1 /* Copyright (C) 2002, 2003, 2005, 2007, 2008, 2009, 2010
2 Free Software Foundation, Inc.
3 Contributed by Andy Vaught
4 F2003 I/O support contributed by Jerry DeLisle
6 This file is part of the GNU Fortran 95 runtime library (libgfortran).
8 Libgfortran is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3, or (at your option)
13 Libgfortran is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 Under Section 7 of GPL version 3, you are granted additional
19 permissions described in the GCC Runtime Library Exception, version
20 3.1, as published by the Free Software Foundation.
22 You should have received a copy of the GNU General Public License and
23 a copy of the GCC Runtime Library Exception along with this program;
24 see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
25 <http://www.gnu.org/licenses/>. */
37 typedef unsigned char uchar;
39 /* read.c -- Deal with formatted reads */
42 /* set_integer()-- All of the integer assignments come here to
43 * actually place the value into memory. */
46 set_integer (void *dest, GFC_INTEGER_LARGEST value, int length)
50 #ifdef HAVE_GFC_INTEGER_16
51 /* length=10 comes about for kind=10 real/complex BOZ, cf. PR41711. */
55 GFC_INTEGER_16 tmp = value;
56 memcpy (dest, (void *) &tmp, length);
62 GFC_INTEGER_8 tmp = value;
63 memcpy (dest, (void *) &tmp, length);
68 GFC_INTEGER_4 tmp = value;
69 memcpy (dest, (void *) &tmp, length);
74 GFC_INTEGER_2 tmp = value;
75 memcpy (dest, (void *) &tmp, length);
80 GFC_INTEGER_1 tmp = value;
81 memcpy (dest, (void *) &tmp, length);
85 internal_error (NULL, "Bad integer kind");
90 /* max_value()-- Given a length (kind), return the maximum signed or
94 max_value (int length, int signed_flag)
96 GFC_UINTEGER_LARGEST value;
97 #if defined HAVE_GFC_REAL_16 || defined HAVE_GFC_REAL_10
103 #if defined HAVE_GFC_REAL_16 || defined HAVE_GFC_REAL_10
107 for (n = 1; n < 4 * length; n++)
108 value = (value << 2) + 3;
114 value = signed_flag ? 0x7fffffffffffffff : 0xffffffffffffffff;
117 value = signed_flag ? 0x7fffffff : 0xffffffff;
120 value = signed_flag ? 0x7fff : 0xffff;
123 value = signed_flag ? 0x7f : 0xff;
126 internal_error (NULL, "Bad integer kind");
133 /* convert_real()-- Convert a character representation of a floating
134 point number to the machine number. Returns nonzero if there is a
135 range problem during conversion. Note: many architectures
136 (e.g. IA-64, HP-PA) require that the storage pointed to by the dest
137 argument is properly aligned for the type in question. */
140 convert_real (st_parameter_dt *dtp, void *dest, const char *buffer, int length)
147 *((GFC_REAL_4*) dest) =
148 #if defined(HAVE_STRTOF)
149 gfc_strtof (buffer, NULL);
151 (GFC_REAL_4) gfc_strtod (buffer, NULL);
156 *((GFC_REAL_8*) dest) = gfc_strtod (buffer, NULL);
159 #if defined(HAVE_GFC_REAL_10) && defined (HAVE_STRTOLD)
161 *((GFC_REAL_10*) dest) = gfc_strtold (buffer, NULL);
165 #if defined(HAVE_GFC_REAL_16) && defined (HAVE_STRTOLD)
167 *((GFC_REAL_16*) dest) = gfc_strtold (buffer, NULL);
172 internal_error (&dtp->common, "Unsupported real kind during IO");
177 generate_error (&dtp->common, LIBERROR_READ_VALUE,
178 "Error during floating point read");
179 next_record (dtp, 1);
187 /* read_l()-- Read a logical value */
190 read_l (st_parameter_dt *dtp, const fnode *f, char *dest, int length)
197 p = read_block_form (dtp, &w);
220 set_integer (dest, (GFC_INTEGER_LARGEST) 1, length);
224 set_integer (dest, (GFC_INTEGER_LARGEST) 0, length);
228 generate_error (&dtp->common, LIBERROR_READ_VALUE,
229 "Bad value on logical read");
230 next_record (dtp, 1);
237 read_utf8 (st_parameter_dt *dtp, int *nbytes)
239 static const uchar masks[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x02, 0x01 };
240 static const uchar patns[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
247 s = read_block_form (dtp, nbytes);
251 /* If this is a short read, just return. */
259 /* The number of leading 1-bits in the first byte indicates how many
261 for (nb = 2; nb < 7; nb++)
262 if ((c & ~masks[nb-1]) == patns[nb-1])
267 c = (c & masks[nb-1]);
270 s = read_block_form (dtp, &nread);
273 /* Decode the bytes read. */
274 for (i = 1; i < nb; i++)
276 gfc_char4_t n = *s++;
278 if ((n & 0xC0) != 0x80)
281 c = ((c << 6) + (n & 0x3F));
284 /* Make sure the shortest possible encoding was used. */
285 if (c <= 0x7F && nb > 1) goto invalid;
286 if (c <= 0x7FF && nb > 2) goto invalid;
287 if (c <= 0xFFFF && nb > 3) goto invalid;
288 if (c <= 0x1FFFFF && nb > 4) goto invalid;
289 if (c <= 0x3FFFFFF && nb > 5) goto invalid;
291 /* Make sure the character is valid. */
292 if (c > 0x7FFFFFFF || (c >= 0xD800 && c <= 0xDFFF))
298 generate_error (&dtp->common, LIBERROR_READ_VALUE, "Invalid UTF-8 encoding");
299 return (gfc_char4_t) '?';
304 read_utf8_char1 (st_parameter_dt *dtp, char *p, int len, int width)
311 len = (width < len) ? len : width;
315 /* Proceed with decoding one character at a time. */
316 for (j = 0; j < len; j++, dest++)
318 c = read_utf8 (dtp, &nbytes);
320 /* Check for a short read and if so, break out. */
324 *dest = c > 255 ? '?' : (uchar) c;
327 /* If there was a short read, pad the remaining characters. */
328 for (i = j; i < len; i++)
334 read_default_char1 (st_parameter_dt *dtp, char *p, int len, int width)
339 s = read_block_form (dtp, &width);
346 m = (width > len) ? len : width;
351 memset (p + m, ' ', n);
356 read_utf8_char4 (st_parameter_dt *dtp, void *p, int len, int width)
362 len = (width < len) ? len : width;
364 dest = (gfc_char4_t *) p;
366 /* Proceed with decoding one character at a time. */
367 for (j = 0; j < len; j++, dest++)
369 *dest = read_utf8 (dtp, &nbytes);
371 /* Check for a short read and if so, break out. */
376 /* If there was a short read, pad the remaining characters. */
377 for (i = j; i < len; i++)
378 *dest++ = (gfc_char4_t) ' ';
384 read_default_char4 (st_parameter_dt *dtp, char *p, int len, int width)
390 s = read_block_form (dtp, &width);
397 m = ((int) width > len) ? len : (int) width;
399 dest = (gfc_char4_t *) p;
401 for (n = 0; n < m; n++, dest++, s++)
402 *dest = (unsigned char ) *s;
404 for (n = 0; n < len - (int) width; n++, dest++)
405 *dest = (unsigned char) ' ';
409 /* read_a()-- Read a character record into a KIND=1 character destination,
410 processing UTF-8 encoding if necessary. */
413 read_a (st_parameter_dt *dtp, const fnode *f, char *p, int length)
419 if (wi == -1) /* '(A)' edit descriptor */
423 /* Read in w characters, treating comma as not a separator. */
424 dtp->u.p.sf_read_comma = 0;
426 if (dtp->u.p.current_unit->flags.encoding == ENCODING_UTF8)
427 read_utf8_char1 (dtp, p, length, w);
429 read_default_char1 (dtp, p, length, w);
431 dtp->u.p.sf_read_comma =
432 dtp->u.p.current_unit->decimal_status == DECIMAL_COMMA ? 0 : 1;
436 /* read_a_char4()-- Read a character record into a KIND=4 character destination,
437 processing UTF-8 encoding if necessary. */
440 read_a_char4 (st_parameter_dt *dtp, const fnode *f, char *p, int length)
445 if (w == -1) /* '(A)' edit descriptor */
448 /* Read in w characters, treating comma as not a separator. */
449 dtp->u.p.sf_read_comma = 0;
451 if (dtp->u.p.current_unit->flags.encoding == ENCODING_UTF8)
452 read_utf8_char4 (dtp, p, length, w);
454 read_default_char4 (dtp, p, length, w);
456 dtp->u.p.sf_read_comma =
457 dtp->u.p.current_unit->decimal_status == DECIMAL_COMMA ? 0 : 1;
460 /* eat_leading_spaces()-- Given a character pointer and a width,
461 * ignore the leading spaces. */
464 eat_leading_spaces (int *width, char *p)
468 if (*width == 0 || *p != ' ')
480 next_char (st_parameter_dt *dtp, char **p, int *w)
495 if (dtp->u.p.blank_status != BLANK_UNSPECIFIED)
496 return ' '; /* return a blank to signal a null */
498 /* At this point, the rest of the field has to be trailing blanks */
512 /* read_decimal()-- Read a decimal integer value. The values here are
516 read_decimal (st_parameter_dt *dtp, const fnode *f, char *dest, int length)
518 GFC_UINTEGER_LARGEST value, maxv, maxv_10;
519 GFC_INTEGER_LARGEST v;
525 p = read_block_form (dtp, &w);
530 p = eat_leading_spaces (&w, p);
533 set_integer (dest, (GFC_INTEGER_LARGEST) 0, length);
537 maxv = max_value (length, 1);
559 /* At this point we have a digit-string */
564 c = next_char (dtp, &p, &w);
570 if (dtp->u.p.blank_status == BLANK_NULL) continue;
571 if (dtp->u.p.blank_status == BLANK_ZERO) c = '0';
574 if (c < '0' || c > '9')
577 if (value > maxv_10 && compile_options.range_check == 1)
583 if (value > maxv - c && compile_options.range_check == 1)
592 set_integer (dest, v, length);
596 generate_error (&dtp->common, LIBERROR_READ_VALUE,
597 "Bad value during integer read");
598 next_record (dtp, 1);
602 generate_error (&dtp->common, LIBERROR_READ_OVERFLOW,
603 "Value overflowed during integer read");
604 next_record (dtp, 1);
609 /* read_radix()-- This function reads values for non-decimal radixes.
610 * The difference here is that we treat the values here as unsigned
611 * values for the purposes of overflow. If minus sign is present and
612 * the top bit is set, the value will be incorrect. */
615 read_radix (st_parameter_dt *dtp, const fnode *f, char *dest, int length,
618 GFC_UINTEGER_LARGEST value, maxv, maxv_r;
619 GFC_INTEGER_LARGEST v;
625 p = read_block_form (dtp, &w);
630 p = eat_leading_spaces (&w, p);
633 set_integer (dest, (GFC_INTEGER_LARGEST) 0, length);
637 maxv = max_value (length, 0);
638 maxv_r = maxv / radix;
659 /* At this point we have a digit-string */
664 c = next_char (dtp, &p, &w);
669 if (dtp->u.p.blank_status == BLANK_NULL) continue;
670 if (dtp->u.p.blank_status == BLANK_ZERO) c = '0';
676 if (c < '0' || c > '1')
681 if (c < '0' || c > '7')
706 c = c - 'a' + '9' + 1;
715 c = c - 'A' + '9' + 1;
729 value = radix * value;
731 if (maxv - c < value)
740 set_integer (dest, v, length);
744 generate_error (&dtp->common, LIBERROR_READ_VALUE,
745 "Bad value during integer read");
746 next_record (dtp, 1);
750 generate_error (&dtp->common, LIBERROR_READ_OVERFLOW,
751 "Value overflowed during integer read");
752 next_record (dtp, 1);
757 /* read_f()-- Read a floating point number with F-style editing, which
758 is what all of the other floating point descriptors behave as. The
759 tricky part is that optional spaces are allowed after an E or D,
760 and the implicit decimal point if a decimal point is not present in
764 read_f (st_parameter_dt *dtp, const fnode *f, char *dest, int length)
766 int w, seen_dp, exponent;
771 int seen_int_digit; /* Seen a digit before the decimal point? */
772 int seen_dec_digit; /* Seen a digit after the decimal point? */
781 /* Read in the next block. */
782 p = read_block_form (dtp, &w);
785 p = eat_leading_spaces (&w, (char*) p);
789 /* In this buffer we're going to re-format the number cleanly to be parsed
790 by convert_real in the end; this assures we're using strtod from the
791 C library for parsing and thus probably get the best accuracy possible.
792 This process may add a '+0.0' in front of the number as well as change the
793 exponent because of an implicit decimal point or the like. Thus allocating
794 strlen ("+0.0e-1000") == 10 characters plus one for NUL more than the
795 original buffer had should be enough. */
796 buffer = gfc_alloca (w + 11);
800 if (*p == '-' || *p == '+')
808 p = eat_leading_spaces (&w, (char*) p);
812 /* Check for Infinity or NaN. */
813 if (unlikely ((w >= 3 && (*p == 'i' || *p == 'I' || *p == 'n' || *p == 'N'))))
818 /* Scan through the buffer keeping track of spaces and parenthesis. We
819 null terminate the string as soon as we see a left paren or if we are
820 BLANK_NULL mode. Leading spaces have already been skipped above,
821 trailing spaces are ignored by converting to '\0'. A space
822 between "NaN" and the optional perenthesis is not permitted. */
829 if (dtp->u.p.blank_status == BLANK_ZERO)
843 if (seen_paren++ != 1)
857 if (seen_paren != 0 && seen_paren != 2)
860 if ((strcmp (save, "inf") == 0) || (strcmp (save, "infinity") == 0))
865 else if (strcmp (save, "nan") != 0)
868 convert_real (dtp, dest, buffer, length);
872 /* Process the mantissa string. */
878 if (dtp->u.p.current_unit->decimal_status != DECIMAL_COMMA)
891 if (dtp->u.p.blank_status == BLANK_ZERO)
896 else if (dtp->u.p.blank_status == BLANK_NULL)
899 /* TODO: Should we check instead that there are only trailing
900 blanks here, as is done below for exponents? */
941 /* No exponent has been seen, so we use the current scale factor. */
942 exponent = - dtp->u.p.scale_factor;
945 /* At this point the start of an exponent has been found. */
947 p = eat_leading_spaces (&w, (char*) p);
948 if (*p == '-' || *p == '+')
956 /* At this point a digit string is required. We calculate the value
957 of the exponent in order to take account of the scale factor and
958 the d parameter before explict conversion takes place. */
963 if (dtp->u.p.blank_status == BLANK_UNSPECIFIED)
965 while (w > 0 && isdigit (*p))
968 exponent += *p - '0';
973 /* Only allow trailing blanks. */
982 else /* BZ or BN status is enabled. */
988 if (dtp->u.p.blank_status == BLANK_ZERO)
991 assert (dtp->u.p.blank_status == BLANK_NULL);
993 else if (!isdigit (*p))
998 exponent += *p - '0';
1006 exponent *= exponent_sign;
1009 /* Use the precision specified in the format if no decimal point has been
1012 exponent -= f->u.real.d;
1014 /* Output a trailing '0' after decimal point if not yet found. */
1015 if (seen_dp && !seen_dec_digit)
1018 /* Print out the exponent to finish the reformatted number. Maximum 4
1019 digits for the exponent. */
1028 exponent = - exponent;
1031 assert (exponent < 10000);
1032 for (dig = 3; dig >= 0; --dig)
1034 out[dig] = (char) ('0' + exponent % 10);
1041 /* Do the actual conversion. */
1042 convert_real (dtp, dest, buffer, length);
1046 /* The value read is zero. */
1051 *((GFC_REAL_4 *) dest) = 0.0;
1055 *((GFC_REAL_8 *) dest) = 0.0;
1058 #ifdef HAVE_GFC_REAL_10
1060 *((GFC_REAL_10 *) dest) = 0.0;
1064 #ifdef HAVE_GFC_REAL_16
1066 *((GFC_REAL_16 *) dest) = 0.0;
1071 internal_error (&dtp->common, "Unsupported real kind during IO");
1076 generate_error (&dtp->common, LIBERROR_READ_VALUE,
1077 "Bad value during floating point read");
1078 next_record (dtp, 1);
1083 /* read_x()-- Deal with the X/TR descriptor. We just read some data
1084 * and never look at it. */
1087 read_x (st_parameter_dt *dtp, int n)
1092 if ((dtp->u.p.current_unit->pad_status == PAD_NO || is_internal_unit (dtp))
1093 && dtp->u.p.current_unit->bytes_left < n)
1094 n = dtp->u.p.current_unit->bytes_left;
1101 if (is_internal_unit (dtp))
1103 p = mem_alloc_r (dtp->u.p.current_unit->s, &length);
1104 if (unlikely (length < n))
1109 if (dtp->u.p.sf_seen_eor)
1112 p = fbuf_read (dtp->u.p.current_unit, &length);
1119 if (length == 0 && dtp->u.p.item_count == 1)
1121 if (dtp->u.p.current_unit->pad_status == PAD_NO)
1134 if (q == '\n' || q == '\r')
1136 /* Unexpected end of line. Set the position. */
1137 fbuf_seek (dtp->u.p.current_unit, n + 1 ,SEEK_CUR);
1138 dtp->u.p.sf_seen_eor = 1;
1140 /* If we encounter a CR, it might be a CRLF. */
1141 if (q == '\r') /* Probably a CRLF */
1143 /* See if there is an LF. Use fbuf_read rather then fbuf_getc so
1144 the position is not advanced unless it really is an LF. */
1146 p = fbuf_read (dtp->u.p.current_unit, &readlen);
1147 if (*p == '\n' && readlen == 1)
1149 dtp->u.p.sf_seen_eor = 2;
1150 fbuf_seek (dtp->u.p.current_unit, 1 ,SEEK_CUR);
1159 fbuf_seek (dtp->u.p.current_unit, n, SEEK_CUR);
1162 if ((dtp->common.flags & IOPARM_DT_HAS_SIZE) != 0)
1163 dtp->u.p.size_used += (GFC_IO_INT) n;
1164 dtp->u.p.current_unit->bytes_left -= n;
1165 dtp->u.p.current_unit->strm_pos += (gfc_offset) n;