIMER_REAL */
+ /* Use the old alarm() interface. */
+ int secs = (int) timeout;
+ if (secs == 0)
+ /* Round TIMEOUTs smaller than 1 to 1, not to zero. This is
+ because alarm(0) means "never deliver the alarm", i.e. "wait
+ forever", which is not what someone who specifies a 0.5s
+ timeout would expect. */
+ secs = 1;
+ alarm (secs);
+#endif /* not ITIMER_REAL */
+}
+
+/* Cancel the alarm set with alarm_set. */
+
+static void
+alarm_cancel (void)
+{
+#ifdef ITIMER_REAL
+ struct itimerval disable;
+ xzero (disable);
+ setitimer (ITIMER_REAL, &disable, NULL);
+#else /* not ITIMER_REAL */
+ alarm (0);
+#endif /* not ITIMER_REAL */
+}
+
+/* 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.
+
+ This works by setting up SIGALRM to be delivered in TIMEOUT seconds
+ using setitimer() or alarm(). The timeout is enforced by
+ longjumping out of the SIGALRM handler. This has several
+ advantages compared to the traditional approach of relying on
+ signals causing system calls to exit with EINTR:
+
+ * The callback function is *forcibly* interrupted after the
+ timeout expires, (almost) regardless of what it was doing and
+ whether it was in a syscall. For example, a calculation that
+ takes a long time is interrupted as reliably as an IO
+ operation.
+
+ * It works with both SYSV and BSD signals because it doesn't
+ depend on the default setting of SA_RESTART.
+
+ * It doesn't special handler setup beyond a simple call to
+ signal(). (It does use sigsetjmp/siglongjmp, but they're
+ optional.)
+
+ The only downside is that, if FUN allocates internal resources that
+ are normally freed prior to exit from the functions, they will be
+ lost in case of timeout. */
int
-run_with_timeout (long timeout, void (*fun) (void *), void *arg)
+run_with_timeout (double timeout, void (*fun) (void *), void *arg)
{
-#ifndef USE_SIGNAL_TIMEOUT
- fun (arg);
- return 0;
-#else
int saved_errno;
if (timeout == 0)
signal (SIGALRM, SIG_DFL);
return 1;
}
- alarm (timeout);
+ alarm_set (timeout);
fun (arg);
/* Preserve errno in case alarm() or signal() modifies it. */
saved_errno = errno;
- alarm (0);
+ alarm_cancel ();
signal (SIGALRM, SIG_DFL);
errno = saved_errno;
return 0;
-#endif
}
+#else /* not USE_SIGNAL_TIMEOUT */
+
+#ifndef WINDOWS
+/* A stub version of run_with_timeout that just calls FUN(ARG). Don't
+ define it under Windows, because Windows has its own version of
+ run_with_timeout that uses threads. */
+
+int
+run_with_timeout (double timeout, void (*fun) (void *), void *arg)
+{
+ fun (arg);
+ return 0;
+}
+#endif /* not WINDOWS */
+#endif /* not USE_SIGNAL_TIMEOUT */
projects / wget / blobdiff