]>
Commit | Line | Data |
---|---|---|
1 | /* Fortran language support routines for GDB, the GNU debugger. | |
2 | ||
3 | Copyright (C) 1993-2025 Free Software Foundation, Inc. | |
4 | ||
5 | Contributed by Motorola. Adapted from the C parser by Farooq Butt | |
6 | (fmbutt@engage.sps.mot.com). | |
7 | ||
8 | This file is part of GDB. | |
9 | ||
10 | This program is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 3 of the License, or | |
13 | (at your option) any later version. | |
14 | ||
15 | This program is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ | |
22 | ||
23 | #include "symtab.h" | |
24 | #include "gdbtypes.h" | |
25 | #include "expression.h" | |
26 | #include "parser-defs.h" | |
27 | #include "language.h" | |
28 | #include "varobj.h" | |
29 | #include "gdbcore.h" | |
30 | #include "f-lang.h" | |
31 | #include "valprint.h" | |
32 | #include "value.h" | |
33 | #include "cp-support.h" | |
34 | #include "charset.h" | |
35 | #include "c-lang.h" | |
36 | #include "target-float.h" | |
37 | #include "gdbarch.h" | |
38 | #include "cli/cli-cmds.h" | |
39 | #include "f-array-walker.h" | |
40 | #include "f-exp.h" | |
41 | ||
42 | #include <math.h> | |
43 | ||
44 | /* Whether GDB should repack array slices created by the user. */ | |
45 | static bool repack_array_slices = false; | |
46 | ||
47 | /* Implement 'show fortran repack-array-slices'. */ | |
48 | static void | |
49 | show_repack_array_slices (struct ui_file *file, int from_tty, | |
50 | struct cmd_list_element *c, const char *value) | |
51 | { | |
52 | gdb_printf (file, _("Repacking of Fortran array slices is %s.\n"), | |
53 | value); | |
54 | } | |
55 | ||
56 | /* Debugging of Fortran's array slicing. */ | |
57 | static bool fortran_array_slicing_debug = false; | |
58 | ||
59 | /* Implement 'show debug fortran-array-slicing'. */ | |
60 | static void | |
61 | show_fortran_array_slicing_debug (struct ui_file *file, int from_tty, | |
62 | struct cmd_list_element *c, | |
63 | const char *value) | |
64 | { | |
65 | gdb_printf (file, _("Debugging of Fortran array slicing is %s.\n"), | |
66 | value); | |
67 | } | |
68 | ||
69 | /* Local functions */ | |
70 | ||
71 | static value *fortran_prepare_argument (struct expression *exp, | |
72 | expr::operation *subexp, | |
73 | int arg_num, bool is_internal_call_p, | |
74 | struct type *func_type, enum noside noside); | |
75 | ||
76 | /* Return the encoding that should be used for the character type | |
77 | TYPE. */ | |
78 | ||
79 | const char * | |
80 | f_language::get_encoding (struct type *type) | |
81 | { | |
82 | const char *encoding; | |
83 | ||
84 | switch (type->length ()) | |
85 | { | |
86 | case 1: | |
87 | encoding = target_charset (type->arch ()); | |
88 | break; | |
89 | case 4: | |
90 | if (type_byte_order (type) == BFD_ENDIAN_BIG) | |
91 | encoding = "UTF-32BE"; | |
92 | else | |
93 | encoding = "UTF-32LE"; | |
94 | break; | |
95 | ||
96 | default: | |
97 | error (_("unrecognized character type")); | |
98 | } | |
99 | ||
100 | return encoding; | |
101 | } | |
102 | ||
103 | /* See language.h. */ | |
104 | ||
105 | struct value * | |
106 | f_language::value_string (struct gdbarch *gdbarch, | |
107 | const char *ptr, ssize_t len) const | |
108 | { | |
109 | struct type *type = language_string_char_type (this, gdbarch); | |
110 | return ::value_string (ptr, len, type); | |
111 | } | |
112 | ||
113 | /* A helper function for the "bound" intrinsics that checks that TYPE | |
114 | is an array. LBOUND_P is true for lower bound; this is used for | |
115 | the error message, if any. */ | |
116 | ||
117 | static void | |
118 | fortran_require_array (struct type *type, bool lbound_p) | |
119 | { | |
120 | type = check_typedef (type); | |
121 | if (type->code () != TYPE_CODE_ARRAY) | |
122 | { | |
123 | if (lbound_p) | |
124 | error (_("LBOUND can only be applied to arrays")); | |
125 | else | |
126 | error (_("UBOUND can only be applied to arrays")); | |
127 | } | |
128 | } | |
129 | ||
130 | /* Create an array containing the lower bounds (when LBOUND_P is true) or | |
131 | the upper bounds (when LBOUND_P is false) of ARRAY (which must be of | |
132 | array type). GDBARCH is the current architecture. */ | |
133 | ||
134 | static struct value * | |
135 | fortran_bounds_all_dims (bool lbound_p, | |
136 | struct gdbarch *gdbarch, | |
137 | struct value *array) | |
138 | { | |
139 | type *array_type = check_typedef (array->type ()); | |
140 | int ndimensions = calc_f77_array_dims (array_type); | |
141 | ||
142 | /* Allocate a result value of the correct type. */ | |
143 | type_allocator alloc (gdbarch); | |
144 | struct type *range | |
145 | = create_static_range_type (alloc, | |
146 | builtin_f_type (gdbarch)->builtin_integer, | |
147 | 1, ndimensions); | |
148 | struct type *elm_type = builtin_f_type (gdbarch)->builtin_integer; | |
149 | struct type *result_type = create_array_type (alloc, elm_type, range); | |
150 | struct value *result = value::allocate (result_type); | |
151 | ||
152 | /* Walk the array dimensions backwards due to the way the array will be | |
153 | laid out in memory, the first dimension will be the most inner. */ | |
154 | LONGEST elm_len = elm_type->length (); | |
155 | for (LONGEST dst_offset = elm_len * (ndimensions - 1); | |
156 | dst_offset >= 0; | |
157 | dst_offset -= elm_len) | |
158 | { | |
159 | LONGEST b; | |
160 | ||
161 | /* Grab the required bound. */ | |
162 | if (lbound_p) | |
163 | b = f77_get_lowerbound (array_type); | |
164 | else | |
165 | b = f77_get_upperbound (array_type); | |
166 | ||
167 | /* And copy the value into the result value. */ | |
168 | struct value *v = value_from_longest (elm_type, b); | |
169 | gdb_assert (dst_offset + v->type ()->length () | |
170 | <= result->type ()->length ()); | |
171 | gdb_assert (v->type ()->length () == elm_len); | |
172 | v->contents_copy (result, dst_offset, 0, elm_len); | |
173 | ||
174 | /* Peel another dimension of the array. */ | |
175 | array_type = array_type->target_type (); | |
176 | } | |
177 | ||
178 | return result; | |
179 | } | |
180 | ||
181 | /* Return the lower bound (when LBOUND_P is true) or the upper bound (when | |
182 | LBOUND_P is false) for dimension DIM_VAL (which must be an integer) of | |
183 | ARRAY (which must be an array). RESULT_TYPE corresponds to the type kind | |
184 | the function should be evaluated in. */ | |
185 | ||
186 | static value * | |
187 | fortran_bounds_for_dimension (bool lbound_p, value *array, value *dim_val, | |
188 | type* result_type) | |
189 | { | |
190 | /* Check the requested dimension is valid for this array. */ | |
191 | type *array_type = check_typedef (array->type ()); | |
192 | int ndimensions = calc_f77_array_dims (array_type); | |
193 | long dim = value_as_long (dim_val); | |
194 | if (dim < 1 || dim > ndimensions) | |
195 | { | |
196 | if (lbound_p) | |
197 | error (_("LBOUND dimension must be from 1 to %d"), ndimensions); | |
198 | else | |
199 | error (_("UBOUND dimension must be from 1 to %d"), ndimensions); | |
200 | } | |
201 | ||
202 | /* Walk the dimensions backwards, due to the ordering in which arrays are | |
203 | laid out the first dimension is the most inner. */ | |
204 | for (int i = ndimensions - 1; i >= 0; --i) | |
205 | { | |
206 | /* If this is the requested dimension then we're done. Grab the | |
207 | bounds and return. */ | |
208 | if (i == dim - 1) | |
209 | { | |
210 | LONGEST b; | |
211 | ||
212 | if (lbound_p) | |
213 | b = f77_get_lowerbound (array_type); | |
214 | else | |
215 | b = f77_get_upperbound (array_type); | |
216 | ||
217 | return value_from_longest (result_type, b); | |
218 | } | |
219 | ||
220 | /* Peel off another dimension of the array. */ | |
221 | array_type = array_type->target_type (); | |
222 | } | |
223 | ||
224 | gdb_assert_not_reached ("failed to find matching dimension"); | |
225 | } | |
226 | ||
227 | /* Return the number of dimensions for a Fortran array or string. */ | |
228 | ||
229 | int | |
230 | calc_f77_array_dims (struct type *array_type) | |
231 | { | |
232 | int ndimen = 1; | |
233 | struct type *tmp_type; | |
234 | ||
235 | if ((array_type->code () == TYPE_CODE_STRING)) | |
236 | return 1; | |
237 | ||
238 | if ((array_type->code () != TYPE_CODE_ARRAY)) | |
239 | error (_("Can't get dimensions for a non-array type")); | |
240 | ||
241 | tmp_type = array_type; | |
242 | ||
243 | while ((tmp_type = tmp_type->target_type ())) | |
244 | { | |
245 | if (tmp_type->code () == TYPE_CODE_ARRAY) | |
246 | ++ndimen; | |
247 | } | |
248 | return ndimen; | |
249 | } | |
250 | ||
251 | /* A class used by FORTRAN_VALUE_SUBARRAY when repacking Fortran array | |
252 | slices. This is a base class for two alternative repacking mechanisms, | |
253 | one for when repacking from a lazy value, and one for repacking from a | |
254 | non-lazy (already loaded) value. */ | |
255 | class fortran_array_repacker_base_impl | |
256 | : public fortran_array_walker_base_impl | |
257 | { | |
258 | public: | |
259 | /* Constructor, DEST is the value we are repacking into. */ | |
260 | fortran_array_repacker_base_impl (struct value *dest) | |
261 | : m_dest (dest), | |
262 | m_dest_offset (0) | |
263 | { /* Nothing. */ } | |
264 | ||
265 | /* When we start processing the inner most dimension, this is where we | |
266 | will be creating values for each element as we load them and then copy | |
267 | them into the M_DEST value. Set a value mark so we can free these | |
268 | temporary values. */ | |
269 | void start_dimension (struct type *index_type, LONGEST nelts, bool inner_p) | |
270 | { | |
271 | if (inner_p) | |
272 | { | |
273 | gdb_assert (!m_mark.has_value ()); | |
274 | m_mark.emplace (); | |
275 | } | |
276 | } | |
277 | ||
278 | /* When we finish processing the inner most dimension free all temporary | |
279 | value that were created. */ | |
280 | void finish_dimension (bool inner_p, bool last_p) | |
281 | { | |
282 | if (inner_p) | |
283 | { | |
284 | gdb_assert (m_mark.has_value ()); | |
285 | m_mark.reset (); | |
286 | } | |
287 | } | |
288 | ||
289 | protected: | |
290 | /* Copy the contents of array element ELT into M_DEST at the next | |
291 | available offset. */ | |
292 | void copy_element_to_dest (struct value *elt) | |
293 | { | |
294 | elt->contents_copy (m_dest, m_dest_offset, 0, | |
295 | elt->type ()->length ()); | |
296 | m_dest_offset += elt->type ()->length (); | |
297 | } | |
298 | ||
299 | /* The value being written to. */ | |
300 | struct value *m_dest; | |
301 | ||
302 | /* The byte offset in M_DEST at which the next element should be | |
303 | written. */ | |
304 | LONGEST m_dest_offset; | |
305 | ||
306 | /* Set and reset to handle removing intermediate values from the | |
307 | value chain. */ | |
308 | std::optional<scoped_value_mark> m_mark; | |
309 | }; | |
310 | ||
311 | /* A class used by FORTRAN_VALUE_SUBARRAY when repacking Fortran array | |
312 | slices. This class is specialised for repacking an array slice from a | |
313 | lazy array value, as such it does not require the parent array value to | |
314 | be loaded into GDB's memory; the parent value could be huge, while the | |
315 | slice could be tiny. */ | |
316 | class fortran_lazy_array_repacker_impl | |
317 | : public fortran_array_repacker_base_impl | |
318 | { | |
319 | public: | |
320 | /* Constructor. TYPE is the type of the slice being loaded from the | |
321 | parent value, so this type will correctly reflect the strides required | |
322 | to find all of the elements from the parent value. ADDRESS is the | |
323 | address in target memory of value matching TYPE, and DEST is the value | |
324 | we are repacking into. */ | |
325 | explicit fortran_lazy_array_repacker_impl (struct type *type, | |
326 | CORE_ADDR address, | |
327 | struct value *dest) | |
328 | : fortran_array_repacker_base_impl (dest), | |
329 | m_addr (address) | |
330 | { /* Nothing. */ } | |
331 | ||
332 | /* Create a lazy value in target memory representing a single element, | |
333 | then load the element into GDB's memory and copy the contents into the | |
334 | destination value. */ | |
335 | void process_element (struct type *elt_type, LONGEST elt_off, | |
336 | LONGEST index, bool last_p) | |
337 | { | |
338 | copy_element_to_dest (value_at_lazy (elt_type, m_addr + elt_off)); | |
339 | } | |
340 | ||
341 | private: | |
342 | /* The address in target memory where the parent value starts. */ | |
343 | CORE_ADDR m_addr; | |
344 | }; | |
345 | ||
346 | /* A class used by FORTRAN_VALUE_SUBARRAY when repacking Fortran array | |
347 | slices. This class is specialised for repacking an array slice from a | |
348 | previously loaded (non-lazy) array value, as such it fetches the | |
349 | element values from the contents of the parent value. */ | |
350 | class fortran_array_repacker_impl | |
351 | : public fortran_array_repacker_base_impl | |
352 | { | |
353 | public: | |
354 | /* Constructor. TYPE is the type for the array slice within the parent | |
355 | value, as such it has stride values as required to find the elements | |
356 | within the original parent value. ADDRESS is the address in target | |
357 | memory of the value matching TYPE. BASE_OFFSET is the offset from | |
358 | the start of VAL's content buffer to the start of the object of TYPE, | |
359 | VAL is the parent object from which we are loading the value, and | |
360 | DEST is the value into which we are repacking. */ | |
361 | explicit fortran_array_repacker_impl (struct type *type, CORE_ADDR address, | |
362 | LONGEST base_offset, | |
363 | struct value *val, struct value *dest) | |
364 | : fortran_array_repacker_base_impl (dest), | |
365 | m_base_offset (base_offset), | |
366 | m_val (val) | |
367 | { | |
368 | gdb_assert (!val->lazy ()); | |
369 | } | |
370 | ||
371 | /* Extract an element of ELT_TYPE at offset (M_BASE_OFFSET + ELT_OFF) | |
372 | from the content buffer of M_VAL then copy this extracted value into | |
373 | the repacked destination value. */ | |
374 | void process_element (struct type *elt_type, LONGEST elt_off, | |
375 | LONGEST index, bool last_p) | |
376 | { | |
377 | struct value *elt | |
378 | = value_from_component (m_val, elt_type, (elt_off + m_base_offset)); | |
379 | copy_element_to_dest (elt); | |
380 | } | |
381 | ||
382 | private: | |
383 | /* The offset into the content buffer of M_VAL to the start of the slice | |
384 | being extracted. */ | |
385 | LONGEST m_base_offset; | |
386 | ||
387 | /* The parent value from which we are extracting a slice. */ | |
388 | struct value *m_val; | |
389 | }; | |
390 | ||
391 | ||
392 | /* Evaluate FORTRAN_ASSOCIATED expressions. Both GDBARCH and LANG are | |
393 | extracted from the expression being evaluated. POINTER is the required | |
394 | first argument to the 'associated' keyword, and TARGET is the optional | |
395 | second argument, this will be nullptr if the user only passed one | |
396 | argument to their use of 'associated'. */ | |
397 | ||
398 | static struct value * | |
399 | fortran_associated (struct gdbarch *gdbarch, const language_defn *lang, | |
400 | struct value *pointer, struct value *target = nullptr) | |
401 | { | |
402 | struct type *result_type = language_bool_type (lang, gdbarch); | |
403 | ||
404 | /* All Fortran pointers should have the associated property, this is | |
405 | how we know the pointer is pointing at something or not. */ | |
406 | struct type *pointer_type = check_typedef (pointer->type ()); | |
407 | if (TYPE_ASSOCIATED_PROP (pointer_type) == nullptr | |
408 | && pointer_type->code () != TYPE_CODE_PTR) | |
409 | error (_("ASSOCIATED can only be applied to pointers")); | |
410 | ||
411 | /* Get an address from POINTER. Fortran (or at least gfortran) models | |
412 | array pointers as arrays with a dynamic data address, so we need to | |
413 | use two approaches here, for real pointers we take the contents of the | |
414 | pointer as an address. For non-pointers we take the address of the | |
415 | content. */ | |
416 | CORE_ADDR pointer_addr; | |
417 | if (pointer_type->code () == TYPE_CODE_PTR) | |
418 | pointer_addr = value_as_address (pointer); | |
419 | else | |
420 | pointer_addr = pointer->address (); | |
421 | ||
422 | /* The single argument case, is POINTER associated with anything? */ | |
423 | if (target == nullptr) | |
424 | { | |
425 | bool is_associated = false; | |
426 | ||
427 | /* If POINTER is an actual pointer and doesn't have an associated | |
428 | property then we need to figure out whether this pointer is | |
429 | associated by looking at the value of the pointer itself. We make | |
430 | the assumption that a non-associated pointer will be set to 0. | |
431 | This is probably true for most targets, but might not be true for | |
432 | everyone. */ | |
433 | if (pointer_type->code () == TYPE_CODE_PTR | |
434 | && TYPE_ASSOCIATED_PROP (pointer_type) == nullptr) | |
435 | is_associated = (pointer_addr != 0); | |
436 | else | |
437 | is_associated = !type_not_associated (pointer_type); | |
438 | return value_from_longest (result_type, is_associated ? 1 : 0); | |
439 | } | |
440 | ||
441 | /* The two argument case, is POINTER associated with TARGET? */ | |
442 | ||
443 | struct type *target_type = check_typedef (target->type ()); | |
444 | ||
445 | struct type *pointer_target_type; | |
446 | if (pointer_type->code () == TYPE_CODE_PTR) | |
447 | pointer_target_type = pointer_type->target_type (); | |
448 | else | |
449 | pointer_target_type = pointer_type; | |
450 | ||
451 | struct type *target_target_type; | |
452 | if (target_type->code () == TYPE_CODE_PTR) | |
453 | target_target_type = target_type->target_type (); | |
454 | else | |
455 | target_target_type = target_type; | |
456 | ||
457 | if (pointer_target_type->code () != target_target_type->code () | |
458 | || (pointer_target_type->code () != TYPE_CODE_ARRAY | |
459 | && (pointer_target_type->length () | |
460 | != target_target_type->length ()))) | |
461 | error (_("arguments to associated must be of same type and kind")); | |
462 | ||
463 | /* If TARGET is not in memory, or the original pointer is specifically | |
464 | known to be not associated with anything, then the answer is obviously | |
465 | false. Alternatively, if POINTER is an actual pointer and has no | |
466 | associated property, then we have to check if its associated by | |
467 | looking the value of the pointer itself. We make the assumption that | |
468 | a non-associated pointer will be set to 0. This is probably true for | |
469 | most targets, but might not be true for everyone. */ | |
470 | if (target->lval () != lval_memory | |
471 | || type_not_associated (pointer_type) | |
472 | || (TYPE_ASSOCIATED_PROP (pointer_type) == nullptr | |
473 | && pointer_type->code () == TYPE_CODE_PTR | |
474 | && pointer_addr == 0)) | |
475 | return value_from_longest (result_type, 0); | |
476 | ||
477 | /* See the comment for POINTER_ADDR above. */ | |
478 | CORE_ADDR target_addr; | |
479 | if (target_type->code () == TYPE_CODE_PTR) | |
480 | target_addr = value_as_address (target); | |
481 | else | |
482 | target_addr = target->address (); | |
483 | ||
484 | /* Wrap the following checks inside a do { ... } while (false) loop so | |
485 | that we can use `break' to jump out of the loop. */ | |
486 | bool is_associated = false; | |
487 | do | |
488 | { | |
489 | /* If the addresses are different then POINTER is definitely not | |
490 | pointing at TARGET. */ | |
491 | if (pointer_addr != target_addr) | |
492 | break; | |
493 | ||
494 | /* If POINTER is a real pointer (i.e. not an array pointer, which are | |
495 | implemented as arrays with a dynamic content address), then this | |
496 | is all the checking that is needed. */ | |
497 | if (pointer_type->code () == TYPE_CODE_PTR) | |
498 | { | |
499 | is_associated = true; | |
500 | break; | |
501 | } | |
502 | ||
503 | /* We have an array pointer. Check the number of dimensions. */ | |
504 | int pointer_dims = calc_f77_array_dims (pointer_type); | |
505 | int target_dims = calc_f77_array_dims (target_type); | |
506 | if (pointer_dims != target_dims) | |
507 | break; | |
508 | ||
509 | /* Now check that every dimension has the same upper bound, lower | |
510 | bound, and stride value. */ | |
511 | int dim = 0; | |
512 | while (dim < pointer_dims) | |
513 | { | |
514 | LONGEST pointer_lowerbound, pointer_upperbound, pointer_stride; | |
515 | LONGEST target_lowerbound, target_upperbound, target_stride; | |
516 | ||
517 | pointer_type = check_typedef (pointer_type); | |
518 | target_type = check_typedef (target_type); | |
519 | ||
520 | struct type *pointer_range = pointer_type->index_type (); | |
521 | struct type *target_range = target_type->index_type (); | |
522 | ||
523 | if (!get_discrete_bounds (pointer_range, &pointer_lowerbound, | |
524 | &pointer_upperbound)) | |
525 | break; | |
526 | ||
527 | if (!get_discrete_bounds (target_range, &target_lowerbound, | |
528 | &target_upperbound)) | |
529 | break; | |
530 | ||
531 | if (pointer_lowerbound != target_lowerbound | |
532 | || pointer_upperbound != target_upperbound) | |
533 | break; | |
534 | ||
535 | /* Figure out the stride (in bits) for both pointer and target. | |
536 | If either doesn't have a stride then we take the element size, | |
537 | but we need to convert to bits (hence the * 8). */ | |
538 | pointer_stride = pointer_range->bounds ()->bit_stride (); | |
539 | if (pointer_stride == 0) | |
540 | pointer_stride | |
541 | = type_length_units (check_typedef | |
542 | (pointer_type->target_type ())) * 8; | |
543 | target_stride = target_range->bounds ()->bit_stride (); | |
544 | if (target_stride == 0) | |
545 | target_stride | |
546 | = type_length_units (check_typedef | |
547 | (target_type->target_type ())) * 8; | |
548 | if (pointer_stride != target_stride) | |
549 | break; | |
550 | ||
551 | ++dim; | |
552 | } | |
553 | ||
554 | if (dim < pointer_dims) | |
555 | break; | |
556 | ||
557 | is_associated = true; | |
558 | } | |
559 | while (false); | |
560 | ||
561 | return value_from_longest (result_type, is_associated ? 1 : 0); | |
562 | } | |
563 | ||
564 | struct value * | |
565 | eval_op_f_associated (struct type *expect_type, | |
566 | struct expression *exp, | |
567 | enum noside noside, | |
568 | enum exp_opcode opcode, | |
569 | struct value *arg1) | |
570 | { | |
571 | return fortran_associated (exp->gdbarch, exp->language_defn, arg1); | |
572 | } | |
573 | ||
574 | struct value * | |
575 | eval_op_f_associated (struct type *expect_type, | |
576 | struct expression *exp, | |
577 | enum noside noside, | |
578 | enum exp_opcode opcode, | |
579 | struct value *arg1, | |
580 | struct value *arg2) | |
581 | { | |
582 | return fortran_associated (exp->gdbarch, exp->language_defn, arg1, arg2); | |
583 | } | |
584 | ||
585 | /* Implement FORTRAN_ARRAY_SIZE expression, this corresponds to the 'SIZE' | |
586 | keyword. RESULT_TYPE corresponds to the type kind the function should be | |
587 | evaluated in, ARRAY is the value that should be an array, though this will | |
588 | not have been checked before calling this function. DIM is optional, if | |
589 | present then it should be an integer identifying a dimension of the | |
590 | array to ask about. As with ARRAY the validity of DIM is not checked | |
591 | before calling this function. | |
592 | ||
593 | Return either the total number of elements in ARRAY (when DIM is | |
594 | nullptr), or the number of elements in dimension DIM. */ | |
595 | ||
596 | static value * | |
597 | fortran_array_size (value *array, value *dim_val, type *result_type) | |
598 | { | |
599 | /* Check that ARRAY is the correct type. */ | |
600 | struct type *array_type = check_typedef (array->type ()); | |
601 | if (array_type->code () != TYPE_CODE_ARRAY) | |
602 | error (_("SIZE can only be applied to arrays")); | |
603 | if (type_not_allocated (array_type) || type_not_associated (array_type)) | |
604 | error (_("SIZE can only be used on allocated/associated arrays")); | |
605 | ||
606 | int ndimensions = calc_f77_array_dims (array_type); | |
607 | int dim = -1; | |
608 | LONGEST result = 0; | |
609 | ||
610 | if (dim_val != nullptr) | |
611 | { | |
612 | if (check_typedef (dim_val->type ())->code () != TYPE_CODE_INT) | |
613 | error (_("DIM argument to SIZE must be an integer")); | |
614 | dim = (int) value_as_long (dim_val); | |
615 | ||
616 | if (dim < 1 || dim > ndimensions) | |
617 | error (_("DIM argument to SIZE must be between 1 and %d"), | |
618 | ndimensions); | |
619 | } | |
620 | ||
621 | /* Now walk over all the dimensions of the array totalling up the | |
622 | elements in each dimension. */ | |
623 | for (int i = ndimensions - 1; i >= 0; --i) | |
624 | { | |
625 | /* If this is the requested dimension then we're done. Grab the | |
626 | bounds and return. */ | |
627 | if (i == dim - 1 || dim == -1) | |
628 | { | |
629 | LONGEST lbound, ubound; | |
630 | struct type *range = array_type->index_type (); | |
631 | ||
632 | if (!get_discrete_bounds (range, &lbound, &ubound)) | |
633 | error (_("failed to find array bounds")); | |
634 | ||
635 | LONGEST dim_size = (ubound - lbound + 1); | |
636 | if (result == 0) | |
637 | result = dim_size; | |
638 | else | |
639 | result *= dim_size; | |
640 | ||
641 | if (dim != -1) | |
642 | break; | |
643 | } | |
644 | ||
645 | /* Peel off another dimension of the array. */ | |
646 | array_type = array_type->target_type (); | |
647 | } | |
648 | ||
649 | return value_from_longest (result_type, result); | |
650 | } | |
651 | ||
652 | /* See f-exp.h. */ | |
653 | ||
654 | struct value * | |
655 | eval_op_f_array_size (struct type *expect_type, | |
656 | struct expression *exp, | |
657 | enum noside noside, | |
658 | enum exp_opcode opcode, | |
659 | struct value *arg1) | |
660 | { | |
661 | gdb_assert (opcode == FORTRAN_ARRAY_SIZE); | |
662 | ||
663 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
664 | return fortran_array_size (arg1, nullptr, result_type); | |
665 | } | |
666 | ||
667 | /* See f-exp.h. */ | |
668 | ||
669 | struct value * | |
670 | eval_op_f_array_size (struct type *expect_type, | |
671 | struct expression *exp, | |
672 | enum noside noside, | |
673 | enum exp_opcode opcode, | |
674 | struct value *arg1, | |
675 | struct value *arg2) | |
676 | { | |
677 | gdb_assert (opcode == FORTRAN_ARRAY_SIZE); | |
678 | ||
679 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
680 | return fortran_array_size (arg1, arg2, result_type); | |
681 | } | |
682 | ||
683 | /* See f-exp.h. */ | |
684 | ||
685 | value *eval_op_f_array_size (type *expect_type, expression *exp, noside noside, | |
686 | exp_opcode opcode, value *arg1, value *arg2, | |
687 | type *kind_arg) | |
688 | { | |
689 | gdb_assert (opcode == FORTRAN_ARRAY_SIZE); | |
690 | gdb_assert (kind_arg->code () == TYPE_CODE_INT); | |
691 | ||
692 | return fortran_array_size (arg1, arg2, kind_arg); | |
693 | } | |
694 | ||
695 | /* Implement UNOP_FORTRAN_SHAPE expression. Both GDBARCH and LANG are | |
696 | extracted from the expression being evaluated. VAL is the value on | |
697 | which 'shape' was used, this can be any type. | |
698 | ||
699 | Return an array of integers. If VAL is not an array then the returned | |
700 | array should have zero elements. If VAL is an array then the returned | |
701 | array should have one element per dimension, with the element | |
702 | containing the extent of that dimension from VAL. */ | |
703 | ||
704 | static struct value * | |
705 | fortran_array_shape (struct gdbarch *gdbarch, const language_defn *lang, | |
706 | struct value *val) | |
707 | { | |
708 | struct type *val_type = check_typedef (val->type ()); | |
709 | ||
710 | /* If we are passed an array that is either not allocated, or not | |
711 | associated, then this is explicitly not allowed according to the | |
712 | Fortran specification. */ | |
713 | if (val_type->code () == TYPE_CODE_ARRAY | |
714 | && (type_not_associated (val_type) || type_not_allocated (val_type))) | |
715 | error (_("The array passed to SHAPE must be allocated or associated")); | |
716 | ||
717 | /* The Fortran specification allows non-array types to be passed to this | |
718 | function, in which case we get back an empty array. | |
719 | ||
720 | Calculate the number of dimensions for the resulting array. */ | |
721 | int ndimensions = 0; | |
722 | if (val_type->code () == TYPE_CODE_ARRAY) | |
723 | ndimensions = calc_f77_array_dims (val_type); | |
724 | ||
725 | /* Allocate a result value of the correct type. */ | |
726 | type_allocator alloc (gdbarch); | |
727 | struct type *range | |
728 | = create_static_range_type (alloc, | |
729 | builtin_type (gdbarch)->builtin_int, | |
730 | 1, ndimensions); | |
731 | struct type *elm_type = builtin_f_type (gdbarch)->builtin_integer; | |
732 | struct type *result_type = create_array_type (alloc, elm_type, range); | |
733 | struct value *result = value::allocate (result_type); | |
734 | LONGEST elm_len = elm_type->length (); | |
735 | ||
736 | /* Walk the array dimensions backwards due to the way the array will be | |
737 | laid out in memory, the first dimension will be the most inner. | |
738 | ||
739 | If VAL was not an array then ndimensions will be 0, in which case we | |
740 | will never go around this loop. */ | |
741 | for (LONGEST dst_offset = elm_len * (ndimensions - 1); | |
742 | dst_offset >= 0; | |
743 | dst_offset -= elm_len) | |
744 | { | |
745 | LONGEST lbound, ubound; | |
746 | ||
747 | if (!get_discrete_bounds (val_type->index_type (), &lbound, &ubound)) | |
748 | error (_("failed to find array bounds")); | |
749 | ||
750 | LONGEST dim_size = (ubound - lbound + 1); | |
751 | ||
752 | /* And copy the value into the result value. */ | |
753 | struct value *v = value_from_longest (elm_type, dim_size); | |
754 | gdb_assert (dst_offset + v->type ()->length () | |
755 | <= result->type ()->length ()); | |
756 | gdb_assert (v->type ()->length () == elm_len); | |
757 | v->contents_copy (result, dst_offset, 0, elm_len); | |
758 | ||
759 | /* Peel another dimension of the array. */ | |
760 | val_type = val_type->target_type (); | |
761 | } | |
762 | ||
763 | return result; | |
764 | } | |
765 | ||
766 | /* See f-exp.h. */ | |
767 | ||
768 | struct value * | |
769 | eval_op_f_array_shape (struct type *expect_type, struct expression *exp, | |
770 | enum noside noside, enum exp_opcode opcode, | |
771 | struct value *arg1) | |
772 | { | |
773 | gdb_assert (opcode == UNOP_FORTRAN_SHAPE); | |
774 | return fortran_array_shape (exp->gdbarch, exp->language_defn, arg1); | |
775 | } | |
776 | ||
777 | /* A helper function for UNOP_ABS. */ | |
778 | ||
779 | struct value * | |
780 | eval_op_f_abs (struct type *expect_type, struct expression *exp, | |
781 | enum noside noside, | |
782 | enum exp_opcode opcode, | |
783 | struct value *arg1) | |
784 | { | |
785 | struct type *type = arg1->type (); | |
786 | switch (type->code ()) | |
787 | { | |
788 | case TYPE_CODE_FLT: | |
789 | { | |
790 | double d | |
791 | = fabs (target_float_to_host_double (arg1->contents ().data (), | |
792 | arg1->type ())); | |
793 | return value_from_host_double (type, d); | |
794 | } | |
795 | case TYPE_CODE_INT: | |
796 | { | |
797 | LONGEST l = value_as_long (arg1); | |
798 | l = llabs (l); | |
799 | return value_from_longest (type, l); | |
800 | } | |
801 | } | |
802 | error (_("ABS of type %s not supported"), TYPE_SAFE_NAME (type)); | |
803 | } | |
804 | ||
805 | /* A helper function for BINOP_MOD. */ | |
806 | ||
807 | struct value * | |
808 | eval_op_f_mod (struct type *expect_type, struct expression *exp, | |
809 | enum noside noside, | |
810 | enum exp_opcode opcode, | |
811 | struct value *arg1, struct value *arg2) | |
812 | { | |
813 | struct type *type = arg1->type (); | |
814 | if (type->code () != arg2->type ()->code ()) | |
815 | error (_("non-matching types for parameters to MOD ()")); | |
816 | switch (type->code ()) | |
817 | { | |
818 | case TYPE_CODE_FLT: | |
819 | { | |
820 | double d1 | |
821 | = target_float_to_host_double (arg1->contents ().data (), | |
822 | arg1->type ()); | |
823 | double d2 | |
824 | = target_float_to_host_double (arg2->contents ().data (), | |
825 | arg2->type ()); | |
826 | double d3 = fmod (d1, d2); | |
827 | return value_from_host_double (type, d3); | |
828 | } | |
829 | case TYPE_CODE_INT: | |
830 | { | |
831 | LONGEST v1 = value_as_long (arg1); | |
832 | LONGEST v2 = value_as_long (arg2); | |
833 | if (v2 == 0) | |
834 | error (_("calling MOD (N, 0) is undefined")); | |
835 | LONGEST v3 = v1 - (v1 / v2) * v2; | |
836 | return value_from_longest (arg1->type (), v3); | |
837 | } | |
838 | } | |
839 | error (_("MOD of type %s not supported"), TYPE_SAFE_NAME (type)); | |
840 | } | |
841 | ||
842 | /* A helper function for the different FORTRAN_CEILING overloads. Calculates | |
843 | CEILING for ARG1 (a float type) and returns it in the requested kind type | |
844 | RESULT_TYPE. */ | |
845 | ||
846 | static value * | |
847 | fortran_ceil_operation (value *arg1, type *result_type) | |
848 | { | |
849 | if (arg1->type ()->code () != TYPE_CODE_FLT) | |
850 | error (_("argument to CEILING must be of type float")); | |
851 | double val = target_float_to_host_double (arg1->contents ().data (), | |
852 | arg1->type ()); | |
853 | val = ceil (val); | |
854 | return value_from_longest (result_type, val); | |
855 | } | |
856 | ||
857 | /* A helper function for FORTRAN_CEILING. */ | |
858 | ||
859 | struct value * | |
860 | eval_op_f_ceil (struct type *expect_type, struct expression *exp, | |
861 | enum noside noside, | |
862 | enum exp_opcode opcode, | |
863 | struct value *arg1) | |
864 | { | |
865 | gdb_assert (opcode == FORTRAN_CEILING); | |
866 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
867 | return fortran_ceil_operation (arg1, result_type); | |
868 | } | |
869 | ||
870 | /* A helper function for FORTRAN_CEILING. */ | |
871 | ||
872 | value * | |
873 | eval_op_f_ceil (type *expect_type, expression *exp, noside noside, | |
874 | exp_opcode opcode, value *arg1, type *kind_arg) | |
875 | { | |
876 | gdb_assert (opcode == FORTRAN_CEILING); | |
877 | gdb_assert (kind_arg->code () == TYPE_CODE_INT); | |
878 | return fortran_ceil_operation (arg1, kind_arg); | |
879 | } | |
880 | ||
881 | /* A helper function for the different FORTRAN_FLOOR overloads. Calculates | |
882 | FLOOR for ARG1 (a float type) and returns it in the requested kind type | |
883 | RESULT_TYPE. */ | |
884 | ||
885 | static value * | |
886 | fortran_floor_operation (value *arg1, type *result_type) | |
887 | { | |
888 | if (arg1->type ()->code () != TYPE_CODE_FLT) | |
889 | error (_("argument to FLOOR must be of type float")); | |
890 | double val = target_float_to_host_double (arg1->contents ().data (), | |
891 | arg1->type ()); | |
892 | val = floor (val); | |
893 | return value_from_longest (result_type, val); | |
894 | } | |
895 | ||
896 | /* A helper function for FORTRAN_FLOOR. */ | |
897 | ||
898 | struct value * | |
899 | eval_op_f_floor (struct type *expect_type, struct expression *exp, | |
900 | enum noside noside, | |
901 | enum exp_opcode opcode, | |
902 | struct value *arg1) | |
903 | { | |
904 | gdb_assert (opcode == FORTRAN_FLOOR); | |
905 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
906 | return fortran_floor_operation (arg1, result_type); | |
907 | } | |
908 | ||
909 | /* A helper function for FORTRAN_FLOOR. */ | |
910 | ||
911 | struct value * | |
912 | eval_op_f_floor (type *expect_type, expression *exp, noside noside, | |
913 | exp_opcode opcode, value *arg1, type *kind_arg) | |
914 | { | |
915 | gdb_assert (opcode == FORTRAN_FLOOR); | |
916 | gdb_assert (kind_arg->code () == TYPE_CODE_INT); | |
917 | return fortran_floor_operation (arg1, kind_arg); | |
918 | } | |
919 | ||
920 | /* A helper function for BINOP_FORTRAN_MODULO. */ | |
921 | ||
922 | struct value * | |
923 | eval_op_f_modulo (struct type *expect_type, struct expression *exp, | |
924 | enum noside noside, | |
925 | enum exp_opcode opcode, | |
926 | struct value *arg1, struct value *arg2) | |
927 | { | |
928 | struct type *type = arg1->type (); | |
929 | if (type->code () != arg2->type ()->code ()) | |
930 | error (_("non-matching types for parameters to MODULO ()")); | |
931 | /* MODULO(A, P) = A - FLOOR (A / P) * P */ | |
932 | switch (type->code ()) | |
933 | { | |
934 | case TYPE_CODE_INT: | |
935 | { | |
936 | LONGEST a = value_as_long (arg1); | |
937 | LONGEST p = value_as_long (arg2); | |
938 | LONGEST result = a - (a / p) * p; | |
939 | if (result != 0 && (a < 0) != (p < 0)) | |
940 | result += p; | |
941 | return value_from_longest (arg1->type (), result); | |
942 | } | |
943 | case TYPE_CODE_FLT: | |
944 | { | |
945 | double a | |
946 | = target_float_to_host_double (arg1->contents ().data (), | |
947 | arg1->type ()); | |
948 | double p | |
949 | = target_float_to_host_double (arg2->contents ().data (), | |
950 | arg2->type ()); | |
951 | double result = fmod (a, p); | |
952 | if (result != 0 && (a < 0.0) != (p < 0.0)) | |
953 | result += p; | |
954 | return value_from_host_double (type, result); | |
955 | } | |
956 | } | |
957 | error (_("MODULO of type %s not supported"), TYPE_SAFE_NAME (type)); | |
958 | } | |
959 | ||
960 | /* A helper function for FORTRAN_CMPLX. */ | |
961 | ||
962 | value * | |
963 | eval_op_f_cmplx (type *expect_type, expression *exp, noside noside, | |
964 | exp_opcode opcode, value *arg1) | |
965 | { | |
966 | gdb_assert (opcode == FORTRAN_CMPLX); | |
967 | ||
968 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_complex; | |
969 | ||
970 | if (arg1->type ()->code () == TYPE_CODE_COMPLEX) | |
971 | return value_cast (result_type, arg1); | |
972 | else | |
973 | return value_literal_complex (arg1, | |
974 | value::zero (arg1->type (), not_lval), | |
975 | result_type); | |
976 | } | |
977 | ||
978 | /* A helper function for FORTRAN_CMPLX. */ | |
979 | ||
980 | struct value * | |
981 | eval_op_f_cmplx (struct type *expect_type, struct expression *exp, | |
982 | enum noside noside, | |
983 | enum exp_opcode opcode, | |
984 | struct value *arg1, struct value *arg2) | |
985 | { | |
986 | if (arg1->type ()->code () == TYPE_CODE_COMPLEX | |
987 | || arg2->type ()->code () == TYPE_CODE_COMPLEX) | |
988 | error (_("Types of arguments for CMPLX called with more then one argument " | |
989 | "must be REAL or INTEGER")); | |
990 | ||
991 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_complex; | |
992 | return value_literal_complex (arg1, arg2, result_type); | |
993 | } | |
994 | ||
995 | /* A helper function for FORTRAN_CMPLX. */ | |
996 | ||
997 | value * | |
998 | eval_op_f_cmplx (type *expect_type, expression *exp, noside noside, | |
999 | exp_opcode opcode, value *arg1, value *arg2, type *kind_arg) | |
1000 | { | |
1001 | gdb_assert (kind_arg->code () == TYPE_CODE_COMPLEX); | |
1002 | if (arg1->type ()->code () == TYPE_CODE_COMPLEX | |
1003 | || arg2->type ()->code () == TYPE_CODE_COMPLEX) | |
1004 | error (_("Types of arguments for CMPLX called with more then one argument " | |
1005 | "must be REAL or INTEGER")); | |
1006 | ||
1007 | return value_literal_complex (arg1, arg2, kind_arg); | |
1008 | } | |
1009 | ||
1010 | /* A helper function for UNOP_FORTRAN_KIND. */ | |
1011 | ||
1012 | struct value * | |
1013 | eval_op_f_kind (struct type *expect_type, struct expression *exp, | |
1014 | enum noside noside, | |
1015 | enum exp_opcode opcode, | |
1016 | struct value *arg1) | |
1017 | { | |
1018 | struct type *type = arg1->type (); | |
1019 | ||
1020 | switch (type->code ()) | |
1021 | { | |
1022 | case TYPE_CODE_STRUCT: | |
1023 | case TYPE_CODE_UNION: | |
1024 | case TYPE_CODE_MODULE: | |
1025 | case TYPE_CODE_FUNC: | |
1026 | error (_("argument to kind must be an intrinsic type")); | |
1027 | } | |
1028 | ||
1029 | if (!type->target_type ()) | |
1030 | return value_from_longest (builtin_type (exp->gdbarch)->builtin_int, | |
1031 | type->length ()); | |
1032 | return value_from_longest (builtin_type (exp->gdbarch)->builtin_int, | |
1033 | type->target_type ()->length ()); | |
1034 | } | |
1035 | ||
1036 | /* A helper function for UNOP_FORTRAN_ALLOCATED. */ | |
1037 | ||
1038 | struct value * | |
1039 | eval_op_f_allocated (struct type *expect_type, struct expression *exp, | |
1040 | enum noside noside, enum exp_opcode op, | |
1041 | struct value *arg1) | |
1042 | { | |
1043 | struct type *type = check_typedef (arg1->type ()); | |
1044 | if (type->code () != TYPE_CODE_ARRAY) | |
1045 | error (_("ALLOCATED can only be applied to arrays")); | |
1046 | struct type *result_type | |
1047 | = builtin_f_type (exp->gdbarch)->builtin_logical; | |
1048 | LONGEST result_value = type_not_allocated (type) ? 0 : 1; | |
1049 | return value_from_longest (result_type, result_value); | |
1050 | } | |
1051 | ||
1052 | /* See f-exp.h. */ | |
1053 | ||
1054 | struct value * | |
1055 | eval_op_f_rank (struct type *expect_type, | |
1056 | struct expression *exp, | |
1057 | enum noside noside, | |
1058 | enum exp_opcode op, | |
1059 | struct value *arg1) | |
1060 | { | |
1061 | gdb_assert (op == UNOP_FORTRAN_RANK); | |
1062 | ||
1063 | struct type *result_type | |
1064 | = builtin_f_type (exp->gdbarch)->builtin_integer; | |
1065 | struct type *type = check_typedef (arg1->type ()); | |
1066 | if (type->code () != TYPE_CODE_ARRAY) | |
1067 | return value_from_longest (result_type, 0); | |
1068 | LONGEST ndim = calc_f77_array_dims (type); | |
1069 | return value_from_longest (result_type, ndim); | |
1070 | } | |
1071 | ||
1072 | /* A helper function for UNOP_FORTRAN_LOC. */ | |
1073 | ||
1074 | struct value * | |
1075 | eval_op_f_loc (struct type *expect_type, struct expression *exp, | |
1076 | enum noside noside, enum exp_opcode op, | |
1077 | struct value *arg1) | |
1078 | { | |
1079 | struct type *result_type; | |
1080 | if (gdbarch_ptr_bit (exp->gdbarch) == 16) | |
1081 | result_type = builtin_f_type (exp->gdbarch)->builtin_integer_s2; | |
1082 | else if (gdbarch_ptr_bit (exp->gdbarch) == 32) | |
1083 | result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
1084 | else | |
1085 | result_type = builtin_f_type (exp->gdbarch)->builtin_integer_s8; | |
1086 | ||
1087 | LONGEST result_value = arg1->address (); | |
1088 | return value_from_longest (result_type, result_value); | |
1089 | } | |
1090 | ||
1091 | namespace expr | |
1092 | { | |
1093 | ||
1094 | /* Called from evaluate to perform array indexing, and sub-range | |
1095 | extraction, for Fortran. As well as arrays this function also | |
1096 | handles strings as they can be treated like arrays of characters. | |
1097 | ARRAY is the array or string being accessed. EXP and NOSIDE are as | |
1098 | for evaluate. */ | |
1099 | ||
1100 | value * | |
1101 | fortran_undetermined::value_subarray (value *array, | |
1102 | struct expression *exp, | |
1103 | enum noside noside) | |
1104 | { | |
1105 | type *original_array_type = check_typedef (array->type ()); | |
1106 | bool is_string_p = original_array_type->code () == TYPE_CODE_STRING; | |
1107 | const std::vector<operation_up> &ops = std::get<1> (m_storage); | |
1108 | int nargs = ops.size (); | |
1109 | ||
1110 | /* Perform checks for ARRAY not being available. The somewhat overly | |
1111 | complex logic here is just to keep backward compatibility with the | |
1112 | errors that we used to get before FORTRAN_VALUE_SUBARRAY was | |
1113 | rewritten. Maybe a future task would streamline the error messages we | |
1114 | get here, and update all the expected test results. */ | |
1115 | if (ops[0]->opcode () != OP_RANGE) | |
1116 | { | |
1117 | if (type_not_associated (original_array_type)) | |
1118 | error (_("no such vector element (vector not associated)")); | |
1119 | else if (type_not_allocated (original_array_type)) | |
1120 | error (_("no such vector element (vector not allocated)")); | |
1121 | } | |
1122 | else | |
1123 | { | |
1124 | if (type_not_associated (original_array_type)) | |
1125 | error (_("array not associated")); | |
1126 | else if (type_not_allocated (original_array_type)) | |
1127 | error (_("array not allocated")); | |
1128 | } | |
1129 | ||
1130 | /* First check that the number of dimensions in the type we are slicing | |
1131 | matches the number of arguments we were passed. */ | |
1132 | int ndimensions = calc_f77_array_dims (original_array_type); | |
1133 | if (nargs != ndimensions) | |
1134 | error (_("Wrong number of subscripts")); | |
1135 | ||
1136 | /* This will be initialised below with the type of the elements held in | |
1137 | ARRAY. */ | |
1138 | struct type *inner_element_type; | |
1139 | ||
1140 | /* Extract the types of each array dimension from the original array | |
1141 | type. We need these available so we can fill in the default upper and | |
1142 | lower bounds if the user requested slice doesn't provide that | |
1143 | information. Additionally unpacking the dimensions like this gives us | |
1144 | the inner element type. */ | |
1145 | std::vector<struct type *> dim_types; | |
1146 | { | |
1147 | dim_types.reserve (ndimensions); | |
1148 | struct type *type = original_array_type; | |
1149 | for (int i = 0; i < ndimensions; ++i) | |
1150 | { | |
1151 | dim_types.push_back (type); | |
1152 | type = type->target_type (); | |
1153 | } | |
1154 | /* TYPE is now the inner element type of the array, we start the new | |
1155 | array slice off as this type, then as we process the requested slice | |
1156 | (from the user) we wrap new types around this to build up the final | |
1157 | slice type. */ | |
1158 | inner_element_type = type; | |
1159 | } | |
1160 | ||
1161 | /* As we analyse the new slice type we need to understand if the data | |
1162 | being referenced is contiguous. Do decide this we must track the size | |
1163 | of an element at each dimension of the new slice array. Initially the | |
1164 | elements of the inner most dimension of the array are the same inner | |
1165 | most elements as the original ARRAY. */ | |
1166 | LONGEST slice_element_size = inner_element_type->length (); | |
1167 | ||
1168 | /* Start off assuming all data is contiguous, this will be set to false | |
1169 | if access to any dimension results in non-contiguous data. */ | |
1170 | bool is_all_contiguous = true; | |
1171 | ||
1172 | /* The TOTAL_OFFSET is the distance in bytes from the start of the | |
1173 | original ARRAY to the start of the new slice. This is calculated as | |
1174 | we process the information from the user. */ | |
1175 | LONGEST total_offset = 0; | |
1176 | ||
1177 | /* A structure representing information about each dimension of the | |
1178 | resulting slice. */ | |
1179 | struct slice_dim | |
1180 | { | |
1181 | /* Constructor. */ | |
1182 | slice_dim (LONGEST l, LONGEST h, LONGEST s, struct type *idx) | |
1183 | : low (l), | |
1184 | high (h), | |
1185 | stride (s), | |
1186 | index (idx) | |
1187 | { /* Nothing. */ } | |
1188 | ||
1189 | /* The low bound for this dimension of the slice. */ | |
1190 | LONGEST low; | |
1191 | ||
1192 | /* The high bound for this dimension of the slice. */ | |
1193 | LONGEST high; | |
1194 | ||
1195 | /* The byte stride for this dimension of the slice. */ | |
1196 | LONGEST stride; | |
1197 | ||
1198 | struct type *index; | |
1199 | }; | |
1200 | ||
1201 | /* The dimensions of the resulting slice. */ | |
1202 | std::vector<slice_dim> slice_dims; | |
1203 | ||
1204 | /* Process the incoming arguments. These arguments are in the reverse | |
1205 | order to the array dimensions, that is the first argument refers to | |
1206 | the last array dimension. */ | |
1207 | if (fortran_array_slicing_debug) | |
1208 | debug_printf ("Processing array access:\n"); | |
1209 | for (int i = 0; i < nargs; ++i) | |
1210 | { | |
1211 | /* For each dimension of the array the user will have either provided | |
1212 | a ranged access with optional lower bound, upper bound, and | |
1213 | stride, or the user will have supplied a single index. */ | |
1214 | struct type *dim_type = dim_types[ndimensions - (i + 1)]; | |
1215 | fortran_range_operation *range_op | |
1216 | = dynamic_cast<fortran_range_operation *> (ops[i].get ()); | |
1217 | if (range_op != nullptr) | |
1218 | { | |
1219 | enum range_flag range_flag = range_op->get_flags (); | |
1220 | ||
1221 | LONGEST low, high, stride; | |
1222 | low = high = stride = 0; | |
1223 | ||
1224 | if ((range_flag & RANGE_LOW_BOUND_DEFAULT) == 0) | |
1225 | low = value_as_long (range_op->evaluate0 (exp, noside)); | |
1226 | else | |
1227 | low = f77_get_lowerbound (dim_type); | |
1228 | if ((range_flag & RANGE_HIGH_BOUND_DEFAULT) == 0) | |
1229 | high = value_as_long (range_op->evaluate1 (exp, noside)); | |
1230 | else | |
1231 | high = f77_get_upperbound (dim_type); | |
1232 | if ((range_flag & RANGE_HAS_STRIDE) == RANGE_HAS_STRIDE) | |
1233 | stride = value_as_long (range_op->evaluate2 (exp, noside)); | |
1234 | else | |
1235 | stride = 1; | |
1236 | ||
1237 | if (stride == 0) | |
1238 | error (_("stride must not be 0")); | |
1239 | ||
1240 | /* Get information about this dimension in the original ARRAY. */ | |
1241 | struct type *target_type = dim_type->target_type (); | |
1242 | struct type *index_type = dim_type->index_type (); | |
1243 | LONGEST lb = f77_get_lowerbound (dim_type); | |
1244 | LONGEST ub = f77_get_upperbound (dim_type); | |
1245 | LONGEST sd = index_type->bit_stride (); | |
1246 | if (sd == 0) | |
1247 | sd = target_type->length () * 8; | |
1248 | ||
1249 | if (fortran_array_slicing_debug) | |
1250 | { | |
1251 | debug_printf ("|-> Range access\n"); | |
1252 | std::string str = type_to_string (dim_type); | |
1253 | debug_printf ("| |-> Type: %s\n", str.c_str ()); | |
1254 | debug_printf ("| |-> Array:\n"); | |
1255 | debug_printf ("| | |-> Low bound: %s\n", plongest (lb)); | |
1256 | debug_printf ("| | |-> High bound: %s\n", plongest (ub)); | |
1257 | debug_printf ("| | |-> Bit stride: %s\n", plongest (sd)); | |
1258 | debug_printf ("| | |-> Byte stride: %s\n", plongest (sd / 8)); | |
1259 | debug_printf ("| | |-> Type size: %s\n", | |
1260 | pulongest (dim_type->length ())); | |
1261 | debug_printf ("| | '-> Target type size: %s\n", | |
1262 | pulongest (target_type->length ())); | |
1263 | debug_printf ("| |-> Accessing:\n"); | |
1264 | debug_printf ("| | |-> Low bound: %s\n", | |
1265 | plongest (low)); | |
1266 | debug_printf ("| | |-> High bound: %s\n", | |
1267 | plongest (high)); | |
1268 | debug_printf ("| | '-> Element stride: %s\n", | |
1269 | plongest (stride)); | |
1270 | } | |
1271 | ||
1272 | /* Check the user hasn't asked for something invalid. */ | |
1273 | if (high > ub || low < lb) | |
1274 | error (_("array subscript out of bounds")); | |
1275 | ||
1276 | /* Calculate what this dimension of the new slice array will look | |
1277 | like. OFFSET is the byte offset from the start of the | |
1278 | previous (more outer) dimension to the start of this | |
1279 | dimension. E_COUNT is the number of elements in this | |
1280 | dimension. REMAINDER is the number of elements remaining | |
1281 | between the last included element and the upper bound. For | |
1282 | example an access '1:6:2' will include elements 1, 3, 5 and | |
1283 | have a remainder of 1 (element #6). */ | |
1284 | LONGEST lowest = std::min (low, high); | |
1285 | LONGEST offset = (sd / 8) * (lowest - lb); | |
1286 | LONGEST e_count = std::abs (high - low) + 1; | |
1287 | e_count = (e_count + (std::abs (stride) - 1)) / std::abs (stride); | |
1288 | LONGEST new_low = 1; | |
1289 | LONGEST new_high = new_low + e_count - 1; | |
1290 | LONGEST new_stride = (sd * stride) / 8; | |
1291 | LONGEST last_elem = low + ((e_count - 1) * stride); | |
1292 | LONGEST remainder = high - last_elem; | |
1293 | if (low > high) | |
1294 | { | |
1295 | offset += std::abs (remainder) * target_type->length (); | |
1296 | if (stride > 0) | |
1297 | error (_("incorrect stride and boundary combination")); | |
1298 | } | |
1299 | else if (stride < 0) | |
1300 | error (_("incorrect stride and boundary combination")); | |
1301 | ||
1302 | /* Is the data within this dimension contiguous? It is if the | |
1303 | newly computed stride is the same size as a single element of | |
1304 | this dimension. */ | |
1305 | bool is_dim_contiguous = (new_stride == slice_element_size); | |
1306 | is_all_contiguous &= is_dim_contiguous; | |
1307 | ||
1308 | if (fortran_array_slicing_debug) | |
1309 | { | |
1310 | debug_printf ("| '-> Results:\n"); | |
1311 | debug_printf ("| |-> Offset = %s\n", plongest (offset)); | |
1312 | debug_printf ("| |-> Elements = %s\n", plongest (e_count)); | |
1313 | debug_printf ("| |-> Low bound = %s\n", plongest (new_low)); | |
1314 | debug_printf ("| |-> High bound = %s\n", | |
1315 | plongest (new_high)); | |
1316 | debug_printf ("| |-> Byte stride = %s\n", | |
1317 | plongest (new_stride)); | |
1318 | debug_printf ("| |-> Last element = %s\n", | |
1319 | plongest (last_elem)); | |
1320 | debug_printf ("| |-> Remainder = %s\n", | |
1321 | plongest (remainder)); | |
1322 | debug_printf ("| '-> Contiguous = %s\n", | |
1323 | (is_dim_contiguous ? "Yes" : "No")); | |
1324 | } | |
1325 | ||
1326 | /* Figure out how big (in bytes) an element of this dimension of | |
1327 | the new array slice will be. */ | |
1328 | slice_element_size = std::abs (new_stride * e_count); | |
1329 | ||
1330 | slice_dims.emplace_back (new_low, new_high, new_stride, | |
1331 | index_type); | |
1332 | ||
1333 | /* Update the total offset. */ | |
1334 | total_offset += offset; | |
1335 | } | |
1336 | else | |
1337 | { | |
1338 | /* There is a single index for this dimension. */ | |
1339 | LONGEST index | |
1340 | = value_as_long (ops[i]->evaluate_with_coercion (exp, noside)); | |
1341 | ||
1342 | /* Get information about this dimension in the original ARRAY. */ | |
1343 | struct type *target_type = dim_type->target_type (); | |
1344 | struct type *index_type = dim_type->index_type (); | |
1345 | LONGEST lb = f77_get_lowerbound (dim_type); | |
1346 | LONGEST ub = f77_get_upperbound (dim_type); | |
1347 | LONGEST sd = index_type->bit_stride () / 8; | |
1348 | if (sd == 0) | |
1349 | sd = target_type->length (); | |
1350 | ||
1351 | if (fortran_array_slicing_debug) | |
1352 | { | |
1353 | debug_printf ("|-> Index access\n"); | |
1354 | std::string str = type_to_string (dim_type); | |
1355 | debug_printf ("| |-> Type: %s\n", str.c_str ()); | |
1356 | debug_printf ("| |-> Array:\n"); | |
1357 | debug_printf ("| | |-> Low bound: %s\n", plongest (lb)); | |
1358 | debug_printf ("| | |-> High bound: %s\n", plongest (ub)); | |
1359 | debug_printf ("| | |-> Byte stride: %s\n", plongest (sd)); | |
1360 | debug_printf ("| | |-> Type size: %s\n", | |
1361 | pulongest (dim_type->length ())); | |
1362 | debug_printf ("| | '-> Target type size: %s\n", | |
1363 | pulongest (target_type->length ())); | |
1364 | debug_printf ("| '-> Accessing:\n"); | |
1365 | debug_printf ("| '-> Index: %s\n", | |
1366 | plongest (index)); | |
1367 | } | |
1368 | ||
1369 | /* If the array has actual content then check the index is in | |
1370 | bounds. An array without content (an unbound array) doesn't | |
1371 | have a known upper bound, so don't error check in that | |
1372 | situation. */ | |
1373 | if (index < lb | |
1374 | || (dim_type->index_type ()->bounds ()->high.is_available () | |
1375 | && index > ub) | |
1376 | || (array->lval () != lval_memory | |
1377 | && dim_type->index_type ()->bounds ()->high.kind () == PROP_UNDEFINED)) | |
1378 | { | |
1379 | if (type_not_associated (dim_type)) | |
1380 | error (_("no such vector element (vector not associated)")); | |
1381 | else if (type_not_allocated (dim_type)) | |
1382 | error (_("no such vector element (vector not allocated)")); | |
1383 | else | |
1384 | error (_("no such vector element")); | |
1385 | } | |
1386 | ||
1387 | /* Calculate using the type stride, not the target type size. */ | |
1388 | LONGEST offset = sd * (index - lb); | |
1389 | total_offset += offset; | |
1390 | } | |
1391 | } | |
1392 | ||
1393 | /* Build a type that represents the new array slice in the target memory | |
1394 | of the original ARRAY, this type makes use of strides to correctly | |
1395 | find only those elements that are part of the new slice. */ | |
1396 | struct type *array_slice_type = inner_element_type; | |
1397 | for (const auto &d : slice_dims) | |
1398 | { | |
1399 | /* Create the range. */ | |
1400 | dynamic_prop p_low, p_high, p_stride; | |
1401 | ||
1402 | p_low.set_const_val (d.low); | |
1403 | p_high.set_const_val (d.high); | |
1404 | p_stride.set_const_val (d.stride); | |
1405 | ||
1406 | type_allocator alloc (d.index->target_type ()); | |
1407 | struct type *new_range | |
1408 | = create_range_type_with_stride (alloc, | |
1409 | d.index->target_type (), | |
1410 | &p_low, &p_high, 0, &p_stride, | |
1411 | true); | |
1412 | array_slice_type | |
1413 | = create_array_type (alloc, array_slice_type, new_range); | |
1414 | } | |
1415 | ||
1416 | if (fortran_array_slicing_debug) | |
1417 | { | |
1418 | debug_printf ("'-> Final result:\n"); | |
1419 | debug_printf (" |-> Type: %s\n", | |
1420 | type_to_string (array_slice_type).c_str ()); | |
1421 | debug_printf (" |-> Total offset: %s\n", | |
1422 | plongest (total_offset)); | |
1423 | debug_printf (" |-> Base address: %s\n", | |
1424 | core_addr_to_string (array->address ())); | |
1425 | debug_printf (" '-> Contiguous = %s\n", | |
1426 | (is_all_contiguous ? "Yes" : "No")); | |
1427 | } | |
1428 | ||
1429 | /* Should we repack this array slice? */ | |
1430 | if (!is_all_contiguous && (repack_array_slices || is_string_p)) | |
1431 | { | |
1432 | /* Build a type for the repacked slice. */ | |
1433 | struct type *repacked_array_type = inner_element_type; | |
1434 | for (const auto &d : slice_dims) | |
1435 | { | |
1436 | /* Create the range. */ | |
1437 | dynamic_prop p_low, p_high, p_stride; | |
1438 | ||
1439 | p_low.set_const_val (d.low); | |
1440 | p_high.set_const_val (d.high); | |
1441 | p_stride.set_const_val (repacked_array_type->length ()); | |
1442 | ||
1443 | type_allocator alloc (d.index->target_type ()); | |
1444 | struct type *new_range | |
1445 | = create_range_type_with_stride (alloc, | |
1446 | d.index->target_type (), | |
1447 | &p_low, &p_high, 0, &p_stride, | |
1448 | true); | |
1449 | repacked_array_type | |
1450 | = create_array_type (alloc, repacked_array_type, new_range); | |
1451 | } | |
1452 | ||
1453 | /* Now copy the elements from the original ARRAY into the packed | |
1454 | array value DEST. */ | |
1455 | struct value *dest = value::allocate (repacked_array_type); | |
1456 | if (array->lazy () | |
1457 | || (total_offset + array_slice_type->length () | |
1458 | > check_typedef (array->type ())->length ())) | |
1459 | { | |
1460 | fortran_array_walker<fortran_lazy_array_repacker_impl> p | |
1461 | (array_slice_type, array->address () + total_offset, dest); | |
1462 | p.walk (); | |
1463 | } | |
1464 | else | |
1465 | { | |
1466 | fortran_array_walker<fortran_array_repacker_impl> p | |
1467 | (array_slice_type, array->address () + total_offset, | |
1468 | total_offset, array, dest); | |
1469 | p.walk (); | |
1470 | } | |
1471 | array = dest; | |
1472 | } | |
1473 | else | |
1474 | { | |
1475 | if (array->lval () == lval_memory) | |
1476 | { | |
1477 | /* If the value we're taking a slice from is not yet loaded, or | |
1478 | the requested slice is outside the values content range then | |
1479 | just create a new lazy value pointing at the memory where the | |
1480 | contents we're looking for exist. */ | |
1481 | if (array->lazy () | |
1482 | || (total_offset + array_slice_type->length () | |
1483 | > check_typedef (array->type ())->length ())) | |
1484 | array = value_at_lazy (array_slice_type, | |
1485 | array->address () + total_offset); | |
1486 | else | |
1487 | array = value_from_contents_and_address | |
1488 | (array_slice_type, array->contents ().data () + total_offset, | |
1489 | array->address () + total_offset); | |
1490 | } | |
1491 | else if (!array->lazy ()) | |
1492 | array = value_from_component (array, array_slice_type, total_offset); | |
1493 | else | |
1494 | error (_("cannot subscript arrays that are not in memory")); | |
1495 | } | |
1496 | ||
1497 | return array; | |
1498 | } | |
1499 | ||
1500 | value * | |
1501 | fortran_undetermined::evaluate (struct type *expect_type, | |
1502 | struct expression *exp, | |
1503 | enum noside noside) | |
1504 | { | |
1505 | value *callee = std::get<0> (m_storage)->evaluate (nullptr, exp, noside); | |
1506 | if (noside == EVAL_AVOID_SIDE_EFFECTS | |
1507 | && is_dynamic_type (callee->type ())) | |
1508 | callee = std::get<0> (m_storage)->evaluate (nullptr, exp, EVAL_NORMAL); | |
1509 | struct type *type = check_typedef (callee->type ()); | |
1510 | enum type_code code = type->code (); | |
1511 | ||
1512 | if (code == TYPE_CODE_PTR) | |
1513 | { | |
1514 | /* Fortran always passes variable to subroutines as pointer. | |
1515 | So we need to look into its target type to see if it is | |
1516 | array, string or function. If it is, we need to switch | |
1517 | to the target value the original one points to. */ | |
1518 | struct type *target_type = check_typedef (type->target_type ()); | |
1519 | ||
1520 | if (target_type->code () == TYPE_CODE_ARRAY | |
1521 | || target_type->code () == TYPE_CODE_STRING | |
1522 | || target_type->code () == TYPE_CODE_FUNC) | |
1523 | { | |
1524 | callee = value_ind (callee); | |
1525 | type = check_typedef (callee->type ()); | |
1526 | code = type->code (); | |
1527 | } | |
1528 | } | |
1529 | ||
1530 | switch (code) | |
1531 | { | |
1532 | case TYPE_CODE_ARRAY: | |
1533 | case TYPE_CODE_STRING: | |
1534 | return value_subarray (callee, exp, noside); | |
1535 | ||
1536 | case TYPE_CODE_PTR: | |
1537 | case TYPE_CODE_FUNC: | |
1538 | case TYPE_CODE_INTERNAL_FUNCTION: | |
1539 | { | |
1540 | /* It's a function call. Allocate arg vector, including | |
1541 | space for the function to be called in argvec[0] and a | |
1542 | termination NULL. */ | |
1543 | const std::vector<operation_up> &actual (std::get<1> (m_storage)); | |
1544 | std::vector<value *> argvec (actual.size ()); | |
1545 | bool is_internal_func = (code == TYPE_CODE_INTERNAL_FUNCTION); | |
1546 | for (int tem = 0; tem < argvec.size (); tem++) | |
1547 | argvec[tem] = fortran_prepare_argument (exp, actual[tem].get (), | |
1548 | tem, is_internal_func, | |
1549 | callee->type (), | |
1550 | noside); | |
1551 | return evaluate_subexp_do_call (exp, noside, callee, argvec, | |
1552 | nullptr, expect_type); | |
1553 | } | |
1554 | ||
1555 | default: | |
1556 | error (_("Cannot perform substring on this type")); | |
1557 | } | |
1558 | } | |
1559 | ||
1560 | value * | |
1561 | fortran_bound_1arg::evaluate (struct type *expect_type, | |
1562 | struct expression *exp, | |
1563 | enum noside noside) | |
1564 | { | |
1565 | bool lbound_p = std::get<0> (m_storage) == FORTRAN_LBOUND; | |
1566 | value *arg1 = std::get<1> (m_storage)->evaluate (nullptr, exp, noside); | |
1567 | fortran_require_array (arg1->type (), lbound_p); | |
1568 | return fortran_bounds_all_dims (lbound_p, exp->gdbarch, arg1); | |
1569 | } | |
1570 | ||
1571 | value * | |
1572 | fortran_bound_2arg::evaluate (struct type *expect_type, | |
1573 | struct expression *exp, | |
1574 | enum noside noside) | |
1575 | { | |
1576 | bool lbound_p = std::get<0> (m_storage) == FORTRAN_LBOUND; | |
1577 | value *arg1 = std::get<1> (m_storage)->evaluate (nullptr, exp, noside); | |
1578 | fortran_require_array (arg1->type (), lbound_p); | |
1579 | ||
1580 | /* User asked for the bounds of a specific dimension of the array. */ | |
1581 | value *arg2 = std::get<2> (m_storage)->evaluate (nullptr, exp, noside); | |
1582 | type *type_arg2 = check_typedef (arg2->type ()); | |
1583 | if (type_arg2->code () != TYPE_CODE_INT) | |
1584 | { | |
1585 | if (lbound_p) | |
1586 | error (_("LBOUND second argument should be an integer")); | |
1587 | else | |
1588 | error (_("UBOUND second argument should be an integer")); | |
1589 | } | |
1590 | ||
1591 | type *result_type = builtin_f_type (exp->gdbarch)->builtin_integer; | |
1592 | return fortran_bounds_for_dimension (lbound_p, arg1, arg2, result_type); | |
1593 | } | |
1594 | ||
1595 | value * | |
1596 | fortran_bound_3arg::evaluate (type *expect_type, | |
1597 | expression *exp, | |
1598 | noside noside) | |
1599 | { | |
1600 | const bool lbound_p = std::get<0> (m_storage) == FORTRAN_LBOUND; | |
1601 | value *arg1 = std::get<1> (m_storage)->evaluate (nullptr, exp, noside); | |
1602 | fortran_require_array (arg1->type (), lbound_p); | |
1603 | ||
1604 | /* User asked for the bounds of a specific dimension of the array. */ | |
1605 | value *arg2 = std::get<2> (m_storage)->evaluate (nullptr, exp, noside); | |
1606 | type *type_arg2 = check_typedef (arg2->type ()); | |
1607 | if (type_arg2->code () != TYPE_CODE_INT) | |
1608 | { | |
1609 | if (lbound_p) | |
1610 | error (_("LBOUND second argument should be an integer")); | |
1611 | else | |
1612 | error (_("UBOUND second argument should be an integer")); | |
1613 | } | |
1614 | ||
1615 | type *kind_arg = std::get<3> (m_storage); | |
1616 | gdb_assert (kind_arg->code () == TYPE_CODE_INT); | |
1617 | ||
1618 | return fortran_bounds_for_dimension (lbound_p, arg1, arg2, kind_arg); | |
1619 | } | |
1620 | ||
1621 | /* Implement STRUCTOP_STRUCT for Fortran. See operation::evaluate in | |
1622 | expression.h for argument descriptions. */ | |
1623 | ||
1624 | value * | |
1625 | fortran_structop_operation::evaluate (struct type *expect_type, | |
1626 | struct expression *exp, | |
1627 | enum noside noside) | |
1628 | { | |
1629 | value *arg1 = std::get<0> (m_storage)->evaluate (nullptr, exp, noside); | |
1630 | const char *str = std::get<1> (m_storage).c_str (); | |
1631 | if (noside == EVAL_AVOID_SIDE_EFFECTS) | |
1632 | { | |
1633 | struct type *type = lookup_struct_elt_type (arg1->type (), str, 1); | |
1634 | ||
1635 | if (type != nullptr && is_dynamic_type (type)) | |
1636 | arg1 = std::get<0> (m_storage)->evaluate (nullptr, exp, EVAL_NORMAL); | |
1637 | } | |
1638 | ||
1639 | value *elt = value_struct_elt (&arg1, {}, str, NULL, "structure"); | |
1640 | ||
1641 | if (noside == EVAL_AVOID_SIDE_EFFECTS) | |
1642 | { | |
1643 | struct type *elt_type = elt->type (); | |
1644 | if (is_dynamic_type (elt_type)) | |
1645 | { | |
1646 | const gdb_byte *valaddr = elt->contents_for_printing ().data (); | |
1647 | CORE_ADDR address = elt->address (); | |
1648 | gdb::array_view<const gdb_byte> view | |
1649 | = gdb::make_array_view (valaddr, elt_type->length ()); | |
1650 | elt_type = resolve_dynamic_type (elt_type, view, address); | |
1651 | } | |
1652 | elt = value::zero (elt_type, elt->lval ()); | |
1653 | } | |
1654 | ||
1655 | return elt; | |
1656 | } | |
1657 | ||
1658 | } /* namespace expr */ | |
1659 | ||
1660 | /* See language.h. */ | |
1661 | ||
1662 | void | |
1663 | f_language::print_array_index (struct type *index_type, LONGEST index, | |
1664 | struct ui_file *stream, | |
1665 | const value_print_options *options) const | |
1666 | { | |
1667 | struct value *index_value = value_from_longest (index_type, index); | |
1668 | ||
1669 | gdb_printf (stream, "("); | |
1670 | value_print (index_value, stream, options); | |
1671 | gdb_printf (stream, ") = "); | |
1672 | } | |
1673 | ||
1674 | /* See language.h. */ | |
1675 | ||
1676 | void | |
1677 | f_language::language_arch_info (struct gdbarch *gdbarch, | |
1678 | struct language_arch_info *lai) const | |
1679 | { | |
1680 | const struct builtin_f_type *builtin = builtin_f_type (gdbarch); | |
1681 | ||
1682 | /* Helper function to allow shorter lines below. */ | |
1683 | auto add = [&] (struct type * t) | |
1684 | { | |
1685 | lai->add_primitive_type (t); | |
1686 | }; | |
1687 | ||
1688 | add (builtin->builtin_character); | |
1689 | add (builtin->builtin_logical); | |
1690 | add (builtin->builtin_logical_s1); | |
1691 | add (builtin->builtin_logical_s2); | |
1692 | add (builtin->builtin_logical_s8); | |
1693 | add (builtin->builtin_real); | |
1694 | add (builtin->builtin_real_s8); | |
1695 | add (builtin->builtin_real_s16); | |
1696 | add (builtin->builtin_complex); | |
1697 | add (builtin->builtin_complex_s8); | |
1698 | add (builtin->builtin_void); | |
1699 | ||
1700 | lai->set_string_char_type (builtin->builtin_character); | |
1701 | lai->set_bool_type (builtin->builtin_logical, "logical"); | |
1702 | } | |
1703 | ||
1704 | /* See language.h. */ | |
1705 | ||
1706 | unsigned int | |
1707 | f_language::search_name_hash (const char *name) const | |
1708 | { | |
1709 | return cp_search_name_hash (name); | |
1710 | } | |
1711 | ||
1712 | /* See language.h. */ | |
1713 | ||
1714 | struct block_symbol | |
1715 | f_language::lookup_symbol_local (const char *scope, | |
1716 | const char *name, | |
1717 | const struct block *block, | |
1718 | const domain_search_flags domain) const | |
1719 | { | |
1720 | return cp_lookup_symbol_imports (scope, name, block, domain); | |
1721 | } | |
1722 | ||
1723 | /* See language.h. */ | |
1724 | ||
1725 | struct block_symbol | |
1726 | f_language::lookup_symbol_nonlocal (const char *name, | |
1727 | const struct block *block, | |
1728 | const domain_search_flags domain) const | |
1729 | { | |
1730 | return cp_lookup_symbol_nonlocal (this, name, block, domain); | |
1731 | } | |
1732 | ||
1733 | /* See language.h. */ | |
1734 | ||
1735 | symbol_name_matcher_ftype * | |
1736 | f_language::get_symbol_name_matcher_inner | |
1737 | (const lookup_name_info &lookup_name) const | |
1738 | { | |
1739 | return cp_get_symbol_name_matcher (lookup_name); | |
1740 | } | |
1741 | ||
1742 | /* Single instance of the Fortran language class. */ | |
1743 | ||
1744 | static f_language f_language_defn; | |
1745 | ||
1746 | static struct builtin_f_type * | |
1747 | build_fortran_types (struct gdbarch *gdbarch) | |
1748 | { | |
1749 | struct builtin_f_type *builtin_f_type = new struct builtin_f_type; | |
1750 | ||
1751 | builtin_f_type->builtin_void = builtin_type (gdbarch)->builtin_void; | |
1752 | ||
1753 | type_allocator alloc (gdbarch); | |
1754 | ||
1755 | builtin_f_type->builtin_character | |
1756 | = alloc.new_type (TYPE_CODE_CHAR, TARGET_CHAR_BIT, "character"); | |
1757 | ||
1758 | builtin_f_type->builtin_logical_s1 | |
1759 | = init_boolean_type (alloc, TARGET_CHAR_BIT, 1, "logical*1"); | |
1760 | ||
1761 | builtin_f_type->builtin_logical_s2 | |
1762 | = init_boolean_type (alloc, gdbarch_short_bit (gdbarch), 1, "logical*2"); | |
1763 | ||
1764 | builtin_f_type->builtin_logical | |
1765 | = init_boolean_type (alloc, gdbarch_int_bit (gdbarch), 1, "logical*4"); | |
1766 | ||
1767 | builtin_f_type->builtin_logical_s8 | |
1768 | = init_boolean_type (alloc, gdbarch_long_long_bit (gdbarch), 1, | |
1769 | "logical*8"); | |
1770 | ||
1771 | builtin_f_type->builtin_integer_s1 | |
1772 | = init_integer_type (alloc, TARGET_CHAR_BIT, 0, "integer*1"); | |
1773 | ||
1774 | builtin_f_type->builtin_integer_s2 | |
1775 | = init_integer_type (alloc, gdbarch_short_bit (gdbarch), 0, "integer*2"); | |
1776 | ||
1777 | builtin_f_type->builtin_integer | |
1778 | = init_integer_type (alloc, gdbarch_int_bit (gdbarch), 0, "integer*4"); | |
1779 | ||
1780 | builtin_f_type->builtin_integer_s8 | |
1781 | = init_integer_type (alloc, gdbarch_long_long_bit (gdbarch), 0, | |
1782 | "integer*8"); | |
1783 | ||
1784 | builtin_f_type->builtin_real | |
1785 | = init_float_type (alloc, gdbarch_float_bit (gdbarch), | |
1786 | "real*4", gdbarch_float_format (gdbarch)); | |
1787 | ||
1788 | builtin_f_type->builtin_real_s8 | |
1789 | = init_float_type (alloc, gdbarch_double_bit (gdbarch), | |
1790 | "real*8", gdbarch_double_format (gdbarch)); | |
1791 | ||
1792 | auto fmt = gdbarch_floatformat_for_type (gdbarch, "real(kind=16)", 128); | |
1793 | if (fmt != nullptr) | |
1794 | builtin_f_type->builtin_real_s16 | |
1795 | = init_float_type (alloc, 128, "real*16", fmt); | |
1796 | else if (gdbarch_long_double_bit (gdbarch) == 128) | |
1797 | builtin_f_type->builtin_real_s16 | |
1798 | = init_float_type (alloc, gdbarch_long_double_bit (gdbarch), | |
1799 | "real*16", gdbarch_long_double_format (gdbarch)); | |
1800 | else | |
1801 | builtin_f_type->builtin_real_s16 | |
1802 | = alloc.new_type (TYPE_CODE_ERROR, 128, "real*16"); | |
1803 | ||
1804 | builtin_f_type->builtin_complex | |
1805 | = init_complex_type ("complex*4", builtin_f_type->builtin_real); | |
1806 | ||
1807 | builtin_f_type->builtin_complex_s8 | |
1808 | = init_complex_type ("complex*8", builtin_f_type->builtin_real_s8); | |
1809 | ||
1810 | if (builtin_f_type->builtin_real_s16->code () == TYPE_CODE_ERROR) | |
1811 | builtin_f_type->builtin_complex_s16 | |
1812 | = alloc.new_type (TYPE_CODE_ERROR, 256, "complex*16"); | |
1813 | else | |
1814 | builtin_f_type->builtin_complex_s16 | |
1815 | = init_complex_type ("complex*16", builtin_f_type->builtin_real_s16); | |
1816 | ||
1817 | return builtin_f_type; | |
1818 | } | |
1819 | ||
1820 | static const registry<gdbarch>::key<struct builtin_f_type> f_type_data; | |
1821 | ||
1822 | const struct builtin_f_type * | |
1823 | builtin_f_type (struct gdbarch *gdbarch) | |
1824 | { | |
1825 | struct builtin_f_type *result = f_type_data.get (gdbarch); | |
1826 | if (result == nullptr) | |
1827 | { | |
1828 | result = build_fortran_types (gdbarch); | |
1829 | f_type_data.set (gdbarch, result); | |
1830 | } | |
1831 | ||
1832 | return result; | |
1833 | } | |
1834 | ||
1835 | /* Command-list for the "set/show fortran" prefix command. */ | |
1836 | static struct cmd_list_element *set_fortran_list; | |
1837 | static struct cmd_list_element *show_fortran_list; | |
1838 | ||
1839 | INIT_GDB_FILE (f_language) | |
1840 | { | |
1841 | add_setshow_prefix_cmd | |
1842 | ("fortran", no_class, | |
1843 | _("Prefix command for changing Fortran-specific settings."), | |
1844 | _("Generic command for showing Fortran-specific settings."), | |
1845 | &set_fortran_list, &show_fortran_list, | |
1846 | &setlist, &showlist); | |
1847 | ||
1848 | add_setshow_boolean_cmd ("repack-array-slices", class_vars, | |
1849 | &repack_array_slices, _("\ | |
1850 | Enable or disable repacking of non-contiguous array slices."), _("\ | |
1851 | Show whether non-contiguous array slices are repacked."), _("\ | |
1852 | When the user requests a slice of a Fortran array then we can either return\n\ | |
1853 | a descriptor that describes the array in place (using the original array data\n\ | |
1854 | in its existing location) or the original data can be repacked (copied) to a\n\ | |
1855 | new location.\n\ | |
1856 | \n\ | |
1857 | When the content of the array slice is contiguous within the original array\n\ | |
1858 | then the result will never be repacked, but when the data for the new array\n\ | |
1859 | is non-contiguous within the original array repacking will only be performed\n\ | |
1860 | when this setting is on."), | |
1861 | NULL, | |
1862 | show_repack_array_slices, | |
1863 | &set_fortran_list, &show_fortran_list); | |
1864 | ||
1865 | /* Debug Fortran's array slicing logic. */ | |
1866 | add_setshow_boolean_cmd ("fortran-array-slicing", class_maintenance, | |
1867 | &fortran_array_slicing_debug, _("\ | |
1868 | Set debugging of Fortran array slicing."), _("\ | |
1869 | Show debugging of Fortran array slicing."), _("\ | |
1870 | When on, debugging of Fortran array slicing is enabled."), | |
1871 | NULL, | |
1872 | show_fortran_array_slicing_debug, | |
1873 | &setdebuglist, &showdebuglist); | |
1874 | } | |
1875 | ||
1876 | /* Ensures that function argument VALUE is in the appropriate form to | |
1877 | pass to a Fortran function. Returns a possibly new value that should | |
1878 | be used instead of VALUE. | |
1879 | ||
1880 | When IS_ARTIFICIAL is true this indicates an artificial argument, | |
1881 | e.g. hidden string lengths which the GNU Fortran argument passing | |
1882 | convention specifies as being passed by value. | |
1883 | ||
1884 | When IS_ARTIFICIAL is false, the argument is passed by pointer. If the | |
1885 | value is already in target memory then return a value that is a pointer | |
1886 | to VALUE. If VALUE is not in memory (e.g. an integer literal), allocate | |
1887 | space in the target, copy VALUE in, and return a pointer to the in | |
1888 | memory copy. */ | |
1889 | ||
1890 | static struct value * | |
1891 | fortran_argument_convert (struct value *value, bool is_artificial) | |
1892 | { | |
1893 | if (!is_artificial) | |
1894 | { | |
1895 | /* If the value is not in the inferior e.g. registers values, | |
1896 | convenience variables and user input. */ | |
1897 | if (value->lval () != lval_memory) | |
1898 | { | |
1899 | struct type *type = value->type (); | |
1900 | const int length = type->length (); | |
1901 | const CORE_ADDR addr | |
1902 | = value_as_long (value_allocate_space_in_inferior (length)); | |
1903 | write_memory (addr, value->contents ().data (), length); | |
1904 | struct value *val = value_from_contents_and_address | |
1905 | (type, value->contents ().data (), addr); | |
1906 | return value_addr (val); | |
1907 | } | |
1908 | else | |
1909 | return value_addr (value); /* Program variables, e.g. arrays. */ | |
1910 | } | |
1911 | return value; | |
1912 | } | |
1913 | ||
1914 | /* Prepare (and return) an argument value ready for an inferior function | |
1915 | call to a Fortran function. EXP and POS are the expressions describing | |
1916 | the argument to prepare. ARG_NUM is the argument number being | |
1917 | prepared, with 0 being the first argument and so on. FUNC_TYPE is the | |
1918 | type of the function being called. | |
1919 | ||
1920 | IS_INTERNAL_CALL_P is true if this is a call to a function of type | |
1921 | TYPE_CODE_INTERNAL_FUNCTION, otherwise this parameter is false. | |
1922 | ||
1923 | NOSIDE has its usual meaning for expression parsing (see eval.c). | |
1924 | ||
1925 | Arguments in Fortran are normally passed by address, we coerce the | |
1926 | arguments here rather than in value_arg_coerce as otherwise the call to | |
1927 | malloc (to place the non-lvalue parameters in target memory) is hit by | |
1928 | this Fortran specific logic. This results in malloc being called with a | |
1929 | pointer to an integer followed by an attempt to malloc the arguments to | |
1930 | malloc in target memory. Infinite recursion ensues. */ | |
1931 | ||
1932 | static value * | |
1933 | fortran_prepare_argument (struct expression *exp, | |
1934 | expr::operation *subexp, | |
1935 | int arg_num, bool is_internal_call_p, | |
1936 | struct type *func_type, enum noside noside) | |
1937 | { | |
1938 | if (is_internal_call_p) | |
1939 | return subexp->evaluate_with_coercion (exp, noside); | |
1940 | ||
1941 | bool is_artificial = ((arg_num >= func_type->num_fields ()) | |
1942 | ? true | |
1943 | : func_type->field (arg_num).is_artificial ()); | |
1944 | ||
1945 | /* If this is an artificial argument, then either, this is an argument | |
1946 | beyond the end of the known arguments, or possibly, there are no known | |
1947 | arguments (maybe missing debug info). | |
1948 | ||
1949 | For these artificial arguments, if the user has prefixed it with '&' | |
1950 | (for address-of), then lets always allow this to succeed, even if the | |
1951 | argument is not actually in inferior memory. This will allow the user | |
1952 | to pass arguments to a Fortran function even when there's no debug | |
1953 | information. | |
1954 | ||
1955 | As we already pass the address of non-artificial arguments, all we | |
1956 | need to do if skip the UNOP_ADDR operator in the expression and mark | |
1957 | the argument as non-artificial. */ | |
1958 | if (is_artificial) | |
1959 | { | |
1960 | expr::unop_addr_operation *addrop | |
1961 | = dynamic_cast<expr::unop_addr_operation *> (subexp); | |
1962 | if (addrop != nullptr) | |
1963 | { | |
1964 | subexp = addrop->get_expression ().get (); | |
1965 | is_artificial = false; | |
1966 | } | |
1967 | } | |
1968 | ||
1969 | struct value *arg_val = subexp->evaluate_with_coercion (exp, noside); | |
1970 | return fortran_argument_convert (arg_val, is_artificial); | |
1971 | } | |
1972 | ||
1973 | /* See f-lang.h. */ | |
1974 | ||
1975 | struct type * | |
1976 | fortran_preserve_arg_pointer (struct value *arg, struct type *type) | |
1977 | { | |
1978 | if (arg->type ()->code () == TYPE_CODE_PTR) | |
1979 | return arg->type (); | |
1980 | return type; | |
1981 | } | |
1982 | ||
1983 | /* See f-lang.h. */ | |
1984 | ||
1985 | CORE_ADDR | |
1986 | fortran_adjust_dynamic_array_base_address_hack (struct type *type, | |
1987 | CORE_ADDR address) | |
1988 | { | |
1989 | gdb_assert (type->code () == TYPE_CODE_ARRAY); | |
1990 | ||
1991 | /* We can't adjust the base address for arrays that have no content. */ | |
1992 | if (type_not_allocated (type) || type_not_associated (type)) | |
1993 | return address; | |
1994 | ||
1995 | int ndimensions = calc_f77_array_dims (type); | |
1996 | LONGEST total_offset = 0; | |
1997 | ||
1998 | /* Walk through each of the dimensions of this array type and figure out | |
1999 | if any of the dimensions are "backwards", that is the base address | |
2000 | for this dimension points to the element at the highest memory | |
2001 | address and the stride is negative. */ | |
2002 | struct type *tmp_type = type; | |
2003 | for (int i = 0 ; i < ndimensions; ++i) | |
2004 | { | |
2005 | /* Grab the range for this dimension and extract the lower and upper | |
2006 | bounds. */ | |
2007 | tmp_type = check_typedef (tmp_type); | |
2008 | struct type *range_type = tmp_type->index_type (); | |
2009 | LONGEST lowerbound, upperbound, stride; | |
2010 | if (!get_discrete_bounds (range_type, &lowerbound, &upperbound)) | |
2011 | error ("failed to get range bounds"); | |
2012 | ||
2013 | /* Figure out the stride for this dimension. */ | |
2014 | struct type *elt_type = check_typedef (tmp_type->target_type ()); | |
2015 | stride = tmp_type->index_type ()->bounds ()->bit_stride (); | |
2016 | if (stride == 0) | |
2017 | stride = type_length_units (elt_type); | |
2018 | else | |
2019 | { | |
2020 | int unit_size | |
2021 | = gdbarch_addressable_memory_unit_size (elt_type->arch ()); | |
2022 | stride /= (unit_size * 8); | |
2023 | } | |
2024 | ||
2025 | /* If this dimension is "backward" then figure out the offset | |
2026 | adjustment required to point to the element at the lowest memory | |
2027 | address, and add this to the total offset. */ | |
2028 | LONGEST offset = 0; | |
2029 | if (stride < 0 && lowerbound < upperbound) | |
2030 | offset = (upperbound - lowerbound) * stride; | |
2031 | total_offset += offset; | |
2032 | tmp_type = tmp_type->target_type (); | |
2033 | } | |
2034 | ||
2035 | /* Adjust the address of this object and return it. */ | |
2036 | address += total_offset; | |
2037 | return address; | |
2038 | } |