1 /* DWARF index writing support for GDB.
3 Copyright (C) 1994-2019 Free Software Foundation, Inc.
5 This file is part of GDB.
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>. */
23 #include "cli/cli-decode.h"
24 #include "gdbsupport/byte-vector.h"
25 #include "gdbsupport/filestuff.h"
26 #include "gdbsupport/gdb_unlinker.h"
27 #include "gdbsupport/pathstuff.h"
28 #include "gdbsupport/scoped_fd.h"
29 #include "complaints.h"
30 #include "dwarf-index-common.h"
32 #include "dwarf2read.h"
33 #include "gdb/gdb-index.h"
42 #include <unordered_map>
43 #include <unordered_set>
45 /* Ensure only legit values are used. */
46 #define DW2_GDB_INDEX_SYMBOL_STATIC_SET_VALUE(cu_index, value) \
48 gdb_assert ((unsigned int) (value) <= 1); \
49 GDB_INDEX_SYMBOL_STATIC_SET_VALUE((cu_index), (value)); \
52 /* Ensure only legit values are used. */
53 #define DW2_GDB_INDEX_SYMBOL_KIND_SET_VALUE(cu_index, value) \
55 gdb_assert ((value) >= GDB_INDEX_SYMBOL_KIND_TYPE \
56 && (value) <= GDB_INDEX_SYMBOL_KIND_OTHER); \
57 GDB_INDEX_SYMBOL_KIND_SET_VALUE((cu_index), (value)); \
60 /* Ensure we don't use more than the alloted nuber of bits for the CU. */
61 #define DW2_GDB_INDEX_CU_SET_VALUE(cu_index, value) \
63 gdb_assert (((value) & ~GDB_INDEX_CU_MASK) == 0); \
64 GDB_INDEX_CU_SET_VALUE((cu_index), (value)); \
67 /* The "save gdb-index" command. */
69 /* Write SIZE bytes from the buffer pointed to by DATA to FILE, with
73 file_write (FILE *file
, const void *data
, size_t size
)
75 if (fwrite (data
, 1, size
, file
) != size
)
76 error (_("couldn't data write to file"));
79 /* Write the contents of VEC to FILE, with error checking. */
81 template<typename Elem
, typename Alloc
>
83 file_write (FILE *file
, const std::vector
<Elem
, Alloc
> &vec
)
86 file_write (file
, vec
.data (), vec
.size () * sizeof (vec
[0]));
89 /* In-memory buffer to prepare data to be written later to a file. */
93 /* Copy DATA to the end of the buffer. */
95 void append_data (const T
&data
)
97 std::copy (reinterpret_cast<const gdb_byte
*> (&data
),
98 reinterpret_cast<const gdb_byte
*> (&data
+ 1),
99 grow (sizeof (data
)));
102 /* Copy CSTR (a zero-terminated string) to the end of buffer. The
103 terminating zero is appended too. */
104 void append_cstr0 (const char *cstr
)
106 const size_t size
= strlen (cstr
) + 1;
107 std::copy (cstr
, cstr
+ size
, grow (size
));
110 /* Store INPUT as ULEB128 to the end of buffer. */
111 void append_unsigned_leb128 (ULONGEST input
)
115 gdb_byte output
= input
& 0x7f;
119 append_data (output
);
125 /* Accept a host-format integer in VAL and append it to the buffer
126 as a target-format integer which is LEN bytes long. */
127 void append_uint (size_t len
, bfd_endian byte_order
, ULONGEST val
)
129 ::store_unsigned_integer (grow (len
), len
, byte_order
, val
);
132 /* Return the size of the buffer. */
135 return m_vec
.size ();
138 /* Return true iff the buffer is empty. */
141 return m_vec
.empty ();
144 /* Write the buffer to FILE. */
145 void file_write (FILE *file
) const
147 ::file_write (file
, m_vec
);
151 /* Grow SIZE bytes at the end of the buffer. Returns a pointer to
152 the start of the new block. */
153 gdb_byte
*grow (size_t size
)
155 m_vec
.resize (m_vec
.size () + size
);
156 return &*(m_vec
.end () - size
);
159 gdb::byte_vector m_vec
;
162 /* An entry in the symbol table. */
163 struct symtab_index_entry
165 /* The name of the symbol. */
167 /* The offset of the name in the constant pool. */
168 offset_type index_offset
;
169 /* A sorted vector of the indices of all the CUs that hold an object
171 std::vector
<offset_type
> cu_indices
;
174 /* The symbol table. This is a power-of-2-sized hash table. */
182 offset_type n_elements
= 0;
183 std::vector
<symtab_index_entry
> data
;
186 /* Find a slot in SYMTAB for the symbol NAME. Returns a reference to
189 Function is used only during write_hash_table so no index format backward
190 compatibility is needed. */
192 static symtab_index_entry
&
193 find_slot (struct mapped_symtab
*symtab
, const char *name
)
195 offset_type index
, step
, hash
= mapped_index_string_hash (INT_MAX
, name
);
197 index
= hash
& (symtab
->data
.size () - 1);
198 step
= ((hash
* 17) & (symtab
->data
.size () - 1)) | 1;
202 if (symtab
->data
[index
].name
== NULL
203 || strcmp (name
, symtab
->data
[index
].name
) == 0)
204 return symtab
->data
[index
];
205 index
= (index
+ step
) & (symtab
->data
.size () - 1);
209 /* Expand SYMTAB's hash table. */
212 hash_expand (struct mapped_symtab
*symtab
)
214 auto old_entries
= std::move (symtab
->data
);
216 symtab
->data
.clear ();
217 symtab
->data
.resize (old_entries
.size () * 2);
219 for (auto &it
: old_entries
)
222 auto &ref
= find_slot (symtab
, it
.name
);
223 ref
= std::move (it
);
227 /* Add an entry to SYMTAB. NAME is the name of the symbol.
228 CU_INDEX is the index of the CU in which the symbol appears.
229 IS_STATIC is one if the symbol is static, otherwise zero (global). */
232 add_index_entry (struct mapped_symtab
*symtab
, const char *name
,
233 int is_static
, gdb_index_symbol_kind kind
,
234 offset_type cu_index
)
236 offset_type cu_index_and_attrs
;
238 ++symtab
->n_elements
;
239 if (4 * symtab
->n_elements
/ 3 >= symtab
->data
.size ())
240 hash_expand (symtab
);
242 symtab_index_entry
&slot
= find_slot (symtab
, name
);
243 if (slot
.name
== NULL
)
246 /* index_offset is set later. */
249 cu_index_and_attrs
= 0;
250 DW2_GDB_INDEX_CU_SET_VALUE (cu_index_and_attrs
, cu_index
);
251 DW2_GDB_INDEX_SYMBOL_STATIC_SET_VALUE (cu_index_and_attrs
, is_static
);
252 DW2_GDB_INDEX_SYMBOL_KIND_SET_VALUE (cu_index_and_attrs
, kind
);
254 /* We don't want to record an index value twice as we want to avoid the
256 We process all global symbols and then all static symbols
257 (which would allow us to avoid the duplication by only having to check
258 the last entry pushed), but a symbol could have multiple kinds in one CU.
259 To keep things simple we don't worry about the duplication here and
260 sort and uniqufy the list after we've processed all symbols. */
261 slot
.cu_indices
.push_back (cu_index_and_attrs
);
264 /* Sort and remove duplicates of all symbols' cu_indices lists. */
267 uniquify_cu_indices (struct mapped_symtab
*symtab
)
269 for (auto &entry
: symtab
->data
)
271 if (entry
.name
!= NULL
&& !entry
.cu_indices
.empty ())
273 auto &cu_indices
= entry
.cu_indices
;
274 std::sort (cu_indices
.begin (), cu_indices
.end ());
275 auto from
= std::unique (cu_indices
.begin (), cu_indices
.end ());
276 cu_indices
.erase (from
, cu_indices
.end ());
281 /* A form of 'const char *' suitable for container keys. Only the
282 pointer is stored. The strings themselves are compared, not the
287 c_str_view (const char *cstr
)
291 bool operator== (const c_str_view
&other
) const
293 return strcmp (m_cstr
, other
.m_cstr
) == 0;
296 /* Return the underlying C string. Note, the returned string is
297 only a reference with lifetime of this object. */
298 const char *c_str () const
304 friend class c_str_view_hasher
;
305 const char *const m_cstr
;
308 /* A std::unordered_map::hasher for c_str_view that uses the right
309 hash function for strings in a mapped index. */
310 class c_str_view_hasher
313 size_t operator () (const c_str_view
&x
) const
315 return mapped_index_string_hash (INT_MAX
, x
.m_cstr
);
319 /* A std::unordered_map::hasher for std::vector<>. */
324 size_t operator () (const std::vector
<T
> &key
) const
326 return iterative_hash (key
.data (),
327 sizeof (key
.front ()) * key
.size (), 0);
331 /* Write the mapped hash table SYMTAB to the data buffer OUTPUT, with
332 constant pool entries going into the data buffer CPOOL. */
335 write_hash_table (mapped_symtab
*symtab
, data_buf
&output
, data_buf
&cpool
)
338 /* Elements are sorted vectors of the indices of all the CUs that
339 hold an object of this name. */
340 std::unordered_map
<std::vector
<offset_type
>, offset_type
,
341 vector_hasher
<offset_type
>>
344 /* We add all the index vectors to the constant pool first, to
345 ensure alignment is ok. */
346 for (symtab_index_entry
&entry
: symtab
->data
)
348 if (entry
.name
== NULL
)
350 gdb_assert (entry
.index_offset
== 0);
352 /* Finding before inserting is faster than always trying to
353 insert, because inserting always allocates a node, does the
354 lookup, and then destroys the new node if another node
355 already had the same key. C++17 try_emplace will avoid
358 = symbol_hash_table
.find (entry
.cu_indices
);
359 if (found
!= symbol_hash_table
.end ())
361 entry
.index_offset
= found
->second
;
365 symbol_hash_table
.emplace (entry
.cu_indices
, cpool
.size ());
366 entry
.index_offset
= cpool
.size ();
367 cpool
.append_data (MAYBE_SWAP (entry
.cu_indices
.size ()));
368 for (const auto index
: entry
.cu_indices
)
369 cpool
.append_data (MAYBE_SWAP (index
));
373 /* Now write out the hash table. */
374 std::unordered_map
<c_str_view
, offset_type
, c_str_view_hasher
> str_table
;
375 for (const auto &entry
: symtab
->data
)
377 offset_type str_off
, vec_off
;
379 if (entry
.name
!= NULL
)
381 const auto insertpair
= str_table
.emplace (entry
.name
, cpool
.size ());
382 if (insertpair
.second
)
383 cpool
.append_cstr0 (entry
.name
);
384 str_off
= insertpair
.first
->second
;
385 vec_off
= entry
.index_offset
;
389 /* While 0 is a valid constant pool index, it is not valid
390 to have 0 for both offsets. */
395 output
.append_data (MAYBE_SWAP (str_off
));
396 output
.append_data (MAYBE_SWAP (vec_off
));
400 typedef std::unordered_map
<partial_symtab
*, unsigned int> psym_index_map
;
402 /* Helper struct for building the address table. */
403 struct addrmap_index_data
405 addrmap_index_data (data_buf
&addr_vec_
, psym_index_map
&cu_index_htab_
)
406 : addr_vec (addr_vec_
), cu_index_htab (cu_index_htab_
)
409 struct objfile
*objfile
;
411 psym_index_map
&cu_index_htab
;
413 /* Non-zero if the previous_* fields are valid.
414 We can't write an entry until we see the next entry (since it is only then
415 that we know the end of the entry). */
417 /* Index of the CU in the table of all CUs in the index file. */
418 unsigned int previous_cu_index
;
419 /* Start address of the CU. */
420 CORE_ADDR previous_cu_start
;
423 /* Write an address entry to ADDR_VEC. */
426 add_address_entry (struct objfile
*objfile
, data_buf
&addr_vec
,
427 CORE_ADDR start
, CORE_ADDR end
, unsigned int cu_index
)
429 addr_vec
.append_uint (8, BFD_ENDIAN_LITTLE
, start
);
430 addr_vec
.append_uint (8, BFD_ENDIAN_LITTLE
, end
);
431 addr_vec
.append_data (MAYBE_SWAP (cu_index
));
434 /* Worker function for traversing an addrmap to build the address table. */
437 add_address_entry_worker (void *datap
, CORE_ADDR start_addr
, void *obj
)
439 struct addrmap_index_data
*data
= (struct addrmap_index_data
*) datap
;
440 struct partial_symtab
*pst
= (struct partial_symtab
*) obj
;
442 if (data
->previous_valid
)
443 add_address_entry (data
->objfile
, data
->addr_vec
,
444 data
->previous_cu_start
, start_addr
,
445 data
->previous_cu_index
);
447 data
->previous_cu_start
= start_addr
;
450 const auto it
= data
->cu_index_htab
.find (pst
);
451 gdb_assert (it
!= data
->cu_index_htab
.cend ());
452 data
->previous_cu_index
= it
->second
;
453 data
->previous_valid
= 1;
456 data
->previous_valid
= 0;
461 /* Write OBJFILE's address map to ADDR_VEC.
462 CU_INDEX_HTAB is used to map addrmap entries to their CU indices
463 in the index file. */
466 write_address_map (struct objfile
*objfile
, data_buf
&addr_vec
,
467 psym_index_map
&cu_index_htab
)
469 struct addrmap_index_data
addrmap_index_data (addr_vec
, cu_index_htab
);
471 /* When writing the address table, we have to cope with the fact that
472 the addrmap iterator only provides the start of a region; we have to
473 wait until the next invocation to get the start of the next region. */
475 addrmap_index_data
.objfile
= objfile
;
476 addrmap_index_data
.previous_valid
= 0;
478 addrmap_foreach (objfile
->partial_symtabs
->psymtabs_addrmap
,
479 add_address_entry_worker
, &addrmap_index_data
);
481 /* It's highly unlikely the last entry (end address = 0xff...ff)
482 is valid, but we should still handle it.
483 The end address is recorded as the start of the next region, but that
484 doesn't work here. To cope we pass 0xff...ff, this is a rare situation
486 if (addrmap_index_data
.previous_valid
)
487 add_address_entry (objfile
, addr_vec
,
488 addrmap_index_data
.previous_cu_start
, (CORE_ADDR
) -1,
489 addrmap_index_data
.previous_cu_index
);
492 /* Return the symbol kind of PSYM. */
494 static gdb_index_symbol_kind
495 symbol_kind (struct partial_symbol
*psym
)
497 domain_enum domain
= psym
->domain
;
498 enum address_class aclass
= psym
->aclass
;
506 return GDB_INDEX_SYMBOL_KIND_FUNCTION
;
508 return GDB_INDEX_SYMBOL_KIND_TYPE
;
510 case LOC_CONST_BYTES
:
511 case LOC_OPTIMIZED_OUT
:
513 return GDB_INDEX_SYMBOL_KIND_VARIABLE
;
515 /* Note: It's currently impossible to recognize psyms as enum values
516 short of reading the type info. For now punt. */
517 return GDB_INDEX_SYMBOL_KIND_VARIABLE
;
519 /* There are other LOC_FOO values that one might want to classify
520 as variables, but dwarf2read.c doesn't currently use them. */
521 return GDB_INDEX_SYMBOL_KIND_OTHER
;
524 return GDB_INDEX_SYMBOL_KIND_TYPE
;
526 return GDB_INDEX_SYMBOL_KIND_OTHER
;
530 /* Add a list of partial symbols to SYMTAB. */
533 write_psymbols (struct mapped_symtab
*symtab
,
534 std::unordered_set
<partial_symbol
*> &psyms_seen
,
535 struct partial_symbol
**psymp
,
537 offset_type cu_index
,
540 for (; count
-- > 0; ++psymp
)
542 struct partial_symbol
*psym
= *psymp
;
544 if (psym
->ginfo
.language
== language_ada
)
545 error (_("Ada is not currently supported by the index; "
546 "use the DWARF 5 index instead"));
548 /* Only add a given psymbol once. */
549 if (psyms_seen
.insert (psym
).second
)
551 gdb_index_symbol_kind kind
= symbol_kind (psym
);
553 add_index_entry (symtab
, symbol_search_name (&psym
->ginfo
),
554 is_static
, kind
, cu_index
);
559 /* A helper struct used when iterating over debug_types. */
560 struct signatured_type_index_data
562 signatured_type_index_data (data_buf
&types_list_
,
563 std::unordered_set
<partial_symbol
*> &psyms_seen_
)
564 : types_list (types_list_
), psyms_seen (psyms_seen_
)
567 struct objfile
*objfile
;
568 struct mapped_symtab
*symtab
;
569 data_buf
&types_list
;
570 std::unordered_set
<partial_symbol
*> &psyms_seen
;
574 /* A helper function that writes a single signatured_type to an
578 write_one_signatured_type (void **slot
, void *d
)
580 struct signatured_type_index_data
*info
581 = (struct signatured_type_index_data
*) d
;
582 struct signatured_type
*entry
= (struct signatured_type
*) *slot
;
583 struct partial_symtab
*psymtab
= entry
->per_cu
.v
.psymtab
;
585 write_psymbols (info
->symtab
,
587 (info
->objfile
->partial_symtabs
->global_psymbols
.data ()
588 + psymtab
->globals_offset
),
589 psymtab
->n_global_syms
, info
->cu_index
,
591 write_psymbols (info
->symtab
,
593 (info
->objfile
->partial_symtabs
->static_psymbols
.data ()
594 + psymtab
->statics_offset
),
595 psymtab
->n_static_syms
, info
->cu_index
,
598 info
->types_list
.append_uint (8, BFD_ENDIAN_LITTLE
,
599 to_underlying (entry
->per_cu
.sect_off
));
600 info
->types_list
.append_uint (8, BFD_ENDIAN_LITTLE
,
601 to_underlying (entry
->type_offset_in_tu
));
602 info
->types_list
.append_uint (8, BFD_ENDIAN_LITTLE
, entry
->signature
);
609 /* Recurse into all "included" dependencies and count their symbols as
610 if they appeared in this psymtab. */
613 recursively_count_psymbols (struct partial_symtab
*psymtab
,
616 for (int i
= 0; i
< psymtab
->number_of_dependencies
; ++i
)
617 if (psymtab
->dependencies
[i
]->user
!= NULL
)
618 recursively_count_psymbols (psymtab
->dependencies
[i
],
621 psyms_seen
+= psymtab
->n_global_syms
;
622 psyms_seen
+= psymtab
->n_static_syms
;
625 /* Recurse into all "included" dependencies and write their symbols as
626 if they appeared in this psymtab. */
629 recursively_write_psymbols (struct objfile
*objfile
,
630 struct partial_symtab
*psymtab
,
631 struct mapped_symtab
*symtab
,
632 std::unordered_set
<partial_symbol
*> &psyms_seen
,
633 offset_type cu_index
)
637 for (i
= 0; i
< psymtab
->number_of_dependencies
; ++i
)
638 if (psymtab
->dependencies
[i
]->user
!= NULL
)
639 recursively_write_psymbols (objfile
, psymtab
->dependencies
[i
],
640 symtab
, psyms_seen
, cu_index
);
642 write_psymbols (symtab
,
644 (objfile
->partial_symtabs
->global_psymbols
.data ()
645 + psymtab
->globals_offset
),
646 psymtab
->n_global_syms
, cu_index
,
648 write_psymbols (symtab
,
650 (objfile
->partial_symtabs
->static_psymbols
.data ()
651 + psymtab
->statics_offset
),
652 psymtab
->n_static_syms
, cu_index
,
656 /* DWARF-5 .debug_names builder. */
660 debug_names (struct dwarf2_per_objfile
*dwarf2_per_objfile
, bool is_dwarf64
,
661 bfd_endian dwarf5_byte_order
)
662 : m_dwarf5_byte_order (dwarf5_byte_order
),
663 m_dwarf32 (dwarf5_byte_order
),
664 m_dwarf64 (dwarf5_byte_order
),
666 ? static_cast<dwarf
&> (m_dwarf64
)
667 : static_cast<dwarf
&> (m_dwarf32
)),
668 m_name_table_string_offs (m_dwarf
.name_table_string_offs
),
669 m_name_table_entry_offs (m_dwarf
.name_table_entry_offs
),
670 m_debugstrlookup (dwarf2_per_objfile
)
673 int dwarf5_offset_size () const
675 const bool dwarf5_is_dwarf64
= &m_dwarf
== &m_dwarf64
;
676 return dwarf5_is_dwarf64
? 8 : 4;
679 /* Is this symbol from DW_TAG_compile_unit or DW_TAG_type_unit? */
680 enum class unit_kind
{ cu
, tu
};
682 /* Insert one symbol. */
683 void insert (const partial_symbol
*psym
, int cu_index
, bool is_static
,
686 const int dwarf_tag
= psymbol_tag (psym
);
689 const char *name
= symbol_search_name (&psym
->ginfo
);
691 if (psym
->ginfo
.language
== language_ada
)
693 /* We want to ensure that the Ada main function's name appears
694 verbatim in the index. However, this name will be of the
695 form "_ada_mumble", and will be rewritten by ada_decode.
696 So, recognize it specially here and add it to the index by
698 if (strcmp (main_name (), name
) == 0)
700 const auto insertpair
701 = m_name_to_value_set
.emplace (c_str_view (name
),
702 std::set
<symbol_value
> ());
703 std::set
<symbol_value
> &value_set
= insertpair
.first
->second
;
704 value_set
.emplace (symbol_value (dwarf_tag
, cu_index
, is_static
,
708 /* In order for the index to work when read back into gdb, it
709 has to supply a funny form of the name: it should be the
710 encoded name, with any suffixes stripped. Using the
711 ordinary encoded name will not work properly with the
712 searching logic in find_name_components_bounds; nor will
713 using the decoded name. Furthermore, an Ada "verbatim"
714 name (of the form "<MumBle>") must be entered without the
715 angle brackets. Note that the current index is unusual,
716 see PR symtab/24820 for details. */
717 std::string decoded
= ada_decode (name
);
718 if (decoded
[0] == '<')
719 name
= (char *) obstack_copy0 (&m_string_obstack
,
720 decoded
.c_str () + 1,
721 decoded
.length () - 2);
723 name
= obstack_strdup (&m_string_obstack
,
724 ada_encode (decoded
.c_str ()));
727 const auto insertpair
728 = m_name_to_value_set
.emplace (c_str_view (name
),
729 std::set
<symbol_value
> ());
730 std::set
<symbol_value
> &value_set
= insertpair
.first
->second
;
731 value_set
.emplace (symbol_value (dwarf_tag
, cu_index
, is_static
, kind
));
734 /* Build all the tables. All symbols must be already inserted.
735 This function does not call file_write, caller has to do it
739 /* Verify the build method has not be called twice. */
740 gdb_assert (m_abbrev_table
.empty ());
741 const size_t name_count
= m_name_to_value_set
.size ();
742 m_bucket_table
.resize
743 (std::pow (2, std::ceil (std::log2 (name_count
* 4 / 3))));
744 m_hash_table
.reserve (name_count
);
745 m_name_table_string_offs
.reserve (name_count
);
746 m_name_table_entry_offs
.reserve (name_count
);
748 /* Map each hash of symbol to its name and value. */
752 decltype (m_name_to_value_set
)::const_iterator it
;
754 std::vector
<std::forward_list
<hash_it_pair
>> bucket_hash
;
755 bucket_hash
.resize (m_bucket_table
.size ());
756 for (decltype (m_name_to_value_set
)::const_iterator it
757 = m_name_to_value_set
.cbegin ();
758 it
!= m_name_to_value_set
.cend ();
761 const char *const name
= it
->first
.c_str ();
762 const uint32_t hash
= dwarf5_djb_hash (name
);
763 hash_it_pair hashitpair
;
764 hashitpair
.hash
= hash
;
766 auto &slot
= bucket_hash
[hash
% bucket_hash
.size()];
767 slot
.push_front (std::move (hashitpair
));
769 for (size_t bucket_ix
= 0; bucket_ix
< bucket_hash
.size (); ++bucket_ix
)
771 const std::forward_list
<hash_it_pair
> &hashitlist
772 = bucket_hash
[bucket_ix
];
773 if (hashitlist
.empty ())
775 uint32_t &bucket_slot
= m_bucket_table
[bucket_ix
];
776 /* The hashes array is indexed starting at 1. */
777 store_unsigned_integer (reinterpret_cast<gdb_byte
*> (&bucket_slot
),
778 sizeof (bucket_slot
), m_dwarf5_byte_order
,
779 m_hash_table
.size () + 1);
780 for (const hash_it_pair
&hashitpair
: hashitlist
)
782 m_hash_table
.push_back (0);
783 store_unsigned_integer (reinterpret_cast<gdb_byte
*>
784 (&m_hash_table
.back ()),
785 sizeof (m_hash_table
.back ()),
786 m_dwarf5_byte_order
, hashitpair
.hash
);
787 const c_str_view
&name
= hashitpair
.it
->first
;
788 const std::set
<symbol_value
> &value_set
= hashitpair
.it
->second
;
789 m_name_table_string_offs
.push_back_reorder
790 (m_debugstrlookup
.lookup (name
.c_str ()));
791 m_name_table_entry_offs
.push_back_reorder (m_entry_pool
.size ());
792 gdb_assert (!value_set
.empty ());
793 for (const symbol_value
&value
: value_set
)
795 int &idx
= m_indexkey_to_idx
[index_key (value
.dwarf_tag
,
801 m_abbrev_table
.append_unsigned_leb128 (idx
);
802 m_abbrev_table
.append_unsigned_leb128 (value
.dwarf_tag
);
803 m_abbrev_table
.append_unsigned_leb128
804 (value
.kind
== unit_kind::cu
? DW_IDX_compile_unit
806 m_abbrev_table
.append_unsigned_leb128 (DW_FORM_udata
);
807 m_abbrev_table
.append_unsigned_leb128 (value
.is_static
808 ? DW_IDX_GNU_internal
809 : DW_IDX_GNU_external
);
810 m_abbrev_table
.append_unsigned_leb128 (DW_FORM_flag_present
);
812 /* Terminate attributes list. */
813 m_abbrev_table
.append_unsigned_leb128 (0);
814 m_abbrev_table
.append_unsigned_leb128 (0);
817 m_entry_pool
.append_unsigned_leb128 (idx
);
818 m_entry_pool
.append_unsigned_leb128 (value
.cu_index
);
821 /* Terminate the list of CUs. */
822 m_entry_pool
.append_unsigned_leb128 (0);
825 gdb_assert (m_hash_table
.size () == name_count
);
827 /* Terminate tags list. */
828 m_abbrev_table
.append_unsigned_leb128 (0);
831 /* Return .debug_names bucket count. This must be called only after
832 calling the build method. */
833 uint32_t bucket_count () const
835 /* Verify the build method has been already called. */
836 gdb_assert (!m_abbrev_table
.empty ());
837 const uint32_t retval
= m_bucket_table
.size ();
839 /* Check for overflow. */
840 gdb_assert (retval
== m_bucket_table
.size ());
844 /* Return .debug_names names count. This must be called only after
845 calling the build method. */
846 uint32_t name_count () const
848 /* Verify the build method has been already called. */
849 gdb_assert (!m_abbrev_table
.empty ());
850 const uint32_t retval
= m_hash_table
.size ();
852 /* Check for overflow. */
853 gdb_assert (retval
== m_hash_table
.size ());
857 /* Return number of bytes of .debug_names abbreviation table. This
858 must be called only after calling the build method. */
859 uint32_t abbrev_table_bytes () const
861 gdb_assert (!m_abbrev_table
.empty ());
862 return m_abbrev_table
.size ();
865 /* Recurse into all "included" dependencies and store their symbols
866 as if they appeared in this psymtab. */
867 void recursively_write_psymbols
868 (struct objfile
*objfile
,
869 struct partial_symtab
*psymtab
,
870 std::unordered_set
<partial_symbol
*> &psyms_seen
,
873 for (int i
= 0; i
< psymtab
->number_of_dependencies
; ++i
)
874 if (psymtab
->dependencies
[i
]->user
!= NULL
)
875 recursively_write_psymbols (objfile
, psymtab
->dependencies
[i
],
876 psyms_seen
, cu_index
);
878 write_psymbols (psyms_seen
,
879 (objfile
->partial_symtabs
->global_psymbols
.data ()
880 + psymtab
->globals_offset
),
881 psymtab
->n_global_syms
, cu_index
, false, unit_kind::cu
);
882 write_psymbols (psyms_seen
,
883 (objfile
->partial_symtabs
->static_psymbols
.data ()
884 + psymtab
->statics_offset
),
885 psymtab
->n_static_syms
, cu_index
, true, unit_kind::cu
);
888 /* Return number of bytes the .debug_names section will have. This
889 must be called only after calling the build method. */
890 size_t bytes () const
892 /* Verify the build method has been already called. */
893 gdb_assert (!m_abbrev_table
.empty ());
894 size_t expected_bytes
= 0;
895 expected_bytes
+= m_bucket_table
.size () * sizeof (m_bucket_table
[0]);
896 expected_bytes
+= m_hash_table
.size () * sizeof (m_hash_table
[0]);
897 expected_bytes
+= m_name_table_string_offs
.bytes ();
898 expected_bytes
+= m_name_table_entry_offs
.bytes ();
899 expected_bytes
+= m_abbrev_table
.size ();
900 expected_bytes
+= m_entry_pool
.size ();
901 return expected_bytes
;
904 /* Write .debug_names to FILE_NAMES and .debug_str addition to
905 FILE_STR. This must be called only after calling the build
907 void file_write (FILE *file_names
, FILE *file_str
) const
909 /* Verify the build method has been already called. */
910 gdb_assert (!m_abbrev_table
.empty ());
911 ::file_write (file_names
, m_bucket_table
);
912 ::file_write (file_names
, m_hash_table
);
913 m_name_table_string_offs
.file_write (file_names
);
914 m_name_table_entry_offs
.file_write (file_names
);
915 m_abbrev_table
.file_write (file_names
);
916 m_entry_pool
.file_write (file_names
);
917 m_debugstrlookup
.file_write (file_str
);
920 /* A helper user data for write_one_signatured_type. */
921 class write_one_signatured_type_data
924 write_one_signatured_type_data (debug_names
&nametable_
,
925 signatured_type_index_data
&&info_
)
926 : nametable (nametable_
), info (std::move (info_
))
928 debug_names
&nametable
;
929 struct signatured_type_index_data info
;
932 /* A helper function to pass write_one_signatured_type to
933 htab_traverse_noresize. */
935 write_one_signatured_type (void **slot
, void *d
)
937 write_one_signatured_type_data
*data
= (write_one_signatured_type_data
*) d
;
938 struct signatured_type_index_data
*info
= &data
->info
;
939 struct signatured_type
*entry
= (struct signatured_type
*) *slot
;
941 data
->nametable
.write_one_signatured_type (entry
, info
);
948 /* Storage for symbol names mapping them to their .debug_str section
950 class debug_str_lookup
954 /* Object costructor to be called for current DWARF2_PER_OBJFILE.
955 All .debug_str section strings are automatically stored. */
956 debug_str_lookup (struct dwarf2_per_objfile
*dwarf2_per_objfile
)
957 : m_abfd (dwarf2_per_objfile
->objfile
->obfd
),
958 m_dwarf2_per_objfile (dwarf2_per_objfile
)
960 dwarf2_read_section (dwarf2_per_objfile
->objfile
,
961 &dwarf2_per_objfile
->str
);
962 if (dwarf2_per_objfile
->str
.buffer
== NULL
)
964 for (const gdb_byte
*data
= dwarf2_per_objfile
->str
.buffer
;
965 data
< (dwarf2_per_objfile
->str
.buffer
966 + dwarf2_per_objfile
->str
.size
);)
968 const char *const s
= reinterpret_cast<const char *> (data
);
969 const auto insertpair
970 = m_str_table
.emplace (c_str_view (s
),
971 data
- dwarf2_per_objfile
->str
.buffer
);
972 if (!insertpair
.second
)
973 complaint (_("Duplicate string \"%s\" in "
974 ".debug_str section [in module %s]"),
975 s
, bfd_get_filename (m_abfd
));
976 data
+= strlen (s
) + 1;
980 /* Return offset of symbol name S in the .debug_str section. Add
981 such symbol to the section's end if it does not exist there
983 size_t lookup (const char *s
)
985 const auto it
= m_str_table
.find (c_str_view (s
));
986 if (it
!= m_str_table
.end ())
988 const size_t offset
= (m_dwarf2_per_objfile
->str
.size
989 + m_str_add_buf
.size ());
990 m_str_table
.emplace (c_str_view (s
), offset
);
991 m_str_add_buf
.append_cstr0 (s
);
995 /* Append the end of the .debug_str section to FILE. */
996 void file_write (FILE *file
) const
998 m_str_add_buf
.file_write (file
);
1002 std::unordered_map
<c_str_view
, size_t, c_str_view_hasher
> m_str_table
;
1004 struct dwarf2_per_objfile
*m_dwarf2_per_objfile
;
1006 /* Data to add at the end of .debug_str for new needed symbol names. */
1007 data_buf m_str_add_buf
;
1010 /* Container to map used DWARF tags to their .debug_names abbreviation
1015 index_key (int dwarf_tag_
, bool is_static_
, unit_kind kind_
)
1016 : dwarf_tag (dwarf_tag_
), is_static (is_static_
), kind (kind_
)
1021 operator== (const index_key
&other
) const
1023 return (dwarf_tag
== other
.dwarf_tag
&& is_static
== other
.is_static
1024 && kind
== other
.kind
);
1027 const int dwarf_tag
;
1028 const bool is_static
;
1029 const unit_kind kind
;
1032 /* Provide std::unordered_map::hasher for index_key. */
1033 class index_key_hasher
1037 operator () (const index_key
&key
) const
1039 return (std::hash
<int>() (key
.dwarf_tag
) << 1) | key
.is_static
;
1043 /* Parameters of one symbol entry. */
1047 const int dwarf_tag
, cu_index
;
1048 const bool is_static
;
1049 const unit_kind kind
;
1051 symbol_value (int dwarf_tag_
, int cu_index_
, bool is_static_
,
1053 : dwarf_tag (dwarf_tag_
), cu_index (cu_index_
), is_static (is_static_
),
1058 operator< (const symbol_value
&other
) const
1078 /* Abstract base class to unify DWARF-32 and DWARF-64 name table
1083 const bfd_endian dwarf5_byte_order
;
1085 explicit offset_vec (bfd_endian dwarf5_byte_order_
)
1086 : dwarf5_byte_order (dwarf5_byte_order_
)
1089 /* Call std::vector::reserve for NELEM elements. */
1090 virtual void reserve (size_t nelem
) = 0;
1092 /* Call std::vector::push_back with store_unsigned_integer byte
1093 reordering for ELEM. */
1094 virtual void push_back_reorder (size_t elem
) = 0;
1096 /* Return expected output size in bytes. */
1097 virtual size_t bytes () const = 0;
1099 /* Write name table to FILE. */
1100 virtual void file_write (FILE *file
) const = 0;
1103 /* Template to unify DWARF-32 and DWARF-64 output. */
1104 template<typename OffsetSize
>
1105 class offset_vec_tmpl
: public offset_vec
1108 explicit offset_vec_tmpl (bfd_endian dwarf5_byte_order_
)
1109 : offset_vec (dwarf5_byte_order_
)
1112 /* Implement offset_vec::reserve. */
1113 void reserve (size_t nelem
) override
1115 m_vec
.reserve (nelem
);
1118 /* Implement offset_vec::push_back_reorder. */
1119 void push_back_reorder (size_t elem
) override
1121 m_vec
.push_back (elem
);
1122 /* Check for overflow. */
1123 gdb_assert (m_vec
.back () == elem
);
1124 store_unsigned_integer (reinterpret_cast<gdb_byte
*> (&m_vec
.back ()),
1125 sizeof (m_vec
.back ()), dwarf5_byte_order
, elem
);
1128 /* Implement offset_vec::bytes. */
1129 size_t bytes () const override
1131 return m_vec
.size () * sizeof (m_vec
[0]);
1134 /* Implement offset_vec::file_write. */
1135 void file_write (FILE *file
) const override
1137 ::file_write (file
, m_vec
);
1141 std::vector
<OffsetSize
> m_vec
;
1144 /* Base class to unify DWARF-32 and DWARF-64 .debug_names output
1145 respecting name table width. */
1149 offset_vec
&name_table_string_offs
, &name_table_entry_offs
;
1151 dwarf (offset_vec
&name_table_string_offs_
,
1152 offset_vec
&name_table_entry_offs_
)
1153 : name_table_string_offs (name_table_string_offs_
),
1154 name_table_entry_offs (name_table_entry_offs_
)
1159 /* Template to unify DWARF-32 and DWARF-64 .debug_names output
1160 respecting name table width. */
1161 template<typename OffsetSize
>
1162 class dwarf_tmpl
: public dwarf
1165 explicit dwarf_tmpl (bfd_endian dwarf5_byte_order_
)
1166 : dwarf (m_name_table_string_offs
, m_name_table_entry_offs
),
1167 m_name_table_string_offs (dwarf5_byte_order_
),
1168 m_name_table_entry_offs (dwarf5_byte_order_
)
1172 offset_vec_tmpl
<OffsetSize
> m_name_table_string_offs
;
1173 offset_vec_tmpl
<OffsetSize
> m_name_table_entry_offs
;
1176 /* Try to reconstruct original DWARF tag for given partial_symbol.
1177 This function is not DWARF-5 compliant but it is sufficient for
1178 GDB as a DWARF-5 index consumer. */
1179 static int psymbol_tag (const struct partial_symbol
*psym
)
1181 domain_enum domain
= psym
->domain
;
1182 enum address_class aclass
= psym
->aclass
;
1190 return DW_TAG_subprogram
;
1192 return DW_TAG_typedef
;
1194 case LOC_CONST_BYTES
:
1195 case LOC_OPTIMIZED_OUT
:
1197 return DW_TAG_variable
;
1199 /* Note: It's currently impossible to recognize psyms as enum values
1200 short of reading the type info. For now punt. */
1201 return DW_TAG_variable
;
1203 /* There are other LOC_FOO values that one might want to classify
1204 as variables, but dwarf2read.c doesn't currently use them. */
1205 return DW_TAG_variable
;
1208 return DW_TAG_structure_type
;
1214 /* Call insert for all partial symbols and mark them in PSYMS_SEEN. */
1215 void write_psymbols (std::unordered_set
<partial_symbol
*> &psyms_seen
,
1216 struct partial_symbol
**psymp
, int count
, int cu_index
,
1217 bool is_static
, unit_kind kind
)
1219 for (; count
-- > 0; ++psymp
)
1221 struct partial_symbol
*psym
= *psymp
;
1223 /* Only add a given psymbol once. */
1224 if (psyms_seen
.insert (psym
).second
)
1225 insert (psym
, cu_index
, is_static
, kind
);
1229 /* A helper function that writes a single signatured_type
1230 to a debug_names. */
1232 write_one_signatured_type (struct signatured_type
*entry
,
1233 struct signatured_type_index_data
*info
)
1235 struct partial_symtab
*psymtab
= entry
->per_cu
.v
.psymtab
;
1237 write_psymbols (info
->psyms_seen
,
1238 (info
->objfile
->partial_symtabs
->global_psymbols
.data ()
1239 + psymtab
->globals_offset
),
1240 psymtab
->n_global_syms
, info
->cu_index
, false,
1242 write_psymbols (info
->psyms_seen
,
1243 (info
->objfile
->partial_symtabs
->static_psymbols
.data ()
1244 + psymtab
->statics_offset
),
1245 psymtab
->n_static_syms
, info
->cu_index
, true,
1248 info
->types_list
.append_uint (dwarf5_offset_size (), m_dwarf5_byte_order
,
1249 to_underlying (entry
->per_cu
.sect_off
));
1254 /* Store value of each symbol. */
1255 std::unordered_map
<c_str_view
, std::set
<symbol_value
>, c_str_view_hasher
>
1256 m_name_to_value_set
;
1258 /* Tables of DWARF-5 .debug_names. They are in object file byte
1260 std::vector
<uint32_t> m_bucket_table
;
1261 std::vector
<uint32_t> m_hash_table
;
1263 const bfd_endian m_dwarf5_byte_order
;
1264 dwarf_tmpl
<uint32_t> m_dwarf32
;
1265 dwarf_tmpl
<uint64_t> m_dwarf64
;
1267 offset_vec
&m_name_table_string_offs
, &m_name_table_entry_offs
;
1268 debug_str_lookup m_debugstrlookup
;
1270 /* Map each used .debug_names abbreviation tag parameter to its
1272 std::unordered_map
<index_key
, int, index_key_hasher
> m_indexkey_to_idx
;
1274 /* Next unused .debug_names abbreviation tag for
1275 m_indexkey_to_idx. */
1278 /* .debug_names abbreviation table. */
1279 data_buf m_abbrev_table
;
1281 /* .debug_names entry pool. */
1282 data_buf m_entry_pool
;
1284 /* Temporary storage for Ada names. */
1285 auto_obstack m_string_obstack
;
1288 /* Return iff any of the needed offsets does not fit into 32-bit
1289 .debug_names section. */
1292 check_dwarf64_offsets (struct dwarf2_per_objfile
*dwarf2_per_objfile
)
1294 for (dwarf2_per_cu_data
*per_cu
: dwarf2_per_objfile
->all_comp_units
)
1296 if (to_underlying (per_cu
->sect_off
) >= (static_cast<uint64_t> (1) << 32))
1299 for (const signatured_type
*sigtype
: dwarf2_per_objfile
->all_type_units
)
1301 const dwarf2_per_cu_data
&per_cu
= sigtype
->per_cu
;
1303 if (to_underlying (per_cu
.sect_off
) >= (static_cast<uint64_t> (1) << 32))
1309 /* The psyms_seen set is potentially going to be largish (~40k
1310 elements when indexing a -g3 build of GDB itself). Estimate the
1311 number of elements in order to avoid too many rehashes, which
1312 require rebuilding buckets and thus many trips to
1316 psyms_seen_size (struct dwarf2_per_objfile
*dwarf2_per_objfile
)
1318 size_t psyms_count
= 0;
1319 for (dwarf2_per_cu_data
*per_cu
: dwarf2_per_objfile
->all_comp_units
)
1321 struct partial_symtab
*psymtab
= per_cu
->v
.psymtab
;
1323 if (psymtab
!= NULL
&& psymtab
->user
== NULL
)
1324 recursively_count_psymbols (psymtab
, psyms_count
);
1326 /* Generating an index for gdb itself shows a ratio of
1327 TOTAL_SEEN_SYMS/UNIQUE_SYMS or ~5. 4 seems like a good bet. */
1328 return psyms_count
/ 4;
1331 /* Assert that FILE's size is EXPECTED_SIZE. Assumes file's seek
1332 position is at the end of the file. */
1335 assert_file_size (FILE *file
, size_t expected_size
)
1337 const auto file_size
= ftell (file
);
1338 if (file_size
== -1)
1339 perror_with_name (("ftell"));
1340 gdb_assert (file_size
== expected_size
);
1343 /* Write a gdb index file to OUT_FILE from all the sections passed as
1347 write_gdbindex_1 (FILE *out_file
,
1348 const data_buf
&cu_list
,
1349 const data_buf
&types_cu_list
,
1350 const data_buf
&addr_vec
,
1351 const data_buf
&symtab_vec
,
1352 const data_buf
&constant_pool
)
1355 const offset_type size_of_header
= 6 * sizeof (offset_type
);
1356 offset_type total_len
= size_of_header
;
1358 /* The version number. */
1359 contents
.append_data (MAYBE_SWAP (8));
1361 /* The offset of the CU list from the start of the file. */
1362 contents
.append_data (MAYBE_SWAP (total_len
));
1363 total_len
+= cu_list
.size ();
1365 /* The offset of the types CU list from the start of the file. */
1366 contents
.append_data (MAYBE_SWAP (total_len
));
1367 total_len
+= types_cu_list
.size ();
1369 /* The offset of the address table from the start of the file. */
1370 contents
.append_data (MAYBE_SWAP (total_len
));
1371 total_len
+= addr_vec
.size ();
1373 /* The offset of the symbol table from the start of the file. */
1374 contents
.append_data (MAYBE_SWAP (total_len
));
1375 total_len
+= symtab_vec
.size ();
1377 /* The offset of the constant pool from the start of the file. */
1378 contents
.append_data (MAYBE_SWAP (total_len
));
1379 total_len
+= constant_pool
.size ();
1381 gdb_assert (contents
.size () == size_of_header
);
1383 contents
.file_write (out_file
);
1384 cu_list
.file_write (out_file
);
1385 types_cu_list
.file_write (out_file
);
1386 addr_vec
.file_write (out_file
);
1387 symtab_vec
.file_write (out_file
);
1388 constant_pool
.file_write (out_file
);
1390 assert_file_size (out_file
, total_len
);
1393 /* Write contents of a .gdb_index section for OBJFILE into OUT_FILE.
1394 If OBJFILE has an associated dwz file, write contents of a .gdb_index
1395 section for that dwz file into DWZ_OUT_FILE. If OBJFILE does not have an
1396 associated dwz file, DWZ_OUT_FILE must be NULL. */
1399 write_gdbindex (struct dwarf2_per_objfile
*dwarf2_per_objfile
, FILE *out_file
,
1402 struct objfile
*objfile
= dwarf2_per_objfile
->objfile
;
1403 mapped_symtab symtab
;
1404 data_buf objfile_cu_list
;
1405 data_buf dwz_cu_list
;
1407 /* While we're scanning CU's create a table that maps a psymtab pointer
1408 (which is what addrmap records) to its index (which is what is recorded
1409 in the index file). This will later be needed to write the address
1411 psym_index_map cu_index_htab
;
1412 cu_index_htab
.reserve (dwarf2_per_objfile
->all_comp_units
.size ());
1414 /* The CU list is already sorted, so we don't need to do additional
1415 work here. Also, the debug_types entries do not appear in
1416 all_comp_units, but only in their own hash table. */
1418 std::unordered_set
<partial_symbol
*> psyms_seen
1419 (psyms_seen_size (dwarf2_per_objfile
));
1420 for (int i
= 0; i
< dwarf2_per_objfile
->all_comp_units
.size (); ++i
)
1422 struct dwarf2_per_cu_data
*per_cu
1423 = dwarf2_per_objfile
->all_comp_units
[i
];
1424 struct partial_symtab
*psymtab
= per_cu
->v
.psymtab
;
1426 /* CU of a shared file from 'dwz -m' may be unused by this main file.
1427 It may be referenced from a local scope but in such case it does not
1428 need to be present in .gdb_index. */
1429 if (psymtab
== NULL
)
1432 if (psymtab
->user
== NULL
)
1433 recursively_write_psymbols (objfile
, psymtab
, &symtab
,
1436 const auto insertpair
= cu_index_htab
.emplace (psymtab
, i
);
1437 gdb_assert (insertpair
.second
);
1439 /* The all_comp_units list contains CUs read from the objfile as well as
1440 from the eventual dwz file. We need to place the entry in the
1441 corresponding index. */
1442 data_buf
&cu_list
= per_cu
->is_dwz
? dwz_cu_list
: objfile_cu_list
;
1443 cu_list
.append_uint (8, BFD_ENDIAN_LITTLE
,
1444 to_underlying (per_cu
->sect_off
));
1445 cu_list
.append_uint (8, BFD_ENDIAN_LITTLE
, per_cu
->length
);
1448 /* Dump the address map. */
1450 write_address_map (objfile
, addr_vec
, cu_index_htab
);
1452 /* Write out the .debug_type entries, if any. */
1453 data_buf types_cu_list
;
1454 if (dwarf2_per_objfile
->signatured_types
)
1456 signatured_type_index_data
sig_data (types_cu_list
,
1459 sig_data
.objfile
= objfile
;
1460 sig_data
.symtab
= &symtab
;
1461 sig_data
.cu_index
= dwarf2_per_objfile
->all_comp_units
.size ();
1462 htab_traverse_noresize (dwarf2_per_objfile
->signatured_types
,
1463 write_one_signatured_type
, &sig_data
);
1466 /* Now that we've processed all symbols we can shrink their cu_indices
1468 uniquify_cu_indices (&symtab
);
1470 data_buf symtab_vec
, constant_pool
;
1471 write_hash_table (&symtab
, symtab_vec
, constant_pool
);
1473 write_gdbindex_1(out_file
, objfile_cu_list
, types_cu_list
, addr_vec
,
1474 symtab_vec
, constant_pool
);
1476 if (dwz_out_file
!= NULL
)
1477 write_gdbindex_1 (dwz_out_file
, dwz_cu_list
, {}, {}, {}, {});
1479 gdb_assert (dwz_cu_list
.empty ());
1482 /* DWARF-5 augmentation string for GDB's DW_IDX_GNU_* extension. */
1483 static const gdb_byte dwarf5_gdb_augmentation
[] = { 'G', 'D', 'B', 0 };
1485 /* Write a new .debug_names section for OBJFILE into OUT_FILE, write
1486 needed addition to .debug_str section to OUT_FILE_STR. Return how
1487 many bytes were expected to be written into OUT_FILE. */
1490 write_debug_names (struct dwarf2_per_objfile
*dwarf2_per_objfile
,
1491 FILE *out_file
, FILE *out_file_str
)
1493 const bool dwarf5_is_dwarf64
= check_dwarf64_offsets (dwarf2_per_objfile
);
1494 struct objfile
*objfile
= dwarf2_per_objfile
->objfile
;
1495 const enum bfd_endian dwarf5_byte_order
1496 = gdbarch_byte_order (get_objfile_arch (objfile
));
1498 /* The CU list is already sorted, so we don't need to do additional
1499 work here. Also, the debug_types entries do not appear in
1500 all_comp_units, but only in their own hash table. */
1502 debug_names
nametable (dwarf2_per_objfile
, dwarf5_is_dwarf64
,
1504 std::unordered_set
<partial_symbol
*>
1505 psyms_seen (psyms_seen_size (dwarf2_per_objfile
));
1506 for (int i
= 0; i
< dwarf2_per_objfile
->all_comp_units
.size (); ++i
)
1508 const dwarf2_per_cu_data
*per_cu
= dwarf2_per_objfile
->all_comp_units
[i
];
1509 partial_symtab
*psymtab
= per_cu
->v
.psymtab
;
1511 /* CU of a shared file from 'dwz -m' may be unused by this main
1512 file. It may be referenced from a local scope but in such
1513 case it does not need to be present in .debug_names. */
1514 if (psymtab
== NULL
)
1517 if (psymtab
->user
== NULL
)
1518 nametable
.recursively_write_psymbols (objfile
, psymtab
, psyms_seen
, i
);
1520 cu_list
.append_uint (nametable
.dwarf5_offset_size (), dwarf5_byte_order
,
1521 to_underlying (per_cu
->sect_off
));
1524 /* Write out the .debug_type entries, if any. */
1525 data_buf types_cu_list
;
1526 if (dwarf2_per_objfile
->signatured_types
)
1528 debug_names::write_one_signatured_type_data
sig_data (nametable
,
1529 signatured_type_index_data (types_cu_list
, psyms_seen
));
1531 sig_data
.info
.objfile
= objfile
;
1532 /* It is used only for gdb_index. */
1533 sig_data
.info
.symtab
= nullptr;
1534 sig_data
.info
.cu_index
= 0;
1535 htab_traverse_noresize (dwarf2_per_objfile
->signatured_types
,
1536 debug_names::write_one_signatured_type
,
1542 /* No addr_vec - DWARF-5 uses .debug_aranges generated by GCC. */
1544 const offset_type bytes_of_header
1545 = ((dwarf5_is_dwarf64
? 12 : 4)
1547 + sizeof (dwarf5_gdb_augmentation
));
1548 size_t expected_bytes
= 0;
1549 expected_bytes
+= bytes_of_header
;
1550 expected_bytes
+= cu_list
.size ();
1551 expected_bytes
+= types_cu_list
.size ();
1552 expected_bytes
+= nametable
.bytes ();
1555 if (!dwarf5_is_dwarf64
)
1557 const uint64_t size64
= expected_bytes
- 4;
1558 gdb_assert (size64
< 0xfffffff0);
1559 header
.append_uint (4, dwarf5_byte_order
, size64
);
1563 header
.append_uint (4, dwarf5_byte_order
, 0xffffffff);
1564 header
.append_uint (8, dwarf5_byte_order
, expected_bytes
- 12);
1567 /* The version number. */
1568 header
.append_uint (2, dwarf5_byte_order
, 5);
1571 header
.append_uint (2, dwarf5_byte_order
, 0);
1573 /* comp_unit_count - The number of CUs in the CU list. */
1574 header
.append_uint (4, dwarf5_byte_order
,
1575 dwarf2_per_objfile
->all_comp_units
.size ());
1577 /* local_type_unit_count - The number of TUs in the local TU
1579 header
.append_uint (4, dwarf5_byte_order
,
1580 dwarf2_per_objfile
->all_type_units
.size ());
1582 /* foreign_type_unit_count - The number of TUs in the foreign TU
1584 header
.append_uint (4, dwarf5_byte_order
, 0);
1586 /* bucket_count - The number of hash buckets in the hash lookup
1588 header
.append_uint (4, dwarf5_byte_order
, nametable
.bucket_count ());
1590 /* name_count - The number of unique names in the index. */
1591 header
.append_uint (4, dwarf5_byte_order
, nametable
.name_count ());
1593 /* abbrev_table_size - The size in bytes of the abbreviations
1595 header
.append_uint (4, dwarf5_byte_order
, nametable
.abbrev_table_bytes ());
1597 /* augmentation_string_size - The size in bytes of the augmentation
1598 string. This value is rounded up to a multiple of 4. */
1599 static_assert (sizeof (dwarf5_gdb_augmentation
) % 4 == 0, "");
1600 header
.append_uint (4, dwarf5_byte_order
, sizeof (dwarf5_gdb_augmentation
));
1601 header
.append_data (dwarf5_gdb_augmentation
);
1603 gdb_assert (header
.size () == bytes_of_header
);
1605 header
.file_write (out_file
);
1606 cu_list
.file_write (out_file
);
1607 types_cu_list
.file_write (out_file
);
1608 nametable
.file_write (out_file
, out_file_str
);
1610 assert_file_size (out_file
, expected_bytes
);
1613 /* This represents an index file being written (work-in-progress).
1615 The data is initially written to a temporary file. When the finalize method
1616 is called, the file is closed and moved to its final location.
1618 On failure (if this object is being destroyed with having called finalize),
1619 the temporary file is closed and deleted. */
1621 struct index_wip_file
1623 index_wip_file (const char *dir
, const char *basename
,
1626 filename
= (std::string (dir
) + SLASH_STRING
+ basename
1629 filename_temp
= make_temp_filename (filename
);
1631 scoped_fd
out_file_fd (gdb_mkostemp_cloexec (filename_temp
.data (),
1633 if (out_file_fd
.get () == -1)
1634 perror_with_name (("mkstemp"));
1636 out_file
= out_file_fd
.to_file ("wb");
1638 if (out_file
== nullptr)
1639 error (_("Can't open `%s' for writing"), filename_temp
.data ());
1641 unlink_file
.emplace (filename_temp
.data ());
1646 /* We want to keep the file. */
1647 unlink_file
->keep ();
1649 /* Close and move the str file in place. */
1650 unlink_file
.reset ();
1651 if (rename (filename_temp
.data (), filename
.c_str ()) != 0)
1652 perror_with_name (("rename"));
1655 std::string filename
;
1656 gdb::char_vector filename_temp
;
1658 /* Order matters here; we want FILE to be closed before
1659 FILENAME_TEMP is unlinked, because on MS-Windows one cannot
1660 delete a file that is still open. So, we wrap the unlinker in an
1661 optional and emplace it once we know the file name. */
1662 gdb::optional
<gdb::unlinker
> unlink_file
;
1664 gdb_file_up out_file
;
1667 /* See dwarf-index-write.h. */
1670 write_psymtabs_to_index (struct dwarf2_per_objfile
*dwarf2_per_objfile
,
1671 const char *dir
, const char *basename
,
1672 const char *dwz_basename
,
1673 dw_index_kind index_kind
)
1675 struct objfile
*objfile
= dwarf2_per_objfile
->objfile
;
1677 if (dwarf2_per_objfile
->using_index
)
1678 error (_("Cannot use an index to create the index"));
1680 if (dwarf2_per_objfile
->types
.size () > 1)
1681 error (_("Cannot make an index when the file has multiple .debug_types sections"));
1683 if (!objfile
->partial_symtabs
->psymtabs
1684 || !objfile
->partial_symtabs
->psymtabs_addrmap
)
1688 if (stat (objfile_name (objfile
), &st
) < 0)
1689 perror_with_name (objfile_name (objfile
));
1691 const char *index_suffix
= (index_kind
== dw_index_kind::DEBUG_NAMES
1692 ? INDEX5_SUFFIX
: INDEX4_SUFFIX
);
1694 index_wip_file
objfile_index_wip (dir
, basename
, index_suffix
);
1695 gdb::optional
<index_wip_file
> dwz_index_wip
;
1697 if (dwz_basename
!= NULL
)
1698 dwz_index_wip
.emplace (dir
, dwz_basename
, index_suffix
);
1700 if (index_kind
== dw_index_kind::DEBUG_NAMES
)
1702 index_wip_file
str_wip_file (dir
, basename
, DEBUG_STR_SUFFIX
);
1704 write_debug_names (dwarf2_per_objfile
, objfile_index_wip
.out_file
.get (),
1705 str_wip_file
.out_file
.get ());
1707 str_wip_file
.finalize ();
1710 write_gdbindex (dwarf2_per_objfile
, objfile_index_wip
.out_file
.get (),
1711 (dwz_index_wip
.has_value ()
1712 ? dwz_index_wip
->out_file
.get () : NULL
));
1714 objfile_index_wip
.finalize ();
1716 if (dwz_index_wip
.has_value ())
1717 dwz_index_wip
->finalize ();
1720 /* Implementation of the `save gdb-index' command.
1722 Note that the .gdb_index file format used by this command is
1723 documented in the GDB manual. Any changes here must be documented
1727 save_gdb_index_command (const char *arg
, int from_tty
)
1729 const char dwarf5space
[] = "-dwarf-5 ";
1730 dw_index_kind index_kind
= dw_index_kind::GDB_INDEX
;
1735 arg
= skip_spaces (arg
);
1736 if (strncmp (arg
, dwarf5space
, strlen (dwarf5space
)) == 0)
1738 index_kind
= dw_index_kind::DEBUG_NAMES
;
1739 arg
+= strlen (dwarf5space
);
1740 arg
= skip_spaces (arg
);
1744 error (_("usage: save gdb-index [-dwarf-5] DIRECTORY"));
1746 for (objfile
*objfile
: current_program_space
->objfiles ())
1750 /* If the objfile does not correspond to an actual file, skip it. */
1751 if (stat (objfile_name (objfile
), &st
) < 0)
1754 struct dwarf2_per_objfile
*dwarf2_per_objfile
1755 = get_dwarf2_per_objfile (objfile
);
1757 if (dwarf2_per_objfile
!= NULL
)
1761 const char *basename
= lbasename (objfile_name (objfile
));
1762 const dwz_file
*dwz
= dwarf2_get_dwz_file (dwarf2_per_objfile
);
1763 const char *dwz_basename
= NULL
;
1766 dwz_basename
= lbasename (dwz
->filename ());
1768 write_psymtabs_to_index (dwarf2_per_objfile
, arg
, basename
,
1769 dwz_basename
, index_kind
);
1771 catch (const gdb_exception_error
&except
)
1773 exception_fprintf (gdb_stderr
, except
,
1774 _("Error while writing index for `%s': "),
1775 objfile_name (objfile
));
1783 _initialize_dwarf_index_write ()
1785 cmd_list_element
*c
= add_cmd ("gdb-index", class_files
,
1786 save_gdb_index_command
, _("\
1787 Save a gdb-index file.\n\
1788 Usage: save gdb-index [-dwarf-5] DIRECTORY\n\
1790 No options create one file with .gdb-index extension for pre-DWARF-5\n\
1791 compatible .gdb_index section. With -dwarf-5 creates two files with\n\
1792 extension .debug_names and .debug_str for DWARF-5 .debug_names section."),
1794 set_cmd_completer (c
, filename_completer
);