1/* Miscellaneous declarations.
2   Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
3   2005, 2006, 2007, 2008, 2009, 2010, 2011 Free Software Foundation,
4   Inc.
5
6This file is part of GNU Wget.
7
8GNU Wget is free software; you can redistribute it and/or modify
9it under the terms of the GNU General Public License as published by
10the Free Software Foundation; either version 3 of the License, or
11(at your option) any later version.
12
13GNU Wget is distributed in the hope that it will be useful,
14but WITHOUT ANY WARRANTY; without even the implied warranty of
15MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16GNU General Public License for more details.
17
18You should have received a copy of the GNU General Public License
19along with Wget.  If not, see <http://www.gnu.org/licenses/>.
20
21Additional permission under GNU GPL version 3 section 7
22
23If you modify this program, or any covered work, by linking or
24combining it with the OpenSSL project's OpenSSL library (or a
25modified version of that library), containing parts covered by the
26terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
27grants you additional permission to convey the resulting work.
28Corresponding Source for a non-source form of such a combination
29shall include the source code for the parts of OpenSSL used as well
30as that of the covered work.  */
31
32/* This file contains declarations that are universally useful and
33   those that don't fit elsewhere.  It also includes sysdep.h which
34   includes some often-needed system includes, like the obnoxious
35   <time.h> inclusion.  */
36
37#ifndef WGET_H
38#define WGET_H
39
40#include "config.h"
41
42#if ((defined _WIN32 || defined __WIN32__) && !defined __CYGWIN__)
43# define WINDOWS
44#endif
45
46/* Include these, so random files need not include them.  */
47#include "sysdep.h"
48
49/* Disable assertions when debug support is not compiled in. */
50#ifndef ENABLE_DEBUG
51# define NDEBUG
52#endif
53
54/* Is OpenSSL or GNUTLS available? */
55#if defined HAVE_LIBSSL || defined HAVE_LIBSSL32 || defined HAVE_LIBGNUTLS
56# define HAVE_SSL
57#endif
58
59/* `gettext (FOO)' is long to write, so we use `_(FOO)'.  If NLS is
60   unavailable, _(STRING) simply returns STRING.  */
61#include "gettext.h"
62#define _(string)   gettext (string)
63
64/* A pseudo function call that serves as a marker for the automated
65   extraction of messages, but does not call gettext().  The run-time
66   translation is done at a different place in the code.  The purpose
67   of the N_("...") call is to make the message snarfer aware that the
68   "..." string needs to be translated.  STRING should be a string
69   literal.  Concatenated strings and other string expressions won't
70   work.  The macro's expansion is not parenthesized, so that it is
71   suitable as initializer for static 'char[]' or 'const char[]'
72   variables.  -- explanation partly taken from GNU make.  */
73#define N_(string) string
74
75#if HAVE_WCWIDTH && HAVE_MBTOWC
76# define USE_NLS_PROGRESS_BAR 1
77#else
78/* Just to be a little paranoid about it. */
79# undef  USE_NLS_PROGRESS_BAR
80#endif
81
82/* I18N NOTE: You will notice that none of the DEBUGP messages are
83   marked as translatable.  This is intentional, for a few reasons:
84
85   1) The debug messages are not meant for the users to look at, but
86   for the developers; as such, they should be considered more like
87   source comments than real program output.
88
89   2) The messages are numerous, and yet they are random and frivolous
90   ("double yuck!" and such).  There would be a lot of work with no
91   gain.
92
93   3) Finally, the debug messages are meant to be a clue for me to
94   debug problems with Wget.  If I get them in a language I don't
95   understand, debugging will become a new challenge of its own!  */
96
97/* locale independent replacement for ctype.h */
98#include "c-ctype.h"
99
100/* Conditionalize the use of GCC's __attribute__((format)) and
101   __builtin_expect features using macros.  */
102
103#if defined(__GNUC__) && __GNUC__ >= 3
104# define GCC_FORMAT_ATTR(a, b) __attribute__ ((format (printf, a, b)))
105# define LIKELY(exp)   __builtin_expect (!!(exp), 1)
106# define UNLIKELY(exp) __builtin_expect ((exp), 0)
107#else
108# define GCC_FORMAT_ATTR(a, b)
109# define LIKELY(exp)   (exp)
110# define UNLIKELY(exp) (exp)
111#endif
112
113/* Execute the following statement if debugging is both enabled at
114   compile-time and requested at run-time; a no-op otherwise.  */
115
116#ifdef ENABLE_DEBUG
117# define IF_DEBUG if (UNLIKELY (opt.debug))
118#else
119# define IF_DEBUG if (0)
120#endif
121
122/* Print ARGS if debugging is enabled and requested, otherwise do
123   nothing.  This must be called with an extra level of parentheses
124   because it's not possible to pass a variable number of arguments to
125   a macro (in portable C89).  ARGS are like arguments to printf.  */
126
127#define DEBUGP(args) do { IF_DEBUG { debug_logprintf args; } } while (0)
128
129/* Pick an integer type large enough for file sizes, content lengths,
130   and such.  Because today's files can be very large, it should be a
131   signed integer at least 64 bits wide.  This can't be typedeffed to
132   off_t because: a) off_t is always 32-bit on Windows, and b) we
133   don't necessarily want to tie having a 64-bit type for internal
134   calculations to having LFS support.  */
135
136#ifdef WINDOWS
137  /* nothing to do, see mswindows.h */
138#elif SIZEOF_LONG >= 8
139  /* long is large enough, so use it. */
140  typedef long wgint;
141# define SIZEOF_WGINT SIZEOF_LONG
142#elif SIZEOF_LONG_LONG >= 8
143  /* long long is large enough and available, use that */
144  typedef long long wgint;
145# define SIZEOF_WGINT SIZEOF_LONG_LONG
146#elif HAVE_INT64_T
147  typedef int64_t wgint;
148# define SIZEOF_WGINT 8
149#elif SIZEOF_OFF_T >= 8
150  /* In case off_t is typedeffed to a large non-standard type that our
151     tests don't find. */
152  typedef off_t wgint;
153# define SIZEOF_WGINT SIZEOF_OFF_T
154#else
155  /* Fall back to using long, which is always available and in most
156     cases large enough. */
157  typedef long wgint;
158# define SIZEOF_WGINT SIZEOF_LONG
159#endif
160
161/* Pick a strtol-compatible function that will work with wgint.  The
162   choices are strtol, strtoll, or our own implementation of strtoll
163   in cmpt.c, activated with NEED_STRTOLL.  */
164
165#ifdef WINDOWS
166  /* nothing to do, see mswindows.h */
167#elif SIZEOF_WGINT == SIZEOF_LONG
168# define str_to_wgint strtol
169#elif SIZEOF_WGINT == SIZEOF_LONG_LONG
170# define str_to_wgint strtoll
171# ifndef HAVE_STRTOLL
172#  define NEED_STRTOLL
173#  define strtoll_type long long
174# endif
175#else
176  /* wgint has a strange size; synthesize strtoll and use it. */
177# define str_to_wgint strtoll
178# define NEED_STRTOLL
179# define strtoll_type wgint
180#endif
181
182#define WGINT_MAX TYPE_MAXIMUM (wgint)
183
184/* Declare our strtoll replacement. */
185#ifdef NEED_STRTOLL
186strtoll_type strtoll (const char *, char **, int);
187#endif
188
189/* Now define a large numeric type useful for storing sizes of *sums*
190   of downloads, such as the value of the --quota option.  This should
191   be a type able to hold 2G+ values even on systems without large
192   file support.  (It is useful to limit Wget's download quota to say
193   10G even if a single file cannot be that large.)
194
195   To make sure we get the largest size possible, we use `double' on
196   systems without a 64-bit integral type.  (Since it is used in very
197   few places in Wget, this is acceptable.)  */
198
199#if SIZEOF_WGINT >= 8
200/* just use wgint */
201typedef wgint SUM_SIZE_INT;
202#else
203/* On systems without LFS, use double, which buys us integers up to 2^53. */
204typedef double SUM_SIZE_INT;
205#endif
206
207#include "options.h"
208
209/* Everything uses this, so include them here directly.  */
210#include <alloca.h>
211#include "xalloc.h"
212
213/* Likewise for logging functions.  */
214#include "log.h"
215
216/* Likewise for quoting functions.  */
217#include "quote.h"
218#include "quotearg.h"
219
220/* Likewise for struct iri definition */
221#include "iri.h"
222
223/* Useful macros used across the code: */
224
225/* The number of elements in an array.  For example:
226   static char a[] = "foo";     -- countof(a) == 4 (note terminating \0)
227   int a[5] = {1, 2};           -- countof(a) == 5
228   char *a[] = {                -- countof(a) == 3
229     "foo", "bar", "baz"
230   }; */
231#define countof(array) (sizeof (array) / sizeof ((array)[0]))
232
233/* Zero out a value.  */
234#define xzero(x) memset (&(x), '\0', sizeof (x))
235
236/* Convert an ASCII hex digit to the corresponding number between 0
237   and 15.  H should be a hexadecimal digit that satisfies isxdigit;
238   otherwise, the result is undefined.  */
239#define XDIGIT_TO_NUM(h) ((h) < 'A' ? (h) - '0' : c_toupper (h) - 'A' + 10)
240#define X2DIGITS_TO_NUM(h1, h2) ((XDIGIT_TO_NUM (h1) << 4) + XDIGIT_TO_NUM (h2))
241
242/* The reverse of the above: convert a number in the [0, 16) range to
243   the ASCII representation of the corresponding hexadecimal digit.
244   `+ 0' is there so you can't accidentally use it as an lvalue.  */
245#define XNUM_TO_DIGIT(x) ("0123456789ABCDEF"[x] + 0)
246#define XNUM_TO_digit(x) ("0123456789abcdef"[x] + 0)
247
248/* Copy the data delimited with BEG and END to alloca-allocated
249   storage, and zero-terminate it.  Arguments are evaluated only once,
250   in the order BEG, END, PLACE.  */
251#define BOUNDED_TO_ALLOCA(beg, end, place) do { \
252  const char *BTA_beg = (beg);                  \
253  int BTA_len = (end) - BTA_beg;                \
254  char **BTA_dest = &(place);                   \
255  *BTA_dest = alloca (BTA_len + 1);             \
256  memcpy (*BTA_dest, BTA_beg, BTA_len);         \
257  (*BTA_dest)[BTA_len] = '\0';                  \
258} while (0)
259
260/* Return non-zero if string bounded between BEG and END is equal to
261   STRING_LITERAL.  The comparison is case-sensitive.  */
262#define BOUNDED_EQUAL(beg, end, string_literal)             \
263  ((end) - (beg) == sizeof (string_literal) - 1             \
264   && !memcmp (beg, string_literal, sizeof (string_literal) - 1))
265
266/* The same as above, except the comparison is case-insensitive. */
267#define BOUNDED_EQUAL_NO_CASE(beg, end, string_literal)         \
268  ((end) - (beg) == sizeof (string_literal) - 1                 \
269   && !strncasecmp (beg, string_literal, sizeof (string_literal) - 1))
270
271/* Like ptr=strdup(str), but allocates the space for PTR on the stack.
272   This cannot be an expression because this is not portable:
273     #define STRDUP_ALLOCA(str) (strcpy (alloca (strlen (str) + 1), str))
274   The problem is that some compilers can't handle alloca() being an
275   argument to a function.  */
276
277#define STRDUP_ALLOCA(ptr, str) do {                \
278  char **SA_dest = &(ptr);                          \
279  const char *SA_src = (str);                       \
280  *SA_dest = (char *)alloca (strlen (SA_src) + 1);  \
281  strcpy (*SA_dest, SA_src);                        \
282} while (0)
283
284/* Generally useful if you want to avoid arbitrary size limits but
285   don't need a full dynamic array.  Assumes that BASEVAR points to a
286   malloced array of TYPE objects (or possibly a NULL pointer, if
287   SIZEVAR is 0), with the total size stored in SIZEVAR.  This macro
288   will realloc BASEVAR as necessary so that it can hold at least
289   NEEDED_SIZE objects.  The reallocing is done by doubling, which
290   ensures constant amortized time per element.  */
291
292#define DO_REALLOC(basevar, sizevar, needed_size, type) do {    \
293  long DR_needed_size = (needed_size);                          \
294  long DR_newsize = 0;                                          \
295  while ((sizevar) < (DR_needed_size)) {                        \
296    DR_newsize = sizevar << 1;                                  \
297    if (DR_newsize < 16)                                        \
298      DR_newsize = 16;                                          \
299    (sizevar) = DR_newsize;                                     \
300  }                                                             \
301  if (DR_newsize)                                               \
302    basevar = xrealloc (basevar, DR_newsize * sizeof (type));   \
303} while (0)
304
305/* Used to print pointers (usually for debugging).  Print pointers
306   using printf ("0x%0*lx", PTR_FORMAT (p)).  (%p is too unpredictable;
307   some implementations prepend 0x, while some don't, and most don't
308   0-pad the address.)  */
309#define PTR_FORMAT(p) (int) (2 * sizeof (void *)), (unsigned long) (p)
310
311/* Find the maximum buffer length needed to print an integer of type `x'
312   in base 10. 24082 / 10000 = 8*log_{10}(2).  */
313#define MAX_INT_TO_STRING_LEN(x) ((sizeof(x) * 24082 / 10000) + 2)
314
315extern const char *exec_name;
316
317/* Document type ("dt") flags */
318enum
319{
320  TEXTHTML             = 0x0001,        /* document is of type text/html
321                                           or application/xhtml+xml */
322  RETROKF              = 0x0002,        /* retrieval was OK */
323  HEAD_ONLY            = 0x0004,        /* only send the HEAD request */
324  SEND_NOCACHE         = 0x0008,        /* send Pragma: no-cache directive */
325  ACCEPTRANGES         = 0x0010,        /* Accept-ranges header was found */
326  ADDED_HTML_EXTENSION = 0x0020,        /* added ".html" extension due to -E */
327  TEXTCSS              = 0x0040         /* document is of type text/css */
328};
329
330/* Universal error type -- used almost everywhere.  Error reporting of
331   this detail is not generally used or needed and should be
332   simplified.  */
333typedef enum
334{
335  NOCONERROR, HOSTERR, CONSOCKERR, CONERROR, CONSSLERR,
336  CONIMPOSSIBLE, NEWLOCATION,
337  FTPOK, FTPLOGINC, FTPLOGREFUSED, FTPPORTERR, FTPSYSERR,
338  FTPNSFOD, FTPUNKNOWNTYPE, FTPRERR,
339  FTPSRVERR, FTPRETRINT, FTPRESTFAIL, URLERROR, FOPENERR,
340  FOPEN_EXCL_ERR, FWRITEERR, HEOF,
341  HERR, RETROK, RECLEVELEXC, WRONGCODE,
342  FTPINVPASV, FTPNOPASV, CONTNOTSUPPORTED, RETRUNNEEDED, RETRFINISHED,
343  READERR, TRYLIMEXC, FILEBADFILE, RANGEERR,
344  RETRBADPATTERN, PROXERR,
345  AUTHFAILED, QUOTEXC, WRITEFAILED, SSLINITFAILED, VERIFCERTERR,
346  UNLINKERR, NEWLOCATION_KEEP_POST, CLOSEFAILED, ATTRMISSING, UNKNOWNATTR,
347  WARC_ERR, WARC_TMP_FOPENERR, WARC_TMP_FWRITEERR
348} uerr_t;
349
350/* 2005-02-19 SMS.
351   Select an appropriate "orig" suffix and a separator character for
352   adding a unique suffix to a file name.
353
354   A VMS ODS2 file system can't tolerate multiple dots.  An ODS5 file
355   system can, but even there not all dots are equal, and heroic effort
356   would be needed to get ".html^.orig" rather than (the less desirable)
357   "^.html.orig".  It's more satisfactory always to use "_orig" on VMS
358   (rather than including "vms.h", testing "ods5_dest", and acting
359   accordingly).
360
361   Note that code in various places assumes that this string is five
362   characters long.
363*/
364# ifdef __VMS
365#  define ORIG_SFX "_orig"
366# else /* def __VMS */
367#  define ORIG_SFX ".orig"
368# endif /* def __VMS [else] */
369
370/* ".NNN" unique-ifying suffix separator character for unique_name() in
371   url.c (and anywhere else).  Note that on VMS, the file system's
372   version numbers solve the problem that unique_name() is designed to
373   handle, obviating this whole exercise.  Other systems may specify a
374   character different from "." here, if desired.
375*/
376# ifndef __VMS
377#  define UNIQ_SEP '.'
378# endif /* ndef __VMS */
379
380#endif /* WGET_H */
381