1 /* Various declarations for language-independent pretty-print subroutines.
2 Copyright (C) 2003, 2004, 2005, 2007, 2008, 2009, 2010
3 Free Software Foundation, Inc.
4 Contributed by Gabriel Dos Reis <gdr@integrable-solutions.net>
6 This file is part of GCC.
8 GCC is free software; you can redistribute it and/or modify it under
9 the terms of the GNU General Public License as published by the Free
10 Software Foundation; either version 3, or (at your option) any later
13 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
14 WARRANTY; without even the implied warranty of MERCHANTABILITY or
15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
18 You should have received a copy of the GNU General Public License
19 along with GCC; see the file COPYING3. If not see
20 <http://www.gnu.org/licenses/>. */
24 #include "coretypes.h"
26 #include "pretty-print.h"
33 #define obstack_chunk_alloc xmalloc
34 #define obstack_chunk_free free
36 /* A pointer to the formatted diagnostic message. */
37 #define pp_formatted_text_data(PP) \
38 ((const char *) obstack_base (pp_base (PP)->buffer->obstack))
40 /* Format an integer given by va_arg (ARG, type-specifier T) where
41 type-specifier is a precision modifier as indicated by PREC. F is
42 a string used to construct the appropriate format-specifier. */
43 #define pp_integer_with_precision(PP, ARG, PREC, T, F) \
48 pp_scalar (PP, "%" F, va_arg (ARG, T)); \
52 pp_scalar (PP, "%l" F, va_arg (ARG, long T)); \
56 pp_scalar (PP, "%" HOST_LONG_LONG_FORMAT F, va_arg (ARG, long long T)); \
65 /* Subroutine of pp_set_maximum_length. Set up PRETTY-PRINTER's
66 internal maximum characters per line. */
68 pp_set_real_maximum_length (pretty_printer *pp)
70 /* If we're told not to wrap lines then do the obvious thing. In case
71 we'll emit prefix only once per message, it is appropriate
72 not to increase unnecessarily the line-length cut-off. */
73 if (!pp_is_wrapping_line (pp)
74 || pp_prefixing_rule (pp) == DIAGNOSTICS_SHOW_PREFIX_ONCE
75 || pp_prefixing_rule (pp) == DIAGNOSTICS_SHOW_PREFIX_NEVER)
76 pp->maximum_length = pp_line_cutoff (pp);
79 int prefix_length = pp->prefix ? strlen (pp->prefix) : 0;
80 /* If the prefix is ridiculously too long, output at least
82 if (pp_line_cutoff (pp) - prefix_length < 32)
83 pp->maximum_length = pp_line_cutoff (pp) + 32;
85 pp->maximum_length = pp_line_cutoff (pp);
89 /* Clear PRETTY-PRINTER's output state. */
91 pp_clear_state (pretty_printer *pp)
93 pp->emitted_prefix = false;
94 pp_indentation (pp) = 0;
97 /* Flush the formatted text of PRETTY-PRINTER onto the attached stream. */
99 pp_write_text_to_stream (pretty_printer *pp)
101 const char *text = pp_formatted_text (pp);
102 fputs (text, pp->buffer->stream);
103 pp_clear_output_area (pp);
106 /* Wrap a text delimited by START and END into PRETTY-PRINTER. */
108 pp_wrap_text (pretty_printer *pp, const char *start, const char *end)
110 bool wrapping_line = pp_is_wrapping_line (pp);
114 /* Dump anything bordered by whitespaces. */
116 const char *p = start;
117 while (p != end && !ISBLANK (*p) && *p != '\n')
120 && p - start >= pp_remaining_character_count_for_line (pp))
122 pp_append_text (pp, start, p);
126 if (start != end && ISBLANK (*start))
131 if (start != end && *start == '\n')
139 /* Same as pp_wrap_text but wrap text only when in line-wrapping mode. */
141 pp_maybe_wrap_text (pretty_printer *pp, const char *start, const char *end)
143 if (pp_is_wrapping_line (pp))
144 pp_wrap_text (pp, start, end);
146 pp_append_text (pp, start, end);
149 /* Append to the output area of PRETTY-PRINTER a string specified by its
150 STARTing character and LENGTH. */
152 pp_append_r (pretty_printer *pp, const char *start, int length)
154 obstack_grow (pp->buffer->obstack, start, length);
155 pp->buffer->line_length += length;
158 /* Insert enough spaces into the output area of PRETTY-PRINTER to bring
159 the column position to the current indentation level, assuming that a
160 newline has just been written to the buffer. */
162 pp_base_indent (pretty_printer *pp)
164 int n = pp_indentation (pp);
167 for (i = 0; i < n; ++i)
171 /* The following format specifiers are recognized as being client independent:
172 %d, %i: (signed) integer in base ten.
173 %u: unsigned integer in base ten.
174 %o: unsigned integer in base eight.
175 %x: unsigned integer in base sixteen.
176 %ld, %li, %lo, %lu, %lx: long versions of the above.
177 %lld, %lli, %llo, %llu, %llx: long long versions.
178 %wd, %wi, %wo, %wu, %wx: HOST_WIDE_INT versions.
182 %m: strerror(text->err_no) - does not consume a value from args_ptr.
186 %': apostrophe (should only be used in untranslated messages;
187 translations should use appropriate punctuation directly).
188 %.*s: a substring the length of which is specified by an argument
190 %Ns: likewise, but length specified as constant in the format string.
191 Flag 'q': quote formatted text (must come immediately after '%').
193 Arguments can be used sequentially, or through %N$ resp. *N$
194 notation Nth argument after the format string. If %N$ / *N$
195 notation is used, it must be used for all arguments, except %m, %%,
196 %<, %> and %', which may not have a number, as they do not consume
197 an argument. When %M$.*N$s is used, M must be N + 1. (This may
198 also be written %M$.*s, provided N is not otherwise used.) The
199 format string must have conversion specifiers with argument numbers
200 1 up to highest argument; each argument may only be used once.
201 A format string can have at most 30 arguments. */
203 /* Formatting phases 1 and 2: render TEXT->format_spec plus
204 TEXT->args_ptr into a series of chunks in PP->buffer->args[].
205 Phase 3 is in pp_base_format_text. */
208 pp_base_format (pretty_printer *pp, text_info *text)
210 output_buffer *buffer = pp->buffer;
213 struct chunk_info *new_chunk_array;
215 unsigned int curarg = 0, chunk = 0, argno;
216 pp_wrapping_mode_t old_wrapping_mode;
217 bool any_unnumbered = false, any_numbered = false;
218 const char **formatters[PP_NL_ARGMAX];
220 /* Allocate a new chunk structure. */
221 new_chunk_array = XOBNEW (&buffer->chunk_obstack, struct chunk_info);
222 new_chunk_array->prev = buffer->cur_chunk_array;
223 buffer->cur_chunk_array = new_chunk_array;
224 args = new_chunk_array->args;
226 /* Formatting phase 1: split up TEXT->format_spec into chunks in
227 PP->buffer->args[]. Even-numbered chunks are to be output
228 verbatim, odd-numbered chunks are format specifiers.
229 %m, %%, %<, %>, and %' are replaced with the appropriate text at
232 memset (formatters, 0, sizeof formatters);
234 for (p = text->format_spec; *p; )
236 while (*p != '\0' && *p != '%')
238 obstack_1grow (&buffer->chunk_obstack, *p);
251 obstack_1grow (&buffer->chunk_obstack, '%');
256 obstack_grow (&buffer->chunk_obstack,
257 open_quote, strlen (open_quote));
263 obstack_grow (&buffer->chunk_obstack,
264 close_quote, strlen (close_quote));
270 const char *errstr = xstrerror (text->err_no);
271 obstack_grow (&buffer->chunk_obstack, errstr, strlen (errstr));
277 /* Handled in phase 2. Terminate the plain chunk here. */
278 obstack_1grow (&buffer->chunk_obstack, '\0');
279 gcc_assert (chunk < PP_NL_ARGMAX * 2);
280 args[chunk++] = XOBFINISH (&buffer->chunk_obstack, const char *);
287 argno = strtoul (p, &end, 10) - 1;
289 gcc_assert (*p == '$');
293 gcc_assert (!any_unnumbered);
298 any_unnumbered = true;
299 gcc_assert (!any_numbered);
301 gcc_assert (argno < PP_NL_ARGMAX);
302 gcc_assert (!formatters[argno]);
303 formatters[argno] = &args[chunk];
306 obstack_1grow (&buffer->chunk_obstack, *p);
309 while (strchr ("qwl+#", p[-1]));
313 /* We handle '%.Ns' and '%.*s' or '%M$.*N$s'
314 (where M == N + 1). */
319 obstack_1grow (&buffer->chunk_obstack, *p);
322 while (ISDIGIT (p[-1]));
323 gcc_assert (p[-1] == 's');
327 gcc_assert (*p == '*');
328 obstack_1grow (&buffer->chunk_obstack, '*');
334 unsigned int argno2 = strtoul (p, &end, 10) - 1;
336 gcc_assert (argno2 == argno - 1);
337 gcc_assert (!any_unnumbered);
338 gcc_assert (*p == '$');
341 formatters[argno2] = formatters[argno];
345 gcc_assert (!any_numbered);
346 formatters[argno+1] = formatters[argno];
349 gcc_assert (*p == 's');
350 obstack_1grow (&buffer->chunk_obstack, 's');
357 obstack_1grow (&buffer->chunk_obstack, '\0');
358 gcc_assert (chunk < PP_NL_ARGMAX * 2);
359 args[chunk++] = XOBFINISH (&buffer->chunk_obstack, const char *);
362 obstack_1grow (&buffer->chunk_obstack, '\0');
363 gcc_assert (chunk < PP_NL_ARGMAX * 2);
364 args[chunk++] = XOBFINISH (&buffer->chunk_obstack, const char *);
367 /* Set output to the argument obstack, and switch line-wrapping and
369 buffer->obstack = &buffer->chunk_obstack;
370 old_wrapping_mode = pp_set_verbatim_wrapping (pp);
372 /* Second phase. Replace each formatter with the formatted text it
375 for (argno = 0; formatters[argno]; argno++)
383 /* We do not attempt to enforce any ordering on the modifier
386 for (p = *formatters[argno];; p++)
411 /* We don't support precision beyond that of "long long". */
412 gcc_assert (precision < 2);
419 gcc_assert (!wide || precision == 0);
422 pp_string (pp, open_quote);
427 pp_character (pp, va_arg (*text->args_ptr, int));
433 pp_wide_integer (pp, va_arg (*text->args_ptr, HOST_WIDE_INT));
435 pp_integer_with_precision
436 (pp, *text->args_ptr, precision, int, "d");
441 pp_scalar (pp, "%" HOST_WIDE_INT_PRINT "o",
442 va_arg (*text->args_ptr, unsigned HOST_WIDE_INT));
444 pp_integer_with_precision
445 (pp, *text->args_ptr, precision, unsigned, "o");
449 pp_string (pp, va_arg (*text->args_ptr, const char *));
453 pp_pointer (pp, va_arg (*text->args_ptr, void *));
458 pp_scalar (pp, HOST_WIDE_INT_PRINT_UNSIGNED,
459 va_arg (*text->args_ptr, unsigned HOST_WIDE_INT));
461 pp_integer_with_precision
462 (pp, *text->args_ptr, precision, unsigned, "u");
467 pp_scalar (pp, HOST_WIDE_INT_PRINT_HEX,
468 va_arg (*text->args_ptr, unsigned HOST_WIDE_INT));
470 pp_integer_with_precision
471 (pp, *text->args_ptr, precision, unsigned, "x");
479 /* We handle '%.Ns' and '%.*s' or '%M$.*N$s'
480 (where M == N + 1). The format string should be verified
481 already from the first phase. */
486 n = strtoul (p, &end, 10);
488 gcc_assert (*p == 's');
492 gcc_assert (*p == '*');
494 gcc_assert (*p == 's');
495 n = va_arg (*text->args_ptr, int);
497 /* This consumes a second entry in the formatters array. */
498 gcc_assert (formatters[argno] == formatters[argno+1]);
502 s = va_arg (*text->args_ptr, const char *);
503 pp_append_text (pp, s, s + n);
511 gcc_assert (pp_format_decoder (pp));
512 ok = pp_format_decoder (pp) (pp, text, p,
513 precision, wide, plus, hash);
519 pp_string (pp, close_quote);
521 obstack_1grow (&buffer->chunk_obstack, '\0');
522 *formatters[argno] = XOBFINISH (&buffer->chunk_obstack, const char *);
525 #ifdef ENABLE_CHECKING
526 for (; argno < PP_NL_ARGMAX; argno++)
527 gcc_assert (!formatters[argno]);
530 /* Revert to normal obstack and wrapping mode. */
531 buffer->obstack = &buffer->formatted_obstack;
532 buffer->line_length = 0;
533 pp_wrapping_mode (pp) = old_wrapping_mode;
537 /* Format of a message pointed to by TEXT. */
539 pp_base_output_formatted_text (pretty_printer *pp)
542 output_buffer *buffer = pp_buffer (pp);
543 struct chunk_info *chunk_array = buffer->cur_chunk_array;
544 const char **args = chunk_array->args;
546 gcc_assert (buffer->obstack == &buffer->formatted_obstack);
547 gcc_assert (buffer->line_length == 0);
549 /* This is a third phase, first 2 phases done in pp_base_format_args.
550 Now we actually print it. */
551 for (chunk = 0; args[chunk]; chunk++)
552 pp_string (pp, args[chunk]);
554 /* Deallocate the chunk structure and everything after it (i.e. the
555 associated series of formatted strings). */
556 buffer->cur_chunk_array = chunk_array->prev;
557 obstack_free (&buffer->chunk_obstack, chunk_array);
560 /* Helper subroutine of output_verbatim and verbatim. Do the appropriate
561 settings needed by BUFFER for a verbatim formatting. */
563 pp_base_format_verbatim (pretty_printer *pp, text_info *text)
565 /* Set verbatim mode. */
566 pp_wrapping_mode_t oldmode = pp_set_verbatim_wrapping (pp);
568 /* Do the actual formatting. */
569 pp_format (pp, text);
570 pp_output_formatted_text (pp);
572 /* Restore previous settings. */
573 pp_wrapping_mode (pp) = oldmode;
576 /* Flush the content of BUFFER onto the attached stream. */
578 pp_base_flush (pretty_printer *pp)
580 pp_write_text_to_stream (pp);
582 fputc ('\n', pp->buffer->stream);
583 fflush (pp->buffer->stream);
584 pp_needs_newline (pp) = false;
587 /* Sets the number of maximum characters per line PRETTY-PRINTER can
588 output in line-wrapping mode. A LENGTH value 0 suppresses
591 pp_base_set_line_maximum_length (pretty_printer *pp, int length)
593 pp_line_cutoff (pp) = length;
594 pp_set_real_maximum_length (pp);
597 /* Clear PRETTY-PRINTER output area text info. */
599 pp_base_clear_output_area (pretty_printer *pp)
601 obstack_free (pp->buffer->obstack, obstack_base (pp->buffer->obstack));
602 pp->buffer->line_length = 0;
605 /* Set PREFIX for PRETTY-PRINTER. */
607 pp_base_set_prefix (pretty_printer *pp, const char *prefix)
610 pp_set_real_maximum_length (pp);
611 pp->emitted_prefix = false;
612 pp_indentation (pp) = 0;
615 /* Free PRETTY-PRINTER's prefix, a previously malloc()'d string. */
617 pp_base_destroy_prefix (pretty_printer *pp)
619 if (pp->prefix != NULL)
621 free (CONST_CAST (char *, pp->prefix));
626 /* Write out PRETTY-PRINTER's prefix. */
628 pp_base_emit_prefix (pretty_printer *pp)
630 if (pp->prefix != NULL)
632 switch (pp_prefixing_rule (pp))
635 case DIAGNOSTICS_SHOW_PREFIX_NEVER:
638 case DIAGNOSTICS_SHOW_PREFIX_ONCE:
639 if (pp->emitted_prefix)
644 pp_indentation (pp) += 3;
647 case DIAGNOSTICS_SHOW_PREFIX_EVERY_LINE:
649 int prefix_length = strlen (pp->prefix);
650 pp_append_r (pp, pp->prefix, prefix_length);
651 pp->emitted_prefix = true;
658 /* Construct a PRETTY-PRINTER with PREFIX and of MAXIMUM_LENGTH
659 characters per line. */
661 pp_construct (pretty_printer *pp, const char *prefix, int maximum_length)
663 memset (pp, 0, sizeof (pretty_printer));
664 pp->buffer = XCNEW (output_buffer);
665 obstack_init (&pp->buffer->chunk_obstack);
666 obstack_init (&pp->buffer->formatted_obstack);
667 pp->buffer->obstack = &pp->buffer->formatted_obstack;
668 pp->buffer->stream = stderr;
669 pp_line_cutoff (pp) = maximum_length;
670 pp_prefixing_rule (pp) = DIAGNOSTICS_SHOW_PREFIX_ONCE;
671 pp_set_prefix (pp, prefix);
672 pp_translate_identifiers (pp) = true;
675 /* Append a string delimited by START and END to the output area of
676 PRETTY-PRINTER. No line wrapping is done. However, if beginning a
677 new line then emit PRETTY-PRINTER's prefix and skip any leading
678 whitespace if appropriate. The caller must ensure that it is
681 pp_base_append_text (pretty_printer *pp, const char *start, const char *end)
683 /* Emit prefix and skip whitespace if we're starting a new line. */
684 if (pp->buffer->line_length == 0)
687 if (pp_is_wrapping_line (pp))
688 while (start != end && *start == ' ')
691 pp_append_r (pp, start, end - start);
694 /* Finishes constructing a NULL-terminated character string representing
695 the PRETTY-PRINTED text. */
697 pp_base_formatted_text (pretty_printer *pp)
699 obstack_1grow (pp->buffer->obstack, '\0');
700 return pp_formatted_text_data (pp);
703 /* Return a pointer to the last character emitted in PRETTY-PRINTER's
704 output area. A NULL pointer means no character available. */
706 pp_base_last_position_in_text (const pretty_printer *pp)
708 const char *p = NULL;
709 struct obstack *text = pp->buffer->obstack;
711 if (obstack_base (text) != obstack_next_free (text))
712 p = ((const char *) obstack_next_free (text)) - 1;
716 /* Return the amount of characters PRETTY-PRINTER can accept to
717 make a full line. Meaningful only in line-wrapping mode. */
719 pp_base_remaining_character_count_for_line (pretty_printer *pp)
721 return pp->maximum_length - pp->buffer->line_length;
725 /* Format a message into BUFFER a la printf. */
727 pp_printf (pretty_printer *pp, const char *msg, ...)
735 text.format_spec = msg;
737 pp_format (pp, &text);
738 pp_output_formatted_text (pp);
743 /* Output MESSAGE verbatim into BUFFER. */
745 pp_verbatim (pretty_printer *pp, const char *msg, ...)
753 text.format_spec = msg;
755 pp_format_verbatim (pp, &text);
761 /* Have PRETTY-PRINTER start a new line. */
763 pp_base_newline (pretty_printer *pp)
765 obstack_1grow (pp->buffer->obstack, '\n');
766 pp->buffer->line_length = 0;
769 /* Have PRETTY-PRINTER add a CHARACTER. */
771 pp_base_character (pretty_printer *pp, int c)
773 if (pp_is_wrapping_line (pp)
774 && pp_remaining_character_count_for_line (pp) <= 0)
780 obstack_1grow (pp->buffer->obstack, c);
781 ++pp->buffer->line_length;
784 /* Append a STRING to the output area of PRETTY-PRINTER; the STRING may
785 be line-wrapped if in appropriate mode. */
787 pp_base_string (pretty_printer *pp, const char *str)
789 pp_maybe_wrap_text (pp, str, str + (str ? strlen (str) : 0));
792 /* Maybe print out a whitespace if needed. */
795 pp_base_maybe_space (pretty_printer *pp)
797 if (pp_base (pp)->padding != pp_none)
800 pp_base (pp)->padding = pp_none;
804 /* The string starting at P has LEN (at least 1) bytes left; if they
805 start with a valid UTF-8 sequence, return the length of that
806 sequence and set *VALUE to the value of that sequence, and
807 otherwise return 0 and set *VALUE to (unsigned int) -1. */
810 decode_utf8_char (const unsigned char *p, size_t len, unsigned int *value)
821 for (t = *p; t & 0x80; t <<= 1)
824 if (utf8_len > len || utf8_len < 2 || utf8_len > 6)
826 *value = (unsigned int) -1;
829 ch = *p & ((1 << (7 - utf8_len)) - 1);
830 for (i = 1; i < utf8_len; i++)
832 unsigned int u = p[i];
833 if ((u & 0xC0) != 0x80)
835 *value = (unsigned int) -1;
838 ch = (ch << 6) | (u & 0x3F);
840 if ( (ch <= 0x7F && utf8_len > 1)
841 || (ch <= 0x7FF && utf8_len > 2)
842 || (ch <= 0xFFFF && utf8_len > 3)
843 || (ch <= 0x1FFFFF && utf8_len > 4)
844 || (ch <= 0x3FFFFFF && utf8_len > 5)
845 || (ch >= 0xD800 && ch <= 0xDFFF))
847 *value = (unsigned int) -1;
860 /* Given IDENT, an identifier in the internal encoding, return a
861 version of IDENT suitable for diagnostics in the locale character
862 set: either IDENT itself, or a garbage-collected string converted
863 to the locale character set and using escape sequences if not
864 representable in the locale character set or containing control
865 characters or invalid byte sequences. Existing backslashes in
866 IDENT are not doubled, so the result may not uniquely specify the
867 contents of an arbitrary byte sequence identifier. */
870 identifier_to_locale (const char *ident)
872 const unsigned char *uid = (const unsigned char *) ident;
873 size_t idlen = strlen (ident);
874 bool valid_printable_utf8 = true;
875 bool all_ascii = true;
878 for (i = 0; i < idlen;)
881 size_t utf8_len = decode_utf8_char (&uid[i], idlen - i, &c);
882 if (utf8_len == 0 || c <= 0x1F || (c >= 0x7F && c <= 0x9F))
884 valid_printable_utf8 = false;
892 /* If IDENT contains invalid UTF-8 sequences (which may occur with
893 attributes putting arbitrary byte sequences in identifiers), or
894 control characters, we use octal escape sequences for all bytes
895 outside printable ASCII. */
896 if (!valid_printable_utf8)
898 char *ret = GGC_NEWVEC (char, 4 * idlen + 1);
900 for (i = 0; i < idlen; i++)
902 if (uid[i] > 0x1F && uid[i] < 0x7F)
906 sprintf (p, "\\%03o", uid[i]);
914 /* Otherwise, if it is valid printable ASCII, or printable UTF-8
915 with the locale character set being UTF-8, IDENT is used. */
916 if (all_ascii || locale_utf8)
919 /* Otherwise IDENT is converted to the locale character set if
921 #if defined ENABLE_NLS && defined HAVE_LANGINFO_CODESET && HAVE_ICONV
922 if (locale_encoding != NULL)
924 iconv_t cd = iconv_open (locale_encoding, "UTF-8");
925 bool conversion_ok = true;
927 if (cd != (iconv_t) -1)
929 size_t ret_alloc = 4 * idlen + 1;
932 /* Repeat the whole conversion process as needed with
933 larger buffers so non-reversible transformations can
934 always be detected. */
935 ICONV_CONST char *inbuf = CONST_CAST (char *, ident);
937 size_t inbytesleft = idlen;
938 size_t outbytesleft = ret_alloc - 1;
941 ret = GGC_NEWVEC (char, ret_alloc);
944 if (iconv (cd, 0, 0, 0, 0) == (size_t) -1)
946 conversion_ok = false;
950 iconv_ret = iconv (cd, &inbuf, &inbytesleft,
951 &outbuf, &outbytesleft);
952 if (iconv_ret == (size_t) -1 || inbytesleft != 0)
963 conversion_ok = false;
967 else if (iconv_ret != 0)
969 conversion_ok = false;
972 /* Return to initial shift state. */
973 if (iconv (cd, 0, 0, &outbuf, &outbytesleft) == (size_t) -1)
984 conversion_ok = false;
998 /* Otherwise, convert non-ASCII characters in IDENT to UCNs. */
1000 char *ret = GGC_NEWVEC (char, 10 * idlen + 1);
1002 for (i = 0; i < idlen;)
1005 size_t utf8_len = decode_utf8_char (&uid[i], idlen - i, &c);
1010 sprintf (p, "\\U%08x", c);