| 1 | /* Support for printing Fortran values for GDB, the GNU debugger. |
| 2 | |
| 3 | Copyright (C) 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2003, 2005, 2006, |
| 4 | 2007, 2008, 2009, 2010 Free Software Foundation, Inc. |
| 5 | |
| 6 | Contributed by Motorola. Adapted from the C definitions by Farooq Butt |
| 7 | (fmbutt@engage.sps.mot.com), additionally worked over by Stan Shebs. |
| 8 | |
| 9 | This file is part of GDB. |
| 10 | |
| 11 | This program is free software; you can redistribute it and/or modify |
| 12 | it under the terms of the GNU General Public License as published by |
| 13 | the Free Software Foundation; either version 3 of the License, or |
| 14 | (at your option) any later version. |
| 15 | |
| 16 | This program is distributed in the hope that it will be useful, |
| 17 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 18 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 19 | GNU General Public License for more details. |
| 20 | |
| 21 | You should have received a copy of the GNU General Public License |
| 22 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ |
| 23 | |
| 24 | #include "defs.h" |
| 25 | #include "gdb_string.h" |
| 26 | #include "symtab.h" |
| 27 | #include "gdbtypes.h" |
| 28 | #include "expression.h" |
| 29 | #include "value.h" |
| 30 | #include "valprint.h" |
| 31 | #include "language.h" |
| 32 | #include "f-lang.h" |
| 33 | #include "frame.h" |
| 34 | #include "gdbcore.h" |
| 35 | #include "command.h" |
| 36 | #include "block.h" |
| 37 | |
| 38 | #if 0 |
| 39 | static int there_is_a_visible_common_named (char *); |
| 40 | #endif |
| 41 | |
| 42 | extern void _initialize_f_valprint (void); |
| 43 | static void info_common_command (char *, int); |
| 44 | static void list_all_visible_commons (char *); |
| 45 | static void f77_create_arrayprint_offset_tbl (struct type *, |
| 46 | struct ui_file *); |
| 47 | static void f77_get_dynamic_length_of_aggregate (struct type *); |
| 48 | |
| 49 | int f77_array_offset_tbl[MAX_FORTRAN_DIMS + 1][2]; |
| 50 | |
| 51 | /* Array which holds offsets to be applied to get a row's elements |
| 52 | for a given array. Array also holds the size of each subarray. */ |
| 53 | |
| 54 | /* The following macro gives us the size of the nth dimension, Where |
| 55 | n is 1 based. */ |
| 56 | |
| 57 | #define F77_DIM_SIZE(n) (f77_array_offset_tbl[n][1]) |
| 58 | |
| 59 | /* The following gives us the offset for row n where n is 1-based. */ |
| 60 | |
| 61 | #define F77_DIM_OFFSET(n) (f77_array_offset_tbl[n][0]) |
| 62 | |
| 63 | int |
| 64 | f77_get_lowerbound (struct type *type) |
| 65 | { |
| 66 | if (TYPE_ARRAY_LOWER_BOUND_IS_UNDEFINED (type)) |
| 67 | error (_("Lower bound may not be '*' in F77")); |
| 68 | |
| 69 | return TYPE_ARRAY_LOWER_BOUND_VALUE (type); |
| 70 | } |
| 71 | |
| 72 | int |
| 73 | f77_get_upperbound (struct type *type) |
| 74 | { |
| 75 | if (TYPE_ARRAY_UPPER_BOUND_IS_UNDEFINED (type)) |
| 76 | { |
| 77 | /* We have an assumed size array on our hands. Assume that |
| 78 | upper_bound == lower_bound so that we show at least 1 element. |
| 79 | If the user wants to see more elements, let him manually ask for 'em |
| 80 | and we'll subscript the array and show him. */ |
| 81 | |
| 82 | return f77_get_lowerbound (type); |
| 83 | } |
| 84 | |
| 85 | return TYPE_ARRAY_UPPER_BOUND_VALUE (type); |
| 86 | } |
| 87 | |
| 88 | /* Obtain F77 adjustable array dimensions */ |
| 89 | |
| 90 | static void |
| 91 | f77_get_dynamic_length_of_aggregate (struct type *type) |
| 92 | { |
| 93 | int upper_bound = -1; |
| 94 | int lower_bound = 1; |
| 95 | int retcode; |
| 96 | |
| 97 | /* Recursively go all the way down into a possibly multi-dimensional |
| 98 | F77 array and get the bounds. For simple arrays, this is pretty |
| 99 | easy but when the bounds are dynamic, we must be very careful |
| 100 | to add up all the lengths correctly. Not doing this right |
| 101 | will lead to horrendous-looking arrays in parameter lists. |
| 102 | |
| 103 | This function also works for strings which behave very |
| 104 | similarly to arrays. */ |
| 105 | |
| 106 | if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_ARRAY |
| 107 | || TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_STRING) |
| 108 | f77_get_dynamic_length_of_aggregate (TYPE_TARGET_TYPE (type)); |
| 109 | |
| 110 | /* Recursion ends here, start setting up lengths. */ |
| 111 | lower_bound = f77_get_lowerbound (type); |
| 112 | upper_bound = f77_get_upperbound (type); |
| 113 | |
| 114 | /* Patch in a valid length value. */ |
| 115 | |
| 116 | TYPE_LENGTH (type) = |
| 117 | (upper_bound - lower_bound + 1) * TYPE_LENGTH (check_typedef (TYPE_TARGET_TYPE (type))); |
| 118 | } |
| 119 | |
| 120 | /* Function that sets up the array offset,size table for the array |
| 121 | type "type". */ |
| 122 | |
| 123 | static void |
| 124 | f77_create_arrayprint_offset_tbl (struct type *type, struct ui_file *stream) |
| 125 | { |
| 126 | struct type *tmp_type; |
| 127 | int eltlen; |
| 128 | int ndimen = 1; |
| 129 | int upper, lower, retcode; |
| 130 | |
| 131 | tmp_type = type; |
| 132 | |
| 133 | while ((TYPE_CODE (tmp_type) == TYPE_CODE_ARRAY)) |
| 134 | { |
| 135 | upper = f77_get_upperbound (tmp_type); |
| 136 | lower = f77_get_lowerbound (tmp_type); |
| 137 | |
| 138 | F77_DIM_SIZE (ndimen) = upper - lower + 1; |
| 139 | |
| 140 | tmp_type = TYPE_TARGET_TYPE (tmp_type); |
| 141 | ndimen++; |
| 142 | } |
| 143 | |
| 144 | /* Now we multiply eltlen by all the offsets, so that later we |
| 145 | can print out array elements correctly. Up till now we |
| 146 | know an offset to apply to get the item but we also |
| 147 | have to know how much to add to get to the next item */ |
| 148 | |
| 149 | ndimen--; |
| 150 | eltlen = TYPE_LENGTH (tmp_type); |
| 151 | F77_DIM_OFFSET (ndimen) = eltlen; |
| 152 | while (--ndimen > 0) |
| 153 | { |
| 154 | eltlen *= F77_DIM_SIZE (ndimen + 1); |
| 155 | F77_DIM_OFFSET (ndimen) = eltlen; |
| 156 | } |
| 157 | } |
| 158 | |
| 159 | |
| 160 | |
| 161 | /* Actual function which prints out F77 arrays, Valaddr == address in |
| 162 | the superior. Address == the address in the inferior. */ |
| 163 | |
| 164 | static void |
| 165 | f77_print_array_1 (int nss, int ndimensions, struct type *type, |
| 166 | const gdb_byte *valaddr, CORE_ADDR address, |
| 167 | struct ui_file *stream, int recurse, |
| 168 | const struct value_print_options *options, |
| 169 | int *elts) |
| 170 | { |
| 171 | int i; |
| 172 | |
| 173 | if (nss != ndimensions) |
| 174 | { |
| 175 | for (i = 0; (i < F77_DIM_SIZE (nss) && (*elts) < options->print_max); i++) |
| 176 | { |
| 177 | fprintf_filtered (stream, "( "); |
| 178 | f77_print_array_1 (nss + 1, ndimensions, TYPE_TARGET_TYPE (type), |
| 179 | valaddr + i * F77_DIM_OFFSET (nss), |
| 180 | address + i * F77_DIM_OFFSET (nss), |
| 181 | stream, recurse, options, elts); |
| 182 | fprintf_filtered (stream, ") "); |
| 183 | } |
| 184 | if (*elts >= options->print_max && i < F77_DIM_SIZE (nss)) |
| 185 | fprintf_filtered (stream, "..."); |
| 186 | } |
| 187 | else |
| 188 | { |
| 189 | for (i = 0; i < F77_DIM_SIZE (nss) && (*elts) < options->print_max; |
| 190 | i++, (*elts)++) |
| 191 | { |
| 192 | val_print (TYPE_TARGET_TYPE (type), |
| 193 | valaddr + i * F77_DIM_OFFSET (ndimensions), |
| 194 | 0, |
| 195 | address + i * F77_DIM_OFFSET (ndimensions), |
| 196 | stream, recurse, options, current_language); |
| 197 | |
| 198 | if (i != (F77_DIM_SIZE (nss) - 1)) |
| 199 | fprintf_filtered (stream, ", "); |
| 200 | |
| 201 | if ((*elts == options->print_max - 1) |
| 202 | && (i != (F77_DIM_SIZE (nss) - 1))) |
| 203 | fprintf_filtered (stream, "..."); |
| 204 | } |
| 205 | } |
| 206 | } |
| 207 | |
| 208 | /* This function gets called to print an F77 array, we set up some |
| 209 | stuff and then immediately call f77_print_array_1() */ |
| 210 | |
| 211 | static void |
| 212 | f77_print_array (struct type *type, const gdb_byte *valaddr, |
| 213 | CORE_ADDR address, struct ui_file *stream, |
| 214 | int recurse, const struct value_print_options *options) |
| 215 | { |
| 216 | int ndimensions; |
| 217 | int elts = 0; |
| 218 | |
| 219 | ndimensions = calc_f77_array_dims (type); |
| 220 | |
| 221 | if (ndimensions > MAX_FORTRAN_DIMS || ndimensions < 0) |
| 222 | error (_("Type node corrupt! F77 arrays cannot have %d subscripts (%d Max)"), |
| 223 | ndimensions, MAX_FORTRAN_DIMS); |
| 224 | |
| 225 | /* Since F77 arrays are stored column-major, we set up an |
| 226 | offset table to get at the various row's elements. The |
| 227 | offset table contains entries for both offset and subarray size. */ |
| 228 | |
| 229 | f77_create_arrayprint_offset_tbl (type, stream); |
| 230 | |
| 231 | f77_print_array_1 (1, ndimensions, type, valaddr, address, stream, |
| 232 | recurse, options, &elts); |
| 233 | } |
| 234 | \f |
| 235 | |
| 236 | /* Print data of type TYPE located at VALADDR (within GDB), which came from |
| 237 | the inferior at address ADDRESS, onto stdio stream STREAM according to |
| 238 | OPTIONS. The data at VALADDR is in target byte order. |
| 239 | |
| 240 | If the data are a string pointer, returns the number of string characters |
| 241 | printed. */ |
| 242 | |
| 243 | int |
| 244 | f_val_print (struct type *type, const gdb_byte *valaddr, int embedded_offset, |
| 245 | CORE_ADDR address, struct ui_file *stream, int recurse, |
| 246 | const struct value_print_options *options) |
| 247 | { |
| 248 | struct gdbarch *gdbarch = get_type_arch (type); |
| 249 | enum bfd_endian byte_order = gdbarch_byte_order (gdbarch); |
| 250 | unsigned int i = 0; /* Number of characters printed */ |
| 251 | struct type *elttype; |
| 252 | LONGEST val; |
| 253 | CORE_ADDR addr; |
| 254 | int index; |
| 255 | |
| 256 | CHECK_TYPEDEF (type); |
| 257 | switch (TYPE_CODE (type)) |
| 258 | { |
| 259 | case TYPE_CODE_STRING: |
| 260 | f77_get_dynamic_length_of_aggregate (type); |
| 261 | LA_PRINT_STRING (stream, builtin_type (gdbarch)->builtin_char, |
| 262 | valaddr, TYPE_LENGTH (type), NULL, 0, options); |
| 263 | break; |
| 264 | |
| 265 | case TYPE_CODE_ARRAY: |
| 266 | fprintf_filtered (stream, "("); |
| 267 | f77_print_array (type, valaddr, address, stream, recurse, options); |
| 268 | fprintf_filtered (stream, ")"); |
| 269 | break; |
| 270 | |
| 271 | case TYPE_CODE_PTR: |
| 272 | if (options->format && options->format != 's') |
| 273 | { |
| 274 | print_scalar_formatted (valaddr, type, options, 0, stream); |
| 275 | break; |
| 276 | } |
| 277 | else |
| 278 | { |
| 279 | addr = unpack_pointer (type, valaddr); |
| 280 | elttype = check_typedef (TYPE_TARGET_TYPE (type)); |
| 281 | |
| 282 | if (TYPE_CODE (elttype) == TYPE_CODE_FUNC) |
| 283 | { |
| 284 | /* Try to print what function it points to. */ |
| 285 | print_address_demangle (gdbarch, addr, stream, demangle); |
| 286 | /* Return value is irrelevant except for string pointers. */ |
| 287 | return 0; |
| 288 | } |
| 289 | |
| 290 | if (options->addressprint && options->format != 's') |
| 291 | fputs_filtered (paddress (gdbarch, addr), stream); |
| 292 | |
| 293 | /* For a pointer to char or unsigned char, also print the string |
| 294 | pointed to, unless pointer is null. */ |
| 295 | if (TYPE_LENGTH (elttype) == 1 |
| 296 | && TYPE_CODE (elttype) == TYPE_CODE_INT |
| 297 | && (options->format == 0 || options->format == 's') |
| 298 | && addr != 0) |
| 299 | i = val_print_string (TYPE_TARGET_TYPE (type), addr, -1, stream, |
| 300 | options); |
| 301 | |
| 302 | /* Return number of characters printed, including the terminating |
| 303 | '\0' if we reached the end. val_print_string takes care including |
| 304 | the terminating '\0' if necessary. */ |
| 305 | return i; |
| 306 | } |
| 307 | break; |
| 308 | |
| 309 | case TYPE_CODE_REF: |
| 310 | elttype = check_typedef (TYPE_TARGET_TYPE (type)); |
| 311 | if (options->addressprint) |
| 312 | { |
| 313 | CORE_ADDR addr |
| 314 | = extract_typed_address (valaddr + embedded_offset, type); |
| 315 | fprintf_filtered (stream, "@"); |
| 316 | fputs_filtered (paddress (gdbarch, addr), stream); |
| 317 | if (options->deref_ref) |
| 318 | fputs_filtered (": ", stream); |
| 319 | } |
| 320 | /* De-reference the reference. */ |
| 321 | if (options->deref_ref) |
| 322 | { |
| 323 | if (TYPE_CODE (elttype) != TYPE_CODE_UNDEF) |
| 324 | { |
| 325 | struct value *deref_val = |
| 326 | value_at |
| 327 | (TYPE_TARGET_TYPE (type), |
| 328 | unpack_pointer (type, valaddr + embedded_offset)); |
| 329 | common_val_print (deref_val, stream, recurse, |
| 330 | options, current_language); |
| 331 | } |
| 332 | else |
| 333 | fputs_filtered ("???", stream); |
| 334 | } |
| 335 | break; |
| 336 | |
| 337 | case TYPE_CODE_FUNC: |
| 338 | if (options->format) |
| 339 | { |
| 340 | print_scalar_formatted (valaddr, type, options, 0, stream); |
| 341 | break; |
| 342 | } |
| 343 | /* FIXME, we should consider, at least for ANSI C language, eliminating |
| 344 | the distinction made between FUNCs and POINTERs to FUNCs. */ |
| 345 | fprintf_filtered (stream, "{"); |
| 346 | type_print (type, "", stream, -1); |
| 347 | fprintf_filtered (stream, "} "); |
| 348 | /* Try to print what function it points to, and its address. */ |
| 349 | print_address_demangle (gdbarch, address, stream, demangle); |
| 350 | break; |
| 351 | |
| 352 | case TYPE_CODE_INT: |
| 353 | if (options->format || options->output_format) |
| 354 | { |
| 355 | struct value_print_options opts = *options; |
| 356 | opts.format = (options->format ? options->format |
| 357 | : options->output_format); |
| 358 | print_scalar_formatted (valaddr, type, &opts, 0, stream); |
| 359 | } |
| 360 | else |
| 361 | { |
| 362 | val_print_type_code_int (type, valaddr, stream); |
| 363 | /* C and C++ has no single byte int type, char is used instead. |
| 364 | Since we don't know whether the value is really intended to |
| 365 | be used as an integer or a character, print the character |
| 366 | equivalent as well. */ |
| 367 | if (TYPE_LENGTH (type) == 1) |
| 368 | { |
| 369 | fputs_filtered (" ", stream); |
| 370 | LA_PRINT_CHAR ((unsigned char) unpack_long (type, valaddr), |
| 371 | type, stream); |
| 372 | } |
| 373 | } |
| 374 | break; |
| 375 | |
| 376 | case TYPE_CODE_FLAGS: |
| 377 | if (options->format) |
| 378 | print_scalar_formatted (valaddr, type, options, 0, stream); |
| 379 | else |
| 380 | val_print_type_code_flags (type, valaddr, stream); |
| 381 | break; |
| 382 | |
| 383 | case TYPE_CODE_FLT: |
| 384 | if (options->format) |
| 385 | print_scalar_formatted (valaddr, type, options, 0, stream); |
| 386 | else |
| 387 | print_floating (valaddr, type, stream); |
| 388 | break; |
| 389 | |
| 390 | case TYPE_CODE_VOID: |
| 391 | fprintf_filtered (stream, "VOID"); |
| 392 | break; |
| 393 | |
| 394 | case TYPE_CODE_ERROR: |
| 395 | fprintf_filtered (stream, "<error type>"); |
| 396 | break; |
| 397 | |
| 398 | case TYPE_CODE_RANGE: |
| 399 | /* FIXME, we should not ever have to print one of these yet. */ |
| 400 | fprintf_filtered (stream, "<range type>"); |
| 401 | break; |
| 402 | |
| 403 | case TYPE_CODE_BOOL: |
| 404 | if (options->format || options->output_format) |
| 405 | { |
| 406 | struct value_print_options opts = *options; |
| 407 | opts.format = (options->format ? options->format |
| 408 | : options->output_format); |
| 409 | print_scalar_formatted (valaddr, type, &opts, 0, stream); |
| 410 | } |
| 411 | else |
| 412 | { |
| 413 | val = extract_unsigned_integer (valaddr, |
| 414 | TYPE_LENGTH (type), byte_order); |
| 415 | if (val == 0) |
| 416 | fprintf_filtered (stream, ".FALSE."); |
| 417 | else if (val == 1) |
| 418 | fprintf_filtered (stream, ".TRUE."); |
| 419 | else |
| 420 | /* Not a legitimate logical type, print as an integer. */ |
| 421 | { |
| 422 | /* Bash the type code temporarily. */ |
| 423 | TYPE_CODE (type) = TYPE_CODE_INT; |
| 424 | f_val_print (type, valaddr, 0, address, stream, recurse, options); |
| 425 | /* Restore the type code so later uses work as intended. */ |
| 426 | TYPE_CODE (type) = TYPE_CODE_BOOL; |
| 427 | } |
| 428 | } |
| 429 | break; |
| 430 | |
| 431 | case TYPE_CODE_COMPLEX: |
| 432 | type = TYPE_TARGET_TYPE (type); |
| 433 | fputs_filtered ("(", stream); |
| 434 | print_floating (valaddr, type, stream); |
| 435 | fputs_filtered (",", stream); |
| 436 | print_floating (valaddr + TYPE_LENGTH (type), type, stream); |
| 437 | fputs_filtered (")", stream); |
| 438 | break; |
| 439 | |
| 440 | case TYPE_CODE_UNDEF: |
| 441 | /* This happens (without TYPE_FLAG_STUB set) on systems which don't use |
| 442 | dbx xrefs (NO_DBX_XREFS in gcc) if a file has a "struct foo *bar" |
| 443 | and no complete type for struct foo in that file. */ |
| 444 | fprintf_filtered (stream, "<incomplete type>"); |
| 445 | break; |
| 446 | |
| 447 | case TYPE_CODE_STRUCT: |
| 448 | case TYPE_CODE_UNION: |
| 449 | /* Starting from the Fortran 90 standard, Fortran supports derived |
| 450 | types. */ |
| 451 | fprintf_filtered (stream, "( "); |
| 452 | for (index = 0; index < TYPE_NFIELDS (type); index++) |
| 453 | { |
| 454 | int offset = TYPE_FIELD_BITPOS (type, index) / 8; |
| 455 | f_val_print (TYPE_FIELD_TYPE (type, index), valaddr + offset, |
| 456 | embedded_offset, address, stream, recurse, options); |
| 457 | if (index != TYPE_NFIELDS (type) - 1) |
| 458 | fputs_filtered (", ", stream); |
| 459 | } |
| 460 | fprintf_filtered (stream, " )"); |
| 461 | break; |
| 462 | |
| 463 | default: |
| 464 | error (_("Invalid F77 type code %d in symbol table."), TYPE_CODE (type)); |
| 465 | } |
| 466 | gdb_flush (stream); |
| 467 | return 0; |
| 468 | } |
| 469 | |
| 470 | static void |
| 471 | list_all_visible_commons (char *funname) |
| 472 | { |
| 473 | SAVED_F77_COMMON_PTR tmp; |
| 474 | |
| 475 | tmp = head_common_list; |
| 476 | |
| 477 | printf_filtered (_("All COMMON blocks visible at this level:\n\n")); |
| 478 | |
| 479 | while (tmp != NULL) |
| 480 | { |
| 481 | if (strcmp (tmp->owning_function, funname) == 0) |
| 482 | printf_filtered ("%s\n", tmp->name); |
| 483 | |
| 484 | tmp = tmp->next; |
| 485 | } |
| 486 | } |
| 487 | |
| 488 | /* This function is used to print out the values in a given COMMON |
| 489 | block. It will always use the most local common block of the |
| 490 | given name */ |
| 491 | |
| 492 | static void |
| 493 | info_common_command (char *comname, int from_tty) |
| 494 | { |
| 495 | SAVED_F77_COMMON_PTR the_common; |
| 496 | COMMON_ENTRY_PTR entry; |
| 497 | struct frame_info *fi; |
| 498 | char *funname = 0; |
| 499 | struct symbol *func; |
| 500 | |
| 501 | /* We have been told to display the contents of F77 COMMON |
| 502 | block supposedly visible in this function. Let us |
| 503 | first make sure that it is visible and if so, let |
| 504 | us display its contents */ |
| 505 | |
| 506 | fi = get_selected_frame (_("No frame selected")); |
| 507 | |
| 508 | /* The following is generally ripped off from stack.c's routine |
| 509 | print_frame_info() */ |
| 510 | |
| 511 | func = find_pc_function (get_frame_pc (fi)); |
| 512 | if (func) |
| 513 | { |
| 514 | /* In certain pathological cases, the symtabs give the wrong |
| 515 | function (when we are in the first function in a file which |
| 516 | is compiled without debugging symbols, the previous function |
| 517 | is compiled with debugging symbols, and the "foo.o" symbol |
| 518 | that is supposed to tell us where the file with debugging symbols |
| 519 | ends has been truncated by ar because it is longer than 15 |
| 520 | characters). |
| 521 | |
| 522 | So look in the minimal symbol tables as well, and if it comes |
| 523 | up with a larger address for the function use that instead. |
| 524 | I don't think this can ever cause any problems; there shouldn't |
| 525 | be any minimal symbols in the middle of a function. |
| 526 | FIXME: (Not necessarily true. What about text labels) */ |
| 527 | |
| 528 | struct minimal_symbol *msymbol = |
| 529 | lookup_minimal_symbol_by_pc (get_frame_pc (fi)); |
| 530 | |
| 531 | if (msymbol != NULL |
| 532 | && (SYMBOL_VALUE_ADDRESS (msymbol) |
| 533 | > BLOCK_START (SYMBOL_BLOCK_VALUE (func)))) |
| 534 | funname = SYMBOL_LINKAGE_NAME (msymbol); |
| 535 | else |
| 536 | funname = SYMBOL_LINKAGE_NAME (func); |
| 537 | } |
| 538 | else |
| 539 | { |
| 540 | struct minimal_symbol *msymbol = |
| 541 | lookup_minimal_symbol_by_pc (get_frame_pc (fi)); |
| 542 | |
| 543 | if (msymbol != NULL) |
| 544 | funname = SYMBOL_LINKAGE_NAME (msymbol); |
| 545 | else /* Got no 'funname', code below will fail. */ |
| 546 | error (_("No function found for frame.")); |
| 547 | } |
| 548 | |
| 549 | /* If comname is NULL, we assume the user wishes to see the |
| 550 | which COMMON blocks are visible here and then return */ |
| 551 | |
| 552 | if (comname == 0) |
| 553 | { |
| 554 | list_all_visible_commons (funname); |
| 555 | return; |
| 556 | } |
| 557 | |
| 558 | the_common = find_common_for_function (comname, funname); |
| 559 | |
| 560 | if (the_common) |
| 561 | { |
| 562 | if (strcmp (comname, BLANK_COMMON_NAME_LOCAL) == 0) |
| 563 | printf_filtered (_("Contents of blank COMMON block:\n")); |
| 564 | else |
| 565 | printf_filtered (_("Contents of F77 COMMON block '%s':\n"), comname); |
| 566 | |
| 567 | printf_filtered ("\n"); |
| 568 | entry = the_common->entries; |
| 569 | |
| 570 | while (entry != NULL) |
| 571 | { |
| 572 | print_variable_and_value (NULL, entry->symbol, fi, gdb_stdout, 0); |
| 573 | entry = entry->next; |
| 574 | } |
| 575 | } |
| 576 | else |
| 577 | printf_filtered (_("Cannot locate the common block %s in function '%s'\n"), |
| 578 | comname, funname); |
| 579 | } |
| 580 | |
| 581 | /* This function is used to determine whether there is a |
| 582 | F77 common block visible at the current scope called 'comname'. */ |
| 583 | |
| 584 | #if 0 |
| 585 | static int |
| 586 | there_is_a_visible_common_named (char *comname) |
| 587 | { |
| 588 | SAVED_F77_COMMON_PTR the_common; |
| 589 | struct frame_info *fi; |
| 590 | char *funname = 0; |
| 591 | struct symbol *func; |
| 592 | |
| 593 | if (comname == NULL) |
| 594 | error (_("Cannot deal with NULL common name!")); |
| 595 | |
| 596 | fi = get_selected_frame (_("No frame selected")); |
| 597 | |
| 598 | /* The following is generally ripped off from stack.c's routine |
| 599 | print_frame_info() */ |
| 600 | |
| 601 | func = find_pc_function (fi->pc); |
| 602 | if (func) |
| 603 | { |
| 604 | /* In certain pathological cases, the symtabs give the wrong |
| 605 | function (when we are in the first function in a file which |
| 606 | is compiled without debugging symbols, the previous function |
| 607 | is compiled with debugging symbols, and the "foo.o" symbol |
| 608 | that is supposed to tell us where the file with debugging symbols |
| 609 | ends has been truncated by ar because it is longer than 15 |
| 610 | characters). |
| 611 | |
| 612 | So look in the minimal symbol tables as well, and if it comes |
| 613 | up with a larger address for the function use that instead. |
| 614 | I don't think this can ever cause any problems; there shouldn't |
| 615 | be any minimal symbols in the middle of a function. |
| 616 | FIXME: (Not necessarily true. What about text labels) */ |
| 617 | |
| 618 | struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (fi->pc); |
| 619 | |
| 620 | if (msymbol != NULL |
| 621 | && (SYMBOL_VALUE_ADDRESS (msymbol) |
| 622 | > BLOCK_START (SYMBOL_BLOCK_VALUE (func)))) |
| 623 | funname = SYMBOL_LINKAGE_NAME (msymbol); |
| 624 | else |
| 625 | funname = SYMBOL_LINKAGE_NAME (func); |
| 626 | } |
| 627 | else |
| 628 | { |
| 629 | struct minimal_symbol *msymbol = |
| 630 | lookup_minimal_symbol_by_pc (fi->pc); |
| 631 | |
| 632 | if (msymbol != NULL) |
| 633 | funname = SYMBOL_LINKAGE_NAME (msymbol); |
| 634 | } |
| 635 | |
| 636 | the_common = find_common_for_function (comname, funname); |
| 637 | |
| 638 | return (the_common ? 1 : 0); |
| 639 | } |
| 640 | #endif |
| 641 | |
| 642 | void |
| 643 | _initialize_f_valprint (void) |
| 644 | { |
| 645 | add_info ("common", info_common_command, |
| 646 | _("Print out the values contained in a Fortran COMMON block.")); |
| 647 | if (xdb_commands) |
| 648 | add_com ("lc", class_info, info_common_command, |
| 649 | _("Print out the values contained in a Fortran COMMON block.")); |
| 650 | } |