For this enum:
typedef unsigned char byte;
enum byte_enum : byte
{
byte_val = 128
};
The unsigned attribute is not set:
(gdb) p byte_val
$1 = -128
That's because it uses the attributes of the 'byte' typedef for the enum.
So this changes it to use the attributes of the underlying 'unsigned char'
instead.
gdb/ChangeLog:
2020-04-03 Hannes Domani <ssbssa@yahoo.de>
PR gdb/25325
* dwarf2/read.c (read_enumeration_type): Fix typed enum attributes.
gdb/testsuite/ChangeLog:
2020-04-03 Hannes Domani <ssbssa@yahoo.de>
PR gdb/25325
* gdb.cp/typed-enum.cc: New test.
* gdb.cp/typed-enum.exp: New file.
While profiling the DWARF reader, I noticed that
partial_die_info::read creates a vector to store attributes. However,
the vector is not needed, as this code only processes a single
attribute at a time.
This patch removes the vector. On my machine, this improves the time
of "./gdb ./gdb" from 2.22 seconds to 1.92 seconds (mean times over 10
runs).
Note that the attribute is initialized by read_attribute, so it does
not need any special initialization. Avoiding this also improves
performance a bit.
Tested on x86-64 Fedora 30. I'm checking this in.
gdb/ChangeLog
2020-04-02 Tom Tromey <tromey@adacore.com>
* dwarf2/read.c (partial_die_info::read): Do not create a vector
of attributes.
When running test-case gdb.ada/access_to_packed_array we have:
...
(gdb) print pack.a^M
$1 = (0 => 1, 2, 3, 4, 5, 6, 7, 8, 9, 10)^M
...
but with target board readnow.exp, we have instead:
...
(gdb) print pack.a^M
'pack.a' has unknown type; cast it to its declared type^M
...
The symbol is normally found by the map_matching_symbols call in
ada-lang.c:add_nonlocal_symbols:
...
for (objfile *objfile : current_program_space->objfiles ())
{
data.objfile = objfile;
objfile->sf->qf->map_matching_symbols (objfile, lookup_name,
domain, global, callback,
(is_wild_match
? NULL : compare_names));
...
which maps onto psym_map_matching_symbols.
Function psym_map_matching_symbols iterates over all the partial symtabs,
and:
- if not expanded, searches in the partial symtab:
- if not found, continues to the next
- if found, expands into full symtab
- searches in the full symtab
However, with -readnow the call maps onto dw2_map_matching_symbols instead,
which is unimplemented, and consequently no symbol is found.
Fix this by detecting -readnow in dw2_map_matching_symbols, and handling that
appropriately given that partial symtabs are not present, and full symtabs
are: iterate over all the symtabs and search them.
Tested on x86_64-linux, with native and target board -readnow.
This removes 217 FAILs with board -readnow.
gdb/ChangeLog:
2020-04-02 Tom de Vries <tdevries@suse.de>
PR ada/24671
* dwarf2/read.c (dw2_map_matching_symbols): Handle -readnow.
When language is set to auto, part of loading an executable is to update the
language accordingly. This is implemented by set_initial_language.
The implementation of set_initial_language works as follows:
- check if any objfile in the progspace has name_of_main/language_of_main
set, and if so, use the first one found. [ This is what you get f.i. when
using dwarf with DW_AT_main_subprogram. ]
- otherwise, check for known names in the minimal symbols, and either:
- use the associated language if any (f.i. for ada), or
- lookup the symbol in the symtab for the name and use the symbol language
(f.i. for c/c++).
The symbol lookup can be slow though.
In the case of the cc1 binary from PR23710 comment 1, getting to the initial
prompt takes ~8s:
...
$ time.sh gdb cc1 -batch -ex "show language"
The current source language is "auto; currently c++".
maxmem: 1272260
real: 8.05
user: 7.73
system: 0.38
...
but if we skip guessing the initial language by setting it instead, it takes
only ~4s:
...
$ time.sh gdb -iex "set language c++" cc1 -batch -ex "show language"
The current source language is "c++".
maxmem: 498272
real: 3.99
user: 3.90
system: 0.15
...
In both cases, we load the partial symbols for the executable, but in the
first case only we also do a lookup of main, which causes the corresponding
partial symtab to be expanded into a full symtab.
Ideally, we'd like to get the language of the symbol without triggering
expansion into a full symtab, and get the speedup without having to set the
language manually.
There's a related fixme in the header comment of set_initial_language:
...
/* Set the initial language.
FIXME: A better solution would be to record the language in the
psymtab when reading partial symbols, and then use it (if known) to
set the language. This would be a win for formats that encode the
language in an easily discoverable place, such as DWARF. For
stabs, we can jump through hoops looking for specially named
symbols or try to intuit the language from the specific type of
stabs we find, but we can't do that until later when we read in
full symbols. */
void
set_initial_language (void)
...
Since we're already tracking the language of partial symbols, use this to set
the language for the main symbol.
Note that this search in partial symbol tables is not guaranteed to yield the
same result as the lookup_symbol_in_language call currently done in
set_initial_language.
Build and reg-tested on x86_64-linux.
gdb/ChangeLog:
2020-04-02 Tom de Vries <tdevries@suse.de>
* dwarf2/read.c (dwarf2_gdb_index_functions,
dwarf2_debug_names_functions): Init lookup_global_symbol_language with
NULL.
* psymtab.c (psym_lookup_global_symbol_language): New function.
(psym_functions): Init psym_lookup_global_symbol_language with
psym_lookup_global_symbol_language.
* symfile-debug.c (debug_sym_quick_functions): Init
lookup_global_symbol_language with NULL.
* symfile.c (set_initial_language): Remove fixme comment.
* symfile.h (struct quick_symbol_functions): Add
lookup_global_symbol_language.
* symtab.c (find_quick_global_symbol_language): New function.
(find_main_name): Use find_quick_global_symbol_language.
gdb/testsuite/ChangeLog:
2020-04-02 Tom de Vries <tdevries@suse.de>
* gdb.base/main-psymtab.exp: New file.
This patch changes how complex types are created. init_complex_type
and arch_complex_type are unified, and complex types are reused, by
attaching them to the underlying scalar type.
gdb/ChangeLog
2020-04-01 Tom Tromey <tom@tromey.com>
* stabsread.c (rs6000_builtin_type, read_sun_floating_type)
(read_range_type): Update.
* mdebugread.c (basic_type): Update.
* go-lang.c (build_go_types): Use init_complex_type.
* gdbtypes.h (struct main_type) <complex_type>: New member.
(init_complex_type): Update.
(arch_complex_type): Don't declare.
* gdbtypes.c (init_complex_type): Remove "objfile" parameter.
Make name if none given. Use alloc_type_copy. Look for cached
complex type.
(arch_complex_type): Remove.
(gdbtypes_post_init): Use init_complex_type.
* f-lang.c (build_fortran_types): Use init_complex_type.
* dwarf2/read.c (read_base_type): Update.
* d-lang.c (build_d_types): Use init_complex_type.
* ctfread.c (read_base_type): Update.
lookup_name_info always copies the name that is passed in. However,
normally a copy is not needed. This patch changes this class to avoid
copying. This required changing the "name" method to return something
else; I chose a gdb::string_view, to avoid excessive calls to strlen
in the code using the lookup_name_info. However, as this class does
not allow an arbitrary string_view, I've also added a c_str method
that guarantees a \0-terminated result -- a pedantic difference but
one that respects the string_view contract, IMO.
gdb/ChangeLog
2020-04-01 Tom Tromey <tromey@adacore.com>
* symtab.h (class lookup_name_info) <lookup_name_info>: Change
"name" parameter to rvalue reference. Initialize m_name_holder.
<lookup_name_info>: New overloads.
<name>: Return gdb::string_view.
<c_str>: New method.
<make_ignore_params>: Update.
<search_name_hash>: Update.
<language_lookup_name>: Return const char *.
<m_name>: Change type.
* symtab.c (demangle_for_lookup_info::demangle_for_lookup_info)
(demangle_for_lookup_info::demangle_for_lookup_info): Update.
(lookup_name_info::match_any): Update.
* psymtab.c (match_partial_symbol, lookup_partial_symbol):
Update.
* minsyms.c (linkage_name_str): Update.
* language.c (default_symbol_name_matcher): Update.
* dwarf2/read.c (mapped_index_base::find_name_components_bounds):
Update.
* ada-lang.c (ada_fold_name): Change parameter to string_view.
(ada_lookup_name_info::ada_lookup_name_info): Update.
(literal_symbol_name_matcher): Update.
This method calls partial_symtab::expand_psymtab on all dependencies of
a psymtab. Given that there is also a partial_symtab::read_symtab
method, I think it would be clearer to name this function
expand_dependencies, rather than read_dependencies.
gdb/ChangeLog:
* psympriv.h (partial_symtab) <read_dependencies>: Rename to...
<expand_dependencies>: ... this.
* psymtab.c (partial_symtab::read_dependencies): Rename to...
(partial_symtab::expand_dependencies): ... this.
* dwarf2/read.c (dwarf2_include_psymtab) <expand_psymtab>:
Update.
(dwarf2_psymtab::expand_psymtab): Update.
* dbxread.c (dbx_psymtab_to_symtab_1): Update.
* mdebugread.c (psymtab_to_symtab_1): Update.
* xcoffread.c (xcoff_psymtab_to_symtab_1): Update.
I noticed that the start of read_attribute_reprocess had the wrong
formatting. This patch fixes it.
gdb/ChangeLog
2020-03-28 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (read_attribute_reprocess): Fix formatting.
This changes dwarf2_get_ref_die_offset and
dwarf2_get_attr_constant_value to be methods on struct attribute.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (handle_data_member_location, dwarf2_add_field)
(mark_common_block_symbol_computed, read_tag_string_type)
(attr_to_dynamic_prop, read_subrange_type): Update.
(dwarf2_get_ref_die_offset, dwarf2_get_attr_constant_value): Move
to be methods on struct attribute.
(skip_one_die, process_imported_unit_die, read_namespace_alias)
(read_call_site_scope, partial_die_info::read)
(partial_die_info::read, lookup_die_type, follow_die_ref):
Update.
* dwarf2/attribute.c (attribute::get_ref_die_offset): New method,
from dwarf2_get_ref_die_offset.
(attribute::constant_value): New method, from
dwarf2_get_attr_constant_value.
* dwarf2/attribute.h (struct attribute) <get_ref_die_offset>:
Declare method.
<constant_value>: New method.
This moves the DWARF debugging functions that stringify various
constants to a new file, dwarf2/stringify.c.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (dwarf_unit_type_name, dwarf_tag_name)
(dwarf_attr_name, dwarf_form_name, dwarf_bool_name)
(dwarf_type_encoding_name): Move to stringify.c.
* Makefile.in (COMMON_SFILES): Add dwarf2/stringify.c.
* dwarf2/stringify.c: New file.
* dwarf2/stringify.h: New file.
This changes lookup_addr_base and lookup_ranges_base to be methods on
die_info.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/die.h (struct die_info) <addr_base, ranges_base>: New
methods.
* dwarf2/read.c (lookup_addr_base): Move to die.h.
(lookup_ranges_base): Likewise.
(read_cutu_die_from_dwo, read_full_die_1): Update.
The sibling_die helper function does not seem to add much value,
considering that many other fields of die_info are directly accessed.
So, this removes it.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (read_import_statement, read_file_scope)
(read_type_unit_scope, inherit_abstract_dies, read_func_scope)
(read_lexical_block_scope, read_call_site_scope)
(dwarf2_get_subprogram_pc_bounds, get_scope_pc_bounds)
(handle_struct_member_die, process_structure_scope)
(update_enumeration_type_from_children)
(process_enumeration_scope, read_array_type, read_common_block)
(read_namespace, read_module, read_subroutine_type): Update.
(sibling_die): Remove.
This changes dwarf2_attr_no_follow to be a method on die_info.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (lookup_addr_base, lookup_ranges_base)
(build_type_psymtabs_reader, read_structure_type)
(read_enumeration_type, read_full_die_1): Update.
(dwarf2_attr_no_follow): Move to die.h.
* dwarf2/die.h (struct die_info) <attr>: New method.
This removes dwarf2_cu::base_known, changing base_address to be a
gdb::optional.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (struct dwarf2_cu) <base_known>: Remove.
<base_address>: Now an optional.
(dwarf2_find_base_address, dwarf2_rnglists_process)
(dwarf2_ranges_process, fill_in_loclist_baton)
(dwarf2_symbol_mark_computed): Update.
This moves struct die_info to a new header, dwarf2/die.h.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (struct die_info): Move to die.h.
* dwarf2/die.h: New file.
This moves some more code out of read.c and into line-header.c.
dwarf_decode_line_header is split into two -- the part remaining in
read.c handles interfacing to the dwarf2_cu; while the part in
line-header.c (more or less) purely handles the actual decoding.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/line-header.h (dwarf_decode_line_header): Declare.
* dwarf2/read.c
(dwarf2_statement_list_fits_in_line_number_section_complaint):
Move to line-header.c.
(read_checked_initial_length_and_offset, read_formatted_entries):
Likewise.
(dwarf_decode_line_header): Split into two.
* dwarf2/line-header.c
(dwarf2_statement_list_fits_in_line_number_section_complaint):
Move from read.c.
(read_checked_initial_length_and_offset, read_formatted_entries):
Likewise.
(dwarf_decode_line_header): New function, split from read.c.
This changes read_indirect_line_string to be a method on
dwarf2_per_objfile. This makes it a bit simpler to share between
files.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.h (struct dwarf2_per_objfile) <read_line_string>:
Declare method.
* dwarf2/read.c (read_attribute_value): Update.
(dwarf2_per_objfile::read_line_string): Rename from
read_indirect_line_string.
(read_formatted_entries): Update.
This moves some more code out of dwarf2/read.c, introducing new files
dwarf2/macro.c and dwarf2/macro.h.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (dwarf2_macro_malformed_definition_complaint)
(macro_start_file, consume_improper_spaces)
(parse_macro_definition, skip_form_bytes, skip_unknown_opcode)
(dwarf_parse_macro_header, dwarf_decode_macro_bytes)
(dwarf_decode_macros): Move to macro.c.
* dwarf2/macro.c: New file.
* dwarf2/macro.h: New file.
* Makefile.in (COMMON_SFILES): Add dwarf2/macro.c.
This moves a string-reading function to be a method on
dwarf2_section_info, and then updates the users.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/section.h (struct dwarf2_section_info) <read_string>: New
method.
* dwarf2/section.c: New method. From
read_indirect_string_at_offset_from.
* dwarf2/read.c (mapped_debug_names::namei_to_name): Update.
(read_indirect_string_at_offset_from): Move to section.c.
(read_indirect_string_at_offset): Rewrite.
(read_indirect_line_string_at_offset): Remove.
(read_indirect_string, read_indirect_line_string)
(dwarf_decode_macro_bytes): Update.
This splits dwarf_decode_macros into two overloads -- one that's
suitable for splitting into a separate file, and one that finds the
correct section and should remain in dwarf2/read.c.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (dwarf_decode_macros): Split into two overloads.
This changes dwarf_decode_macro_bytes to accept a buildsym_compunit
rather than a dwarf2_cu. This enables some subsequent changes; and
also makes the function accept a "more specific" parameter.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (macro_start_file): Change "cu" parameter to
"builder".
(dwarf_decode_macro_bytes): Likewise. Add dwarf2_per_objfile
parameter.
(dwarf_decode_macros): Update.
This changes read_indirect_string_from_dwz to be a method on the
dwz_file, and adds a new dwarf2/dwz.c file.
gdb/ChangeLog
2020-03-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (read_attribute_value): Update.
(read_indirect_string_from_dwz): Move to dwz.c; change into
method.
(dwarf_decode_macro_bytes): Update.
* dwarf2/dwz.h (struct dwz_file) <read_string>: Declare method.
* dwarf2/dwz.c: New file.
* Makefile.in (COMMON_SFILES): Add dwz.c.
The DWARF standard appendix E.1 describes techniques that can be used for
compression and deduplication: DIEs can be factored out into a new compilation
unit, and referenced using DW_FORM_ref_addr.
Such a new compilation unit can either use a DW_TAG_compile_unit or
DW_TAG_partial_unit. If a DW_TAG_compile_unit is used, its contents is
evaluated by consumers as though it were an ordinary compilation unit. If a
DW_TAG_partial_unit is used, it's only considered by consumers in the context
of a DW_TAG_imported_unit.
An example of when DW_TAG_partial_unit is required is when the factored out
DIEs are not top-level, f.i. because they were children of a namespace. In
such a case the corresponding DW_TAG_imported_unit will occur as child of the
namespace.
In the case of factoring out DIEs from c++ compilation units, we can factor
out into a new DW_TAG_compile_unit, and no DW_TAG_imported_unit is required.
This begs the question how to interpret a top-level DW_TAG_imported_unit of a
c++ DW_TAG_compile_unit compilation unit. The semantics of
DW_TAG_imported_unit describe that the imported unit logically appears at the
point of the DW_TAG_imported_unit entry. But it's not clear what the effect
should be in this case, since all the imported DIEs are already globally
visible anyway, due to the use of DW_TAG_compile_unit.
So, skip top-level imports of c++ DW_TAG_compile_unit compilation units in
process_imported_unit_die.
Using the cc1 binary from PR23710 comment 1 and setting a breakpoint on do_rpo_vn:
...
$ gdb \
-batch \
-iex "maint set dwarf max-cache-age 316" \
-iex "set language c++" \
-ex "b do_rpo_vn" \
cc1
...
we get a 8.1% reduction in execution time, due to reducing the number of
partial symtabs expanded into full symtabs from 212 to 175.
Build and reg-tested on x86_64-linux.
gdb/ChangeLog:
2020-03-17 Tom de Vries <tdevries@suse.de>
PR gdb/23710
* dwarf2/read.h (struct dwarf2_per_cu_data): Add unit_type and lang
fields.
* dwarf2/read.c (process_psymtab_comp_unit): Initialize unit_type and lang
fields.
(process_imported_unit_die): Skip import of c++ CUs.
PR gdb/25663 points out that dwarf2_name will cache a value in the
bcache and then return a substring. However, this substring return is
only done on the branch that caches the value -- so if the function is
called twice with the same arguments, it will return different values.
This patch fixes this problem.
This area is strange. We cache the entire demangled string, but only
return the suffix. I looked at caching just the suffix, but it turns
out that anonymous_struct_prefix assumes that the entire string is
stored. Also weird is that this code is demangling the linkage name
and then storing the demangled form back into the linkage name
attribute -- that seems bad, because what if some code wants to find
the actual linkage name?
Fixing these issues was non-trivial, though; and in the meantime this
patch seems like an improvement. Regression tested on x86-64
Fedora 30.
gdb/ChangeLog
2020-03-16 Tom Tromey <tromey@adacore.com>
PR gdb/25663:
* dwarf2/read.c (dwarf2_name): Strip leading namespaces after
putting value into bcache.
This commit brings support for the DWARF line table is_stmt field to
GDB. The is_stmt field is used by the compiler when a single source
line is split into multiple assembler instructions, especially if the
assembler instructions are interleaved with instruction from other
source lines.
The compiler will set the is_stmt flag false from some instructions
from the source lines, these instructions are not a good place to
insert a breakpoint in order to stop at the source line.
Instructions which are marked with the is_stmt flag true are a good
place to insert a breakpoint for that source line.
Currently GDB ignores all instructions for which is_stmt is false.
This is fine in a lot of cases, however, there are some cases where
this means the debug experience is not as good as it could be.
Consider stopping at a random instruction, currently this instruction
will be attributed to the last line table entry before this point for
which is_stmt was true - as these are the only line table entries that
GDB tracks. This can easily be incorrect in code with even a low
level of optimisation.
With is_stmt tracking in place, when stopping at a random instruction
we now attribute the instruction back to the real source line, even
when is_stmt is false for that instruction in the line table.
When inserting breakpoints we still select line table entries for
which is_stmt is true, so the breakpoint placing behaviour should not
change.
When stepping though code (at the line level, not the instruction
level) we will still stop at instruction where is_stmt is true, I
think this is more likely to be the desired behaviour.
Instruction stepping is, of course, unchanged, stepping one
instruction at a time, but we should now report more accurate line
table information with each instruction step.
The original motivation for this work was a patch posted by Bernd
here:
https://sourceware.org/ml/gdb-patches/2019-11/msg00792.html
As part of that thread it was suggested that many issues would be
resolved if GDB supported line table views, this isn't something I've
attempted in this patch, though reading the spec, it seems like this
would be a useful feature to support in GDB in the future. The spec
is here:
http://dwarfstd.org/ShowIssue.php?issue=170427.1
And Bernd gives a brief description of the benefits here:
https://sourceware.org/ml/gdb-patches/2020-01/msg00147.html
With that all said, I think that there is benefit to having proper
is_stmt support regardless of whether we have views support, so I
think we should consider getting this in first, and then building view
support on top of this.
The gdb.cp/step-and-next-inline.exp test is based off a test proposed
by Bernd Edlinger in this message:
https://sourceware.org/ml/gdb-patches/2019-12/msg00842.html
gdb/ChangeLog:
* buildsym-legacy.c (record_line): Pass extra parameter to
record_line.
* buildsym.c (buildsym_compunit::record_line): Take an extra
parameter, reduce duplication in the line table, and record the
is_stmt flag in the line table.
* buildsym.h (buildsym_compunit::record_line): Add extra
parameter.
* disasm.c (do_mixed_source_and_assembly_deprecated): Ignore
non-statement lines.
* dwarf2/read.c (dwarf_record_line_1): Add extra parameter, pass
this to the symtab builder.
(dwarf_finish_line): Pass extra parameter to dwarf_record_line_1.
(lnp_state_machine::record_line): Pass a suitable is_stmt flag
through to dwarf_record_line_1.
* infrun.c (process_event_stop_test): When stepping, don't stop at
a non-statement instruction, and only refresh the step info when
we land in the middle of a line's range. Also add an extra
comment.
* jit.c (jit_symtab_line_mapping_add_impl): Initialise is_stmt
field.
* record-btrace.c (btrace_find_line_range): Only record lines
marked as is-statement.
* stack.c (frame_show_address): Show the frame address if we are
in a non-statement sal.
* symmisc.c (dump_symtab_1): Print the is_stmt flag.
(maintenance_print_one_line_table): Print a header for the is_stmt
column, and include is_stmt information in the output.
* symtab.c (find_pc_sect_line): Find lines marked as statements in
preference to non-statements.
(find_pcs_for_symtab_line): Prefer is-statement entries.
(find_line_common): Likewise.
* symtab.h (struct linetable_entry): Add is_stmt field.
(struct symtab_and_line): Likewise.
* xcoffread.c (arrange_linetable): Initialise is_stmt field when
arranging the line table.
gdb/testsuite/ChangeLog:
* gdb.cp/step-and-next-inline.cc: New file.
* gdb.cp/step-and-next-inline.exp: New file.
* gdb.cp/step-and-next-inline.h: New file.
* gdb.dwarf2/dw2-is-stmt.c: New file.
* gdb.dwarf2/dw2-is-stmt.exp: New file.
* gdb.dwarf2/dw2-is-stmt-2.c: New file.
* gdb.dwarf2/dw2-is-stmt-2.exp: New file.
* gdb.dwarf2/dw2-ranges-base.exp: Update line table pattern.
Gcc supports an option -feliminate-dwarf2-dups (up until gcc-7, removed in
gcc-8).
When running tests with target board unix/-feliminate-dwarf2-dups, we run
into:
...
(gdb) PASS: gdb.ada/arraydim.exp: print m'length(3)
ptype global_3dim_for_gdb_testing^M
type = array (Unexpected type in ada_discrete_type_low_bound.^M
(gdb) FAIL: gdb.ada/arraydim.exp: ptype global_3dim_for_gdb_testing
...
The DWARF for the variable global_3dim_for_gdb_testing looks as follows:
...
<0><824>: Abbrev Number: 1 (DW_TAG_compile_unit)
<825> DW_AT_name : src/gdb/testsuite/gdb.ada/arraydim/inc.c
<1><832>: Abbrev Number: 2 (DW_TAG_array_type)
<833> DW_AT_type : <0x874>
<2><837>: Abbrev Number: 3 (DW_TAG_subrange_type)
<838> DW_AT_type : <0x84a>
<83c> DW_AT_upper_bound : 0
<2><83d>: Abbrev Number: 3 (DW_TAG_subrange_type)
<83e> DW_AT_type : <0x84a>
<842> DW_AT_upper_bound : 1
<2><843>: Abbrev Number: 3 (DW_TAG_subrange_type)
<844> DW_AT_type : <0x84a>
<848> DW_AT_upper_bound : 2
<2><849>: Abbrev Number: 0
<1><84a>: Abbrev Number: 4 (DW_TAG_typedef)
<84b> DW_AT_type : <0x86d>
<1><84f>: Abbrev Number: 0
<0><85b>: Abbrev Number: 5 (DW_TAG_compile_unit)
<861> DW_AT_name : src/gdb/testsuite/gdb.ada/arraydim/inc.c
<1><86d>: Abbrev Number: 6 (DW_TAG_base_type)
<86e> DW_AT_byte_size : 8
<86f> DW_AT_encoding : 7 (unsigned)
<870> DW_AT_name : long unsigned int
<1><874>: Abbrev Number: 7 (DW_TAG_base_type)
<875> DW_AT_byte_size : 4
<876> DW_AT_encoding : 5 (signed)
<877> DW_AT_name : int
<1><87b>: Abbrev Number: 8 (DW_TAG_variable)
<87c> DW_AT_name : global_3dim_for_gdb_testing
<882> DW_AT_type : <0x832>
<886> DW_AT_external : 1
...
The DWARF contains an anonymous typedef at 0x84a, referring to 0x86d.
Strictly speaking, the anonymous typedef is illegal DWARF, because a
DW_TAG_typedef is defined to have an DW_AT_name attribute containing the name
of the typedef as it appears in the source program.
The DWARF reading code creates a corresponding type for this typedef, which
goes on to confuse the code handling arrays.
Rather than trying to support the type representing this anonymous typedef in
all the locations where it causes problems, fix this by treating the anonymous
typedef as a forwarder DIE in the DWARF reader.
Tested on x86_64-linux, with target boards unix and
unix/-feliminate-dwarf2-dups.
This fixes ~85 failures for unix/-feliminate-dwarf2-dups.
gdb/ChangeLog:
2020-03-07 Tom de Vries <tdevries@suse.de>
* dwarf2/read.c (read_typedef): Treat anonymous typedef as forwarder
DIE.
This introduces a string cache on the per-BFD object, replacing the
macro and filename caches. Both of these caches just store strings,
so this consolidation by itself saves a little memory (about the size
of a bcache per objfile).
Then this patch switches some allocations on the objfile obstack to
use this bcache instead. This saves more space; and turns out to be a
bit faster as well.
Here are the before and after "maint time" + "maint space" results of
"file ./gdb":
Command execution time: 4.664021 (cpu), 4.728518 (wall)
Space used: 39190528 (+29212672 for this command)
Command execution time: 4.216209 (cpu), 4.107023 (wall)
Space used: 36667392 (+26689536 for this command)
The main interface to the string cache is a new pair of overloaded
methods, objfile::intern.
gdb/ChangeLog
2020-03-04 Tom Tromey <tom@tromey.com>
* symmisc.c (print_symbol_bcache_statistics)
(print_objfile_statistics): Update.
* symfile.c (allocate_symtab): Use intern.
* psymtab.c (partial_symtab::partial_symtab): Use intern.
* objfiles.h (struct objfile_per_bfd_storage) <filename_cache,
macro_cache>: Remove.
<string_cache>: New member.
(struct objfile) <intern>: New methods.
* elfread.c (elf_symtab_read): Use intern.
* dwarf2/read.c (fixup_go_packaging): Intern package name.
(dwarf2_compute_name, dwarf2_physname)
(create_dwo_unit_in_dwp_v1, create_dwo_unit_in_dwp_v2): Intern
names.
(guess_partial_die_structure_name): Update.
(partial_die_info::fixup): Intern name.
(dwarf2_canonicalize_name): Change parameter to objfile. Intern
name.
(dwarf2_name): Intern name. Update.
* buildsym.c (buildsym_compunit::get_macro_table): Use
string_cache.
Running anything with the fission.exp board fails since commit
c0ab21c22b ("Replace init_cutu_and_read_dies with a class").
GDB crashes while reading the DWARF info. cu is NULL in
read_signatured_type:
Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
0x000055555780663e in read_signatured_type
sig_type=0x6210000c3600) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:22782
22782 gdb_assert (cu->die_hash == NULL);
(top-gdb) bt
#0 0x000055555780663e in read_signatured_type (sig_type=0x6210000c3600) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:22782
#1 0x00005555578062dd in load_full_type_unit (per_cu=0x6210000c3600) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:22758
#2 0x00005555577c5fb7 in queue_and_load_dwo_tu (slot=0x60600007fc00, info=0x6210000c34e0) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:12674
#3 0x0000555559934232 in htab_traverse_noresize (htab=0x60b000063670, callback=0x5555577c5e61 <queue_and_load_dwo_tu(void**, void*)>, info=0x6210000c34e0)
at /home/simark/src/binutils-gdb/libiberty/hashtab.c:775
#4 0x00005555577c6252 in queue_and_load_all_dwo_tus (per_cu=0x6210000c34e0) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:12701
#5 0x000055555777ebd8 in dw2_do_instantiate_symtab (per_cu=0x6210000c34e0, skip_partial=false) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:2371
#6 0x000055555777eea2 in dw2_instantiate_symtab (per_cu=0x6210000c34e0, skip_partial=false) at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:2395
#7 0x0000555557786ab6 in dw2_lookup_symbol (objfile=0x614000007240, block_index=GLOBAL_BLOCK, name=0x602000025310 "main", domain=VAR_DOMAIN)
at /home/simark/src/binutils-gdb/gdb/dwarf2/read.c:3539
After creating the reader object, the reader.cu field should not be
NULL. By checking the commit previous to the faulty one mentioned
above, I noticed that the cu field is normally set by
init_cu_die_reader, called from read_cutu_die_from_dwo, itself called
from cutu_reader::init_tu_and_read_dwo_dies, itself called from
cutu_reader's constructor.
However, cutu_reader::init_tu_and_read_dwo_dies calls
read_cutu_die_from_dwo, passing a pointer to a local `die_reader_specs`
variable. So it's the `cu` field of that object that gets set.
cutu_reader itself is a `die_reader_specs` (it inherits from it), and
the intention was most likely to pass `this` to read_cutu_die_from_dwo.
This way, the fields of the cutu_reader object, which
read_signatured_type will use, are set.
With this, I am able to use:
make check RUNTESTFLAGS='--target_board=fission'
and it looks much better. There are still some failures to be
investigated, but that's the usual state of the testsuite.
gdb/ChangeLog:
* dwarf2/read.c (cutu_reader::init_tu_and_read_dwo_dies): Remove
reader variable, pass `this` to read_cutu_die_from_dwo.
I noticed that there's no real reason to have field_info::nfields in
the DWARF reader. It simply mirrors information that is already
available. This patch removes it, in favor of a convenience method.
gdb/ChangeLog
2020-02-27 Tom Tromey <tromey@adacore.com>
* dwarf2/read.c (struct field_info) <nfields>: Now a method, not a
member.
(dwarf2_add_field): Don't update nfields.
(dwarf2_attach_fields_to_type, process_structure_scope): Update.
Include files are represented by a partial symtab, but don't expand to
anything. From dwarf2_psymtab::expand_psymtab:
if (per_cu == NULL)
{
/* It's an include file, no symbols to read for it.
Everything is in the parent symtab. */
readin = true;
return;
}
This patch introduces a new specialization of partial_symtab to handle
this case. In addition to being slightly smaller, I believe an
include file is the only situation where a DWARF psymtab can result in
a null compunit_symtab. This adds an assert to that effect as well.
This change will simplify one of the psymtab sharing patches.
gdb/ChangeLog
2020-02-26 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (struct dwarf2_include_psymtab): New.
(dwarf2_create_include_psymtab): Use dwarf2_include_psymtab.
(dwarf2_psymtab::expand_psymtab, dwarf2_psymtab::readin_p)
(dwarf2_psymtab::get_compunit_symtab): Remove null checks for
per_cu_data.
debuginfod is a lightweight web service that indexes ELF/DWARF debugging
resources by build-id and serves them over HTTP.
This patch enables GDB to query debuginfod servers for separate debug
files and source code when it is otherwise not able to find them.
GDB can be built with debuginfod using the --with-debuginfod configure
option.
This requires that libdebuginfod be installed and found at configure time.
debuginfod is packaged with elfutils, starting with version 0.178.
For more information see https://sourceware.org/elfutils/.
Tested on x86_64 Fedora 31.
gdb/ChangeLog:
2020-02-26 Aaron Merey <amerey@redhat.com>
* Makefile.in: Handle optional debuginfod support.
* NEWS: Update.
* README: Add --with-debuginfod summary.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac: Handle optional debuginfod support.
* debuginfod-support.c: debuginfod helper functions.
* debuginfod-support.h: Ditto.
* doc/gdb.texinfo: Add --with-debuginfod to configure options
summary.
* dwarf2/read.c (dwarf2_get_dwz_file): Query debuginfod servers
when a dwz file cannot be found.
* elfread.c (elf_symfile_read): Query debuginfod servers when a
debuginfo file cannot be found.
* source.c (open_source_file): Query debuginfod servers when a
source file cannot be found.
* top.c (print_gdb_configuration): Include
--{with,without}-debuginfod in the output.
gdb/testsuite/ChangeLog:
2020-02-26 Aaron Merey <amerey@redhat.com>
* gdb.debuginfod: New directory for debuginfod tests.
* gdb.debuginfod/main.c: New test file.
* gdb.debuginfod/fetch_src_and_symbols.exp: New tests.
All these functions have their implementations in dwarf2/read.c, so move
their declarations to dwarf2/read.h. Move the doc to the header, at the
same time.
gdb/ChangeLog:
* dwarf2/loc.h (dwarf2_fetch_die_loc_sect_off,
dwarf2_fetch_die_loc_cu_off, dwarf2_fetch_constant_bytes,
dwarf2_fetch_die_type_sect_off): Move to...
* dwarf2/read.h (dwarf2_fetch_die_loc_sect_off,
dwarf2_fetch_die_loc_cu_off, dwarf2_fetch_constant_bytes,
dwarf2_fetch_die_type_sect_off): ... here.
* dwarf2/read.c (dwarf2_fetch_die_loc_sect_off,
dwarf2_fetch_die_loc_cu_off, dwarf2_fetch_constant_bytes,
dwarf2_fetch_die_type_sect_off): Move doc to header file.
The implementation is in dwarf2/read.c, so the declaration belongs in
dwarf2/read.h. Also, move the documentation there.
gdb/ChangeLog:
* dwarf2/loc.h (dwarf2_read_addr_index): Move...
* dwarf2/read.h (dwarf2_read_addr_index): ... here.
* dwarf2/read.c (dwarf2_read_addr_index): Move doc to header.
Since its implementation is in dwarf2/read.c, its declaration belongs in
dwarf2/read.h. Move the documentation to the .h at the same time.
gdb/ChangeLog:
* loc.h (dwarf2_get_die_type): Move to...
* read.h (dwarf2_get_die_type): ... here.
* read.c (dwarf2_get_die_type): Move doc to header.
I noticed that setup_type_unit_groups leaks the symtab vector -- it
allocates this with XNEWVEC, but from what I can tell, nothing frees
it. This patch changes it to use XOBNEWVEC.
Also, the type_unit_unshareable::num_symtabs member is assigned but
never read. So, this removes it.
gdb/ChangeLog
2020-02-24 Tom Tromey <tom@tromey.com>
* dwarf2/read.h (struct type_unit_unshareable) <num_symtabs>:
Remove.
* dwarf2/read.c (dwarf2_cu::setup_type_unit_groups): Use
XOBNEWVEC.
This converts the IS_TYPE_UNIT_GROUP to a method on
dwarf2_per_cu_data.
gdb/ChangeLog
2020-02-24 Tom Tromey <tom@tromey.com>
* dwarf2/read.h (struct dwarf2_per_cu_data) <type_unit_group_p>:
New method.
* dwarf2/read.c (IS_TYPE_UNIT_GROUP): Remove.
(dw2_do_instantiate_symtab, dw2_get_file_names)
(build_type_psymtab_dependencies, load_full_type_unit): Update.
This simplifies the setting and clearing of reading_partial_symbols,
by using scoped_restore in the function that reads partial symbols.
gdb/ChangeLog
2020-02-24 Tom Tromey <tom@tromey.com>
* dwarf2read.c (dwarf2_build_psymtabs_hard): Use
make_scoped_restore.
(dwarf2_psymtab::read_symtab): Don't clear
reading_partial_symbols.
I am sending this patch on behalf of kmoy@google.com, who discovered the bug
and wrote the fix.
gdb_demangle can return null for strings that don't properly demangle. The null
check was mistakenly removed in commit 43816ebc33. Without this check, GDB
aborts when loading symbols from some binaries.
gdb/ChangeLog
2020-02-21 Ali Tamur <tamur@google.com>
* dwarf2/read.c (dwarf2_name): Add null check.
dwarf2_find_containing_comp_unit has this in its binary search:
if (mid_cu->is_dwz > offset_in_dwz
|| (mid_cu->is_dwz == offset_in_dwz
&& mid_cu->sect_off + mid_cu->length >= sect_off))
high = mid;
The intent here is to determine whether SECT_OFF appears in or before
MID_CU.
I believe this has an off-by-one error, and that the check should use
">" rather than ">=". If the two side are equal, then SECT_OFF
actually appears at the start of the next CU.
I've had this patch kicking around for ages but I forget how I found
the problem.
gdb/ChangeLog
2020-02-20 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (dwarf2_find_containing_comp_unit): Use ">", not
">=", in binary search.
(dwarf2_find_containing_comp_unit): New overload.
(run_test): New self-test.
(_initialize_dwarf2_read): Register new test.
This is a simple cleanup. These functions used to use the objfile's
obstack for allocation in the hash tables, but they don't anymore.
Remove the unnecessary objfile parameters, which in turn allows removing
some local variables.
gdb/ChangeLog:
* dwarf2/read.c (allocate_signatured_type_table,
allocate_dwo_unit_table, allocate_type_unit_groups_table,
allocate_dwo_file_hash_table, allocate_dwp_loaded_cutus_table):
Remove objfile parameter, update all callers.
I have come across some uses cases where it would be desirable to treat
an enum that has duplicate values as a "flag enum". For example, this
one here [1]:
enum membarrier_cmd {
MEMBARRIER_CMD_QUERY = 0,
MEMBARRIER_CMD_GLOBAL = (1 << 0),
MEMBARRIER_CMD_GLOBAL_EXPEDITED = (1 << 1),
MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED = (1 << 2),
MEMBARRIER_CMD_PRIVATE_EXPEDITED = (1 << 3),
MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED = (1 << 4),
MEMBARRIER_CMD_PRIVATE_EXPEDITED_SYNC_CORE = (1 << 5),
MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_SYNC_CORE = (1 << 6),
/* Alias for header backward compatibility. */
MEMBARRIER_CMD_SHARED = MEMBARRIER_CMD_GLOBAL,
};
The last enumerator is kept for backwards compatibility. Without this
patch, this enumeration wouldn't be considered a flag enum, because two
enumerators collide. With this patch, it would be considered a flag
enum, and the value 3 would be printed as:
MEMBARRIER_CMD_GLOBAL | MEMBARRIER_CMD_GLOBAL_EXPEDITED
Although if people prefer, we could display both MEMBARRIER_CMD_GLOBAL
and MEMBARRIER_CMD_SHARED in the result. It wouldn't be wrong, and
could perhaps be useful in case a bit may have multiple meanings
(depending on some other bit value).
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/uapi/linux/membarrier.h?id=0bf999f9c5e74c7ecf9dafb527146601e5c848b9#n125
gdb/ChangeLog:
* dwarf2/read.c (update_enumeration_type_from_children): Allow
flag enums to contain duplicate enumerators.
* valprint.c (generic_val_print_enum_1): Update comment.
gdb/testsuite/ChangeLog:
* gdb.base/printcmds.c (enum flag_enum): Add FE_TWO_LEGACY
enumerator.
GDB has this feature where if an enum looks like it is meant to
represent binary flags, it will present the values of that type as a
bitwise OR of the flags that are set in the value.
The original motivation for this patch is to fix this behavior:
enum hello { AAA = 0x1, BBB = 0xf0 };
(gdb) p (enum hello) 0x11
$1 = (AAA | BBB)
This is wrong because the bits set in BBB (0xf0) are not all set in the
value 0x11, but GDB presents it as if they all were.
I think that enumerations with enumerators that have more than one bit
set should simply not qualify as "flag enum", as far as this
heuristic is concerned. I'm not sure what it means to have flags of
more than one bit. So this is what this patch implements.
I have added an assert in generic_val_print_enum_1 to make sure the flag
enum types respect that, in case they are used by other debug info
readers, in the future.
I've enhanced the gdb.base/printcmds.exp test to cover this case. I've
also added tests for printing flag enums with value 0, both when the
enumeration has and doesn't have an enumerator for value 0.
gdb/ChangeLog:
* dwarf2/read.c: Include "count-one-bits.h".
(update_enumeration_type_from_children): If an enumerator has
multiple bits set, don't treat the enumeration as a "flag enum".
* valprint.c (generic_val_print_enum_1): Assert that enumerators
of flag enums have 0 or 1 bit set.
gdb/testsuite/ChangeLog:
* gdb.base/printcmds.c (enum flag_enum): Prefix enumerators with
FE_, add FE_NONE.
(three): Update.
(enum flag_enum_without_zero): New enum.
(flag_enum_without_zero): New variable.
(enum not_flag_enum): New enum.
(three_not_flag): New variable.
* gdb.base/printcmds.exp (test_artificial_arrays): Update.
(test_print_enums): Add more tests for printing flag enums.
I see for some program at gdb startup:
...
Samples: 102K of event 'cycles:pu', Event count (approx.): 91710925103
Overhead Command Shared Object Symbol
15.21% gdb gdb [.]
lnp_state_machine::handle_special
...
where the divisions are the places we stall. The following
micro-optimizes things but it smells like m_line_header->line_range
is constant, likewise probably m_line_header->maximum_ops_per_instruction
so eventually the divisions could be avoided completely with some
lookup table.
Well. Micro-optimizing with this patch improves things
(don't expect [load] CSE over the gdbarch_adjust_dwarf2_line call).
Build and reg-tested on x86_64-linux.
gdb/ChangeLog:
2020-02-14 Richard Biener <rguenther@suse.de>
* dwarf2/read.c (lnp_state_machine::handle_special_opcode): Apply CSE
on expression with division operators.
When passed in CXXFLAGS, -Wstrict-null-sentinel triggers twice in a
gdb/gdbserver build.
Fix the two occurrences.
Build and reg-tested on x86_64-linux.
gdb/ChangeLog:
2020-02-10 Tom de Vries <tdevries@suse.de>
* dwarf2/read.c (process_psymtab_comp_unit_reader): Cast concat NULL
sentinel to char *.
gdbsupport/ChangeLog:
2020-02-10 Tom de Vries <tdevries@suse.de>
* environ.c (gdb_environ::set): Cast concat NULL sentinel to char *.
Say we're debugging a test-case with CUs with name "<artificial>", meaning
not originating from a single file compilation, and use the verbose setting:
...
$ gdb -iex "set verbose on" -batch cc1
Reading symbols from cc1...
Reading in symbols for <artificial>... \
and /tmp/trunk/gcc/attribs.c... \
...
and /tmp/trunk/gcc/tree-ssa-reassoc.c... \
done.
...
From the "/tmp/trunk/gcc/attribs.c" message, it's clear which CU is loaded. But
that's not the case for the "<artificial>" message.
The message uses the filename field of struct partial_symtab, which is
documented like this:
...
/* Name of the source file which this partial_symtab defines,
or if the psymtab is anonymous then a descriptive name for
debugging purposes, or "". It must not be NULL. */
...
So, fix this by setting the filename field to a more descriptive name than
"<artificial>", by appending the CU offset.
This way, we print instead:
...
$ gdb -iex "set verbose on" -batch cc1
Reading symbols from cc1...
Reading in symbols for <artificial>@0x41146d9 \
and /tmp/trunk/gcc/attribs.c... \
... \
and /tmp/trunk/gcc/tree-ssa-reassoc.c... \
done.
...
Build and reg-tested on x86_64-linux.
gdb/ChangeLog:
2020-02-09 Tom de Vries <tdevries@suse.de>
* dwarf2read.c (process_psymtab_comp_unit_reader): Append CU offset to
filename if it matches "<artificial>".
cutu_reader has a "keep" parameter, which is used to decide what to do
with a new CU when the reader is destroyed. Most code does not try to
preserve the CU, so this patch removes this parameter and instead adds
a new method that users can call to preserve the CU on the chain.
gdb/ChangeLog
2020-02-08 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (class cutu_reader) <cutu_reader,
init_tu_and_read_dwo_dies>: Remove "keep" parameter.
<keep>: Declare method.
<m_keep>: Remove member.
<~cutu_reader>: Remove.
(cutu_reader::init_tu_and_read_dwo_dies): Update.
(cutu_reader::cutu_reader): Update.
(cutu_reader::keep): Rename from ~cutu_reader.
(process_psymtab_comp_unit, build_type_psymtabs_1)
(process_skeletonless_type_unit, load_partial_comp_unit)
(load_full_comp_unit, dwarf2_read_addr_index)
(read_signatured_type): Update.
Change-Id: I859b1c64313569d76d46317c14e9b077ebc3a27b
