This commit adds a mechanism for GDB to detect the linetable opcode
DW_LNS_set_epilogue_begin. This opcode is set by compilers to indicate
that a certain instruction marks the point where the frame is destroyed.
While the standard allows for multiple points marked with epilogue_begin
in the same function, for performance reasons, the function that
searches for the epilogue address will only find the last address that
sets this flag for a given block.
This commit also changes amd64_stack_frame_destroyed_p_1 to attempt to
use the epilogue begin directly, and only if an epilogue can't be found
will it attempt heuristics based on the current instruction.
Finally, this commit also changes the dwarf assembler to be able to emit
epilogue-begin instructions, to make it easier to test this patch
Approved-By: Tom Tromey <tom@tromey.com>
Fortran provides additional entry points for subroutines and functions.
These entry points may use only a subset (or a different set) of the
parameters of the original subroutine. The entry points may be described
via the DWARF tag DW_TAG_entry_point.
This commit adds support for parsing the DW_TAG_entry_point DWARF tag.
Currently, between ifx/ifort/gfortran, only ifort is actually emitting
this tag. Both, ifx and gfortran use the DW_TAG_subprogram tag as
workaround/alternative. Thus, this patch really only adds more ifort
support. Even so, some of the attached tests still fail for ifort, due
to some wrong line info generated for the entry points in ifort.
After this patch it is possible to set a breakpoint in gdb with the
ifort compiled example at the entry points 'foo' and 'foobar', which was not
possible before.
As gcc and ifx do not emit the tag I also added a test to gdb.dwarf2
which uses some underlying c compiled code and adds some Fortran style DWARF
to it emitting the DW_TAG_entry_point. Before this patch it was not
possible to actually define breakpoint at the entry point tags.
For gfortran there actually exists a bug on bugzilla, asking for the use
of DW_TAG_entry_point over DW_TAG_subprogram:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=37134
This patch was originally posted here
https://sourceware.org/legacy-ml/gdb-patches/2017-07/msg00317.html
but its review/pinging got lost after a while. I reworked it to fit the
current GDB.
Co-authored-by: Bernhard Heckel <bernhard.heckel@intel.com>
Co-authored-by: Tim Wiederhake <tim.wiederhake@intel.com>
Approved-by: Tom Tromey <tom@tromey.com>
In dwarf2_get_pc_bounds we were writing unchecked to *lowpc. This
commit adds a gdb_assert to first check that lowpc != nullptr.
Approved-by: Tom Tromey <tom@tromey.com>
This commit is in preparation of the next commit. There, we will add
a second variation to retrieve the pc bounds for DIEs tagged with
DW_TAG_entry_point. Instead of dwarf_get_pc_bounds_ranges_or_highlow_pc
we will call a separate method for entry points. As the validity checks
at the endo f dwarf2_get_pc_bounds are the same for both variants,
we introduced the new dwarf_get_pc_bounds_ranges_or_highlow_pc method,
outsourcing part of dwarf2_get_pc_bounds.
This commit should have no functional impact on GDB.
Approved-by: Tom Tromey <tom@tromey.com>
This reverts commit 1c04f72368 ("[gdb/symtab] Fix assert in set_length"), due
to a regression reported in PR29572, and implements a different fix for PR29453.
The fix is to not use the CU table in a .debug_names section to construct
all_units, but instead use create_all_units, and then verify the CU
table from .debug_names. This also fixes PR25969, so remove the KFAIL.
Approved-By: Tom Tromey <tom@tromey.com>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29572
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=25969
compute_delayed_physnames does this:
size_t len = strlen (physname);
...
if (physname[len] == ')') /* shortcut */
break;
However, physname[len] will always be \0.
This patch changes it to the correct len-1.
This changes gdb to use the C++17 [[fallthrough]] attribute rather
than special comments.
This was mostly done by script, but I neglected a few spellings and so
also fixed it up by hand.
I suspect this fixes the bug mentioned below, by switching to a
standard approach that, presumably, clang supports.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23159
Approved-By: John Baldwin <jhb@FreeBSD.org>
Approved-By: Luis Machado <luis.machado@arm.com>
Approved-By: Pedro Alves <pedro@palves.net>
This changes nested types and member functions to use the new
'accessibility' enum, rather than separate private/protected flags.
This is done for consistency, but it also lets us simplify some other
code in the next patch.
Acked-By: Simon Marchi <simon.marchi@efficios.com>
Reviewed-by: Keith Seitz <keiths@redhat.com>
This removes some byte vectors from cplus_struct_type, moving the
information into bitfields in holes in struct field.
A new 'enum accessibility' is added to hold some of this information.
A similar enum is removed from c-varobj.c.
Note that the stabs reader treats "ignored" as an accessibility.
However, the stabs texinfo documents this as a public field that is
optimized out -- unfortunately nobody has updated the stabs reader to
use the better value-based optimized-out machinery. I looked and
apparently gcc never emitted this visibility value, so whatever
compiler generated this stab is unknown. I left a comment in
gdbtypes.h to this effect.
Acked-By: Simon Marchi <simon.marchi@efficios.com>
Reviewed-by: Keith Seitz <keiths@redhat.com>
This patch removes all uses of to_string(const std::string_view&) and
use the std::string ctor or implicit conversion from std::string_view to
std::string instead.
A later patch will remove this gdb::to_string while removing
gdbsupport/gdb_string_view.h.
Change-Id: I877cde557a0727be7b0435107e3c7a2aac165895
Approved-By: Tom Tromey <tom@tromey.com>
Approved-By: Pedro Alves <pedro@palves.net>
Given that GDB now requires a C++17, replace all uses of
gdb::string_view with std::string_view.
This change has mostly been done automatically:
- gdb::string_view -> std::string_view
- #include "gdbsupport/gdb_string_view.h" -> #include <string_view>
One things which got brought up during review is that gdb::stging_view
does support being built from "nullptr" while std::sting_view does not.
Two places are manually adjusted to account for this difference:
gdb/tui/tui-io.c:tui_getc_1 and
gdbsupport/format.h:format_piece::format_piece.
The above automatic change transformed
"gdb::to_string (const gdb::string_view &)" into
"gdb::to_string (const std::string_view &)". The various direct users
of this function are now explicitly including
"gdbsupport/gdb_string_view.h". A later patch will remove the users of
gdb::to_string.
The implementation and tests of gdb::string_view are unchanged, they will
be removed in a following patch.
Change-Id: Ibb806a7e9c79eb16a55c87c6e41ad396fecf0207
Approved-By: Tom Tromey <tom@tromey.com>
Approved-By: Pedro Alves <pedro@palves.net>
Since GDB now requires C++17, we don't need the internally maintained
gdb::optional implementation. This patch does the following replacing:
- gdb::optional -> std::optional
- gdb::in_place -> std::in_place
- #include "gdbsupport/gdb_optional.h" -> #include <optional>
This change has mostly been done automatically. One exception is
gdbsupport/thread-pool.* which did not use the gdb:: prefix as it
already lives in the gdb namespace.
Change-Id: I19a92fa03e89637bab136c72e34fd351524f65e9
Approved-By: Tom Tromey <tom@tromey.com>
Approved-By: Pedro Alves <pedro@palves.net>
gdb::make_unique is a wrapper around std::make_unique when compiled with
C++17. Now that C++17 is required, use std::make_unique directly in the
codebase, and remove gdb::make_unique.
Change-Id: I80b615e46e4b7c097f09d78e579a9bdce00254ab
Approved-By: Tom Tromey <tom@tromey.com>
Approved-By: Pedro Alves <pedro@palves.net
dwarf2/read.c:new_symbol has some extra braces in a couple of 'case's.
These read weirdly to me, and since they aren't necessary, this patch
removes the braces and reindents the bodies. Tested by rebuilding.
While working on background reading of DWARF, I came across the
DWZ-reading code. This code can query the user (via the debuginfod
support) -- something that cannot be done off the main thread.
Looking into it, I realized that this code can be run much earlier,
avoiding this problem. Digging a bit deeper, I also found a
discrepancy here between how the DWARF reader works in "readnow" mode
as compared to the normal modes.
This patch cleans this up by trying to read the DWZ file earlier, and
also by having the DWARF reader convert any exception here into a
warning. This unifies the various cases, but also makes it so that
errors do not prevent gdb from continuing on to the extent possible.
Regression tested on x86-64 Fedora 38.
When running test-case gdb.tui/tui-layout-asm-short-prog.exp on AlmaLinux 9.2
ppc64le, I run into:
...
FAIL: gdb.tui/tui-layout-asm-short-prog.exp: check asm box contents
...
The problem is that we get:
...
7 [ No Assembly Available ]
...
because tui_get_begin_asm_address doesn't succeed.
In more detail, tui_get_begin_asm_address calls:
...
find_line_pc (sal.symtab, sal.line, &addr);
...
with:
...
(gdb) p *sal.symtab
$5 = {next = 0x130393c0, m_compunit = 0x130392f0, m_linetable = 0x0,
filename = "tui-layout-asm-short-prog.S",
filename_for_id = "$gdb/build/gdb/testsuite/tui-layout-asm-short-prog.S",
m_language = language_asm, fullname = 0x0}
(gdb) p sal.line
$6 = 1
...
The problem is the filename_for_id which is the source file prefixed with the
compilation dir rather than the source dir.
This is due to faulty debug info generated by gas, PR28629:
...
<1a> DW_AT_name : tui-layout-asm-short-prog.S
<1e> DW_AT_comp_dir : $gdb/build/gdb/testsuite
<22> DW_AT_producer : GNU AS 2.35.2
...
The DW_AT_name is relative, and it's relative to the DW_AT_comp_dir entry,
making the effective name $gdb/build/gdb/testsuite/tui-layout-asm-short-prog.S.
The bug is fixed starting version 2.38, where we get instead:
...
<1a> DW_AT_name :
$gdb/src/gdb/testsuite/gdb.tui/tui-layout-asm-short-prog.S
<1e> DW_AT_comp_dir : $gdb/build/gdb/testsuite
<22> DW_AT_producer : GNU AS 2.38
...
Work around the faulty debug info by constructing the filename_for_id using
the second directory from the directory table in the .debug_line header:
...
The Directory Table (offset 0x22, lines 2, columns 1):
Entry Name
0 $gdb/build/gdb/testsuite
1 $gdb/src/gdb/testsuite/gdb.tui
...
Note that the used gas contains a backport of commit 3417bfca67 ("GAS:
DWARF-5: Ensure that the 0'th entry in the directory table contains the
current working directory."), because directory 0 is correct. With the
unpatched 2.35.2 release the directory 0 entry is incorrect: it's a copy of
entry 1.
Add a dwarf assembly test-case that reflects the debug info as generated by
unpatched gas 2.35.2.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
In the interest of shrinking dwarf2/read.c a little more, this patch
moves the code that deciphers .debug_aranges into a new file.
Reviewed-By: Guinevere Larsen <blarsen@redhat.com>
While working on background DWARF reading, I found a race case that I
tracked down to the handling of the .debug_aranges section. Currently
the section data is only read in after the CUs have all been created.
However, there's no real reason to do this -- it seems fine to read it
a little earlier, when all the other necessary sections are read in.
This patch makes this change, and updates the
read_addrmap_from_aranges API to assert that the section is read in.
This patch slightly changes the read_addrmap_from_aranges API as well,
to reject an empty section. This seems better to me than what the
current code does, which is try to read an empty section but then do
no work.
Regression tested on x86-64 Fedora 38.
Reviewed-By: Guinevere Larsen <blarsen@redhat.com>
I noticed a few more style issues in commit 8b9c08edda ("[gdb/symtab] Add
name_of_main and language_of_main to the DWARF index"), after checking it
with gcc's check_GNU_style.{sh,py}.
Fix these.
Build on x86_64-linux.
The recent change to record the DWARF language in the per-CU data
yielded a race warning in my testing:
ThreadSanitizer: data race ../../binutils-gdb/gdb/dwarf2/read.c:21779 in prepare_one_comp_unit
This patch fixes the bug by applying the same style of fix that was
done for the ordinary (gdb) language.
I wonder if this code could be improved. Requiring an atomic for the
language in particular seems unfortunate, as it is often consulted
during index finalization. However, I haven't investigated this.
Regression tested on x86-64 Fedora 38.
Reviewed-by: Tom de Vries <tdevries@suse.de>
Post-commit review pointed out a few style issues in commit 8b9c08edda
("[gdb/symtab] Add name_of_main and language_of_main to the DWARF index").
Fix these.
Tested on x86_64-linux.
Reported-By: Tom Tromey <tom@tromey.com>
Approved-By: Tom Tromey <tom@tromey.com>
When using glibc debuginfo generated with gas 2.39, we run into PR gas/29517:
...
$ gdb -q -batch a.out -ex start -ex "p (char *)strstr (\"haha\", \"ah\")"
Temporary breakpoint 1 at 0x40051b: file hello.c, line 6.
Temporary breakpoint 1, main () at hello.c:6
6 printf ("hello\n");
Invalid cast.
...
while without glibc debuginfo installed we get the expected result:
...
$n = 0x7ffff7daa1b1 "aha"
...
and likewise with glibc debuginfo generated with gas 2.40.
The strstr ifunc resolves to __strstr_sse2_unaligned. The problem is that gas
generates dwarf that states that the return type is void:
...
<1><3e1e58>: Abbrev Number: 2 (DW_TAG_subprogram)
<3e1e59> DW_AT_name : __strstr_sse2_unaligned
<3e1e5d> DW_AT_external : 1
<3e1e5e> DW_AT_low_pc : 0xbbd2e
<3e1e66> DW_AT_high_pc : 0xbc1c3
...
while the return type should be a DW_TAG_unspecified_type, as is the case
with gas 2.40.
We can still use the workaround of casting to another function type for both
__strstr_sse2_unaligned:
...
(gdb) p ((char * (*) (const char *, const char *))__strstr_sse2_unaligned) \
("haha", "ah")
$n = 0x7ffff7daa211 "aha"
...
and strstr (which requires using *strstr to dereference the ifunc before we
cast):
...
gdb) p ((char * (*) (const char *, const char *))*strstr) ("haha", "ah")
$n = 0x7ffff7daa251 "aha"
...
but that's a bit cumbersome to use.
Work around this in the dwarf reader, such that we have instead:
...
(gdb) p (char *)strstr ("haha", "ah")
$n = 0x7ffff7daa1b1 "aha"
...
This also requires fixing producer_is_gcc to stop returning true for
producer "GNU AS 2.39.0".
Tested on x86_64-linux.
Approved-By: Andrew Burgess <aburgess@redhat.com>
PR symtab/30911
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30911
This patch adds a new section to the DWARF index containing the name
and the language of the main function symbol, gathered from
`cooked_index::get_main`, if available. Currently, for lack of a better name,
this section is called the "shortcut table". The way this name is both saved and
applied upon an index being loaded in mirrors how it is done in
`cooked_index_functions`, more specifically, the full name of the main function
symbol is saved and `set_objfile_main_name` is used to apply it after it is
loaded.
The main use case for this patch is in improving startup times when dealing with
large binaries. Currently, when an index is used, GDB has to expand symtabs
until it finds out what the language of the main function symbol is. For some
large executables, this may take a considerable amount of time to complete,
slowing down startup. This patch bypasses that operation by having both the name
and language of the main function symbol be provided ahead of time by the index.
In my testing (a binary with about 1.8GB worth of DWARF data) this change brings
startup time down from about 34 seconds to about 1.5 seconds.
When testing the patch with target board cc-with-gdb-index, test-case
gdb.fortran/nested-funcs-2.exp starts failing, but this is due to a
pre-existing issue, filed as PR symtab/30946.
Tested on x86_64-linux, with target board unix and cc-with-gdb-index.
PR symtab/24549
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24549
Approved-By: Tom de Vries <tdevries@suse.de>
I noticed a comment by an include and remembered that I think these
don't really provide much value -- sometimes they are just editorial,
and sometimes they are obsolete. I think it's better to just remove
them. Tested by rebuilding.
Approved-By: Andrew Burgess <aburgess@redhat.com>
There are two methods to factor out type information in a dwarf4 executable:
- use -fdebug-info-types to generate type units in a .debug_types section, and
- use dwz to create partial units.
The dwz method has an extra benefit: it also allows to factor out information
between executables into a newly created .dwz file, pointed to by a
.gnu_debugaltlink section.
There is nothing prohibiting a .gnu_debugaltlink file to contain a
.debug_types section.
It's just not generated by dwz or any other tool atm, and consequently gdb has
no support for it. Enhancement PR symtab/30838 is open about the lack of
support.
Make the current situation explicit by emitting a dwarf error:
...
(gdb) file struct-with-sig-2^M
Reading symbols from struct-with-sig-2...^M
Dwarf Error: .debug_types section not supported in dwz file^M
...
and add an assert in write_gdbindex:
...
+ /* See enhancement PR symtab/30838. */
+ gdb_assert (!(per_cu->is_dwz && per_cu->is_debug_types));
...
to clarify why we can use:
...
data_buf &cu_list = (per_cu->is_debug_types
? types_cu_list
: per_cu->is_dwz ? dwz_cu_list : objfile_cu_list);
...
The test-case is a modified copy from gdb.dwarf2/struct-with-sig.exp, so it
keeps the copyright years range.
Tested on x86_64-linux.
Tested-By: Guinevere Larsen <blarsen@redhat.com>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30838
This changes main_type to hold a language, and updates the debug
readers to set this field. This is done by adding the language to the
type-allocator object.
Note that the non-DWARF readers are changed on a "best effort" basis.
This patch also reimplements type::is_array_like to use the type's
language, and it adds a new type::is_string_like as well. This in
turn lets us change the Python implementation of these methods to
simply defer to the type.
In one spot, it will be convenient for a subsequent patch if the CU is
passed to a type-creation helper function. In another spot, remove
the redundant 'objfile' parameter to another such function.
init_fixed_point_type currently takes an objfile and creates its own
type allocator. However, for a later patch it is more convenient if
this function accepts a type allocator. This patch makes this change.
In this commit:
commit 48ac197b0c
Date: Fri Nov 19 10:12:44 2021 -0700
Handle multiple addresses in call_site_target
a buffer overflow bug was introduced when the following code was
added:
CORE_ADDR *saved = XOBNEWVAR (&objfile->objfile_obstack, CORE_ADDR,
addresses.size ());
std::copy (addresses.begin (), addresses.end (), saved);
The definition of XOBNEWVAR is (from libiberty.h):
#define XOBNEWVAR(O, T, S) ((T *) obstack_alloc ((O), (S)))
So 'saved' is going to point to addresses.size () bytes of memory,
however, the std::copy will write addresses.size () number of
CORE_ADDR sized entries to the address pointed to by 'saved', this is
going to result in memory corruption.
The mistake is that we should have used XOBNEWVEC, which allocates a
vector of entries, the definition of XOBNEWVEC is:
#define XOBNEWVEC(O, T, N) \
((T *) obstack_alloc ((O), sizeof (T) * (N)))
Which means we will have set aside enough space to create a copy of
the contents of the addresses vector.
I'm not sure how to create a test for this problem, this issue cropped
up when debugging a particular i686 built binary, which just happened
to trigger a glibc assertion (likely due to random memory corruption),
debugging the same binary built for x86-64 appeared to work just fine.
Using valgrind on the failing GDB binary pointed straight to the cause
of the problem, and with this patch in place there are no longer
valgrind errors in this area.
If anyone has ideas for a test I'm happy to work on something.
Co-Authored-By: Keith Seitz <keiths@redhat.com>
Approved-By: Tom Tromey <tom@tromey.com>
This adds a new enum constant, TYPE_SPECIFIC_RUST_STUFF, and changes
the DWARF reader to set this on Rust types. This will be used as a
flag in a later patch.
Note that the size of the type_specific_field bitfield had to be
increased. I checked that this did not impact the size of main_type.
A user noticed that gdb would crash when showing a backtrace.
Investigation showed this to be a crash in the DWARF reader when
handling a "pragma export" symbol. The bug here is that earlier code
decides to eliminate the symbol, but the export code tries to add it
anyway -- but to a NULL list.
Add these two methods, rename the field to m_bitsize to make it pseudo
private.
Change-Id: Ief95e5cf106e72f2c22ae47b033d0fa47202b413
Approved-By: Tom Tromey <tom@tromey.com>
Add these two methods, rename the field to m_artificial to make it
pseudo private.
Change-Id: If3a3825473d1d79bb586a8a074b87bba9b43fb1a
Approved-By: Tom Tromey <tom@tromey.com>
This patch merges the DW_TAG_typedef case in new_symbol with some
other type-related cases. These all have identical code.
Reviewed-By: Guinevere Larsen <blarsen@redhat.com>
Handle the remaining uses of TYPE_ALLOC, either by:
- replacing with TYPE_ZALLOC, or
- adding a comment explaining why zero-initialization is not necessary.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
I noticed some cases of TYPE_ALLOC followed by B_CLRALL.
Replace these with TYPE_ZALLOC.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
After finding this code in buildsym_compunit::finish_block_internal:
...
ftype->set_fields
((struct field *)
TYPE_ALLOC (ftype, nparams * sizeof (struct field)));
...
and fixing PR30810 by using TYPE_ZALLOC, I wondered if there were more
locations that needed fixing.
I decided to make things easier to spot by factoring out a new function
alloc_fields:
...
/* Allocate the fields array of this type, with NFIELDS elements. If INIT,
zero-initialize the allocated memory. */
void
type::alloc_fields (unsigned int nfields, bool init = true);
...
where:
- a regular use would be "alloc_fields (nfields)", and
- an exceptional use that needed no initialization would be
"alloc_fields (nfields, false)".
Pretty soon I discovered that most of the latter cases are due to
initialization by memcpy, so I added two variants of copy_fields as well.
After this rewrite there are 8 uses of set_fields left:
...
gdb/coffread.c: type->set_fields (nullptr);
gdb/coffread.c: type->set_fields (nullptr);
gdb/coffread.c: type->set_fields (nullptr);
gdb/eval.c: type->set_fields
gdb/gdbtypes.c: type->set_fields (args);
gdb/gdbtypes.c: t->set_fields (XRESIZEVEC (struct field, t->fields (),
gdb/dwarf2/read.c: type->set_fields (new_fields);
gdb/dwarf2/read.c: sub_type->set_fields (sub_type->fields () + 1);
...
These fall into the following categories:
- set to nullptr (coffread.c),
- type not owned by objfile or gdbarch (eval.c), and
- modifying an existing fields array, like adding an element at the end or
dropping an element at the start (the rest).
Tested on x86_64-linux.
Building gdb with gcc 7.5.0 and -flto -O2 -flto-partition=one generates a
self-referencing DIE:
...
<2><91dace>: Abbrev Number: 405 (DW_TAG_label)
<91dad0> DW_AT_abstract_origin: <0x91dace>
...
When encountering the self-reference DIE in inherit_abstract_dies we loop
following the abstract origin, effectively hanging gdb.
Fix this by handling self-referencing DIEs in the loop in
inherit_abstract_dies.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/30799
https://sourceware.org/bugzilla/show_bug.cgi?id=30799
While GDB is still C++11, lets add a gdb::make_unique template
function that can be used to create std::unique_ptr objects, just like
the C++14 std::make_unique.
If GDB is being compiled with a C++14 compiler then the new
gdb::make_unique function will delegate to the std::make_unique. I
checked with gcc, and at -O1 and above gdb::make_unique will be
optimised away completely in this case.
If C++14 (or later) becomes our minimum, then it will be easy enough
to go through the code and replace gdb::make_unique with
std::make_unique later on.
I've make use of this function in all the places I think this can
easily be used, though I'm sure I've probably missed some.
Should be no user visible changes after this commit.
Approved-By: Tom Tromey <tom@tromey.com>
When building gdb with -std=c++20, I run into:
...
gdb/dwarf2/read.c:2709:3: error: invalid conversion from ‘const char8_t*’ to \
‘const char*’ [-fpermissive]
2709 | u8"u8função",
| ^~~~~~~~~~~~
| |
| const char8_t*
...
Fix this by making the conversion explicit.
Tested on x86_64-linux.
While working on a dwarf assembly test-case I accidentally created the
following pathological dwarf:
...
<1><be>: Abbrev Number: 3 (DW_TAG_class_type)
<bf> DW_AT_name : c1
<c2> DW_AT_specification: <0xbe>
...
and noticed gdb segfaulting during cooked index creating due to running out of
stack. This is a regression from gdb-12, where gdb just hung.
Fix this by inhibiting the scan_attributes self-recursion for self-references.
The same test-case with -readnow makes gdb hang, so also fix this in
dwarf2_attr and follow_die_ref.
Note that this doesn't fix the same problems for the more complicated case of:
...
<1><be>: Abbrev Number: 3 (DW_TAG_class_type)
<bf> DW_AT_name : c1
<c2> DW_AT_specification: <0xc6>
<1><c6>: Abbrev Number: 4 (DW_TAG_class_type)
<c7> DW_AT_name : c2
<ca> DW_AT_specification: <0xbe>
...
but the approach for deciding whether to fix pathological dwarf cases is as
per PR27981 comment 3:
...
yes if it is cheap/obvious, and no if it is something complicated or expensive.
...
and at this point I'm not sure whether fixing this will fall in the first
category.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
Test-case gdb.dwarf2/pr13961.exp contains:
...
<1><25>: Abbrev Number: 8 (DW_TAG_class_type)
<26> DW_AT_specification: <0x2a>
<1><2a>: Abbrev Number: 2 (DW_TAG_class_type)
<2b> DW_AT_name : foo
<2f> DW_AT_byte_size : 4
<30> DW_AT_decl_file : 1
<31> DW_AT_decl_line : 1
<32> DW_AT_sibling : <0x44>
...
The DIE at 0x25 contains an intra-CU forward reference, and is deferred during
DIE indexing in the cooked_index, by adding it to m_deferred_entries.
The resulting cooked index entries are:
...
[25] ((cooked_index_entry *) 0x333b5d0)
name: foo
canonical: foo
qualified: foo
DWARF tag: DW_TAG_class_type
flags: 0x0 []
DIE offset: 0x2a
parent: ((cooked_index_entry *) 0)
[26] ((cooked_index_entry *) 0x333b630)
name: foo
canonical: foo
qualified: foo::foo
DWARF tag: DW_TAG_class_type
flags: 0x0 []
DIE offset: 0x25
parent: ((cooked_index_entry *) 0x333b5d0) [foo]
...
Notice that 0x2a is the parent of 0x25, and that this is why the qualified
name of 0x25 is "foo::foo", which is incorrect, it's supposed to be "foo".
The parent is set here in cooked_indexer::make_index:
...
for (const auto &entry : m_deferred_entries)
{
void *obj = m_die_range_map.find (entry.spec_offset);
cooked_index_entry *parent = static_cast<cooked_index_entry *> (obj);
m_index_storage->add (entry.die_offset, entry.tag, entry.flags,
entry.name, parent, m_per_cu);
}
...
and AFAICT, we store in m_die_range_map the parent of the respective
spec_offset DIE (though that's not clear from the comment describing it).
So, the root cause of this is that when we lookup the parent for DIE 0x25, we get
m_die_range_map.find (0x2a) == 0x2a.
This is an off-by-one error, fixed in cooked_indexer::recurse by:
...
- CORE_ADDR start = form_addr (parent_entry->die_offset,
+ CORE_ADDR start = form_addr (parent_entry->die_offset + 1,
...
which gives us:
...
[12] ((cooked_index_entry *) 0x41e21f0)
name: foo
canonical: foo
qualified: foo
DWARF tag: DW_TAG_class_type
flags: 0x0 []
DIE offset: 0x25
parent: ((cooked_index_entry *) 0)
[13] ((cooked_index_entry *) 0x41e2190)
name: foo
canonical: foo
qualified: foo
DWARF tag: DW_TAG_class_type
flags: 0x0 []
DIE offset: 0x2a
parent: ((cooked_index_entry *) 0)
...
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/30739
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30739
When loading an executable using "file a.out", the language is set according
to a.out, which can involve looking up the language of symbol "main", which
will cause the symtab expansion for the containing CU.
Expansion of lto debug info can be slow, so in commit d321419811 ("[gdb] Use
partial symbol table to find language for main") a feature was added to avoid
the symtab expansion.
This feature stopped working after commit 7f4307436f ("Fix "start" for D,
Rust, etc").
[ The commit addresses problems related to command start, which requires finding
the main function:
- for language D, "main" was found instead of "D main", and
- for Rust, the correct function was found, but attributed the wrong name
(not fully qualified). ]
Reimplement the feature by adding
cooked_index_functions::lookup_global_symbol_language.
Tested on x86_64-linux.
PR symtab/30661
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30661
I noticed that some spots in gdb call bfd_set_cacheable after opening
a BFD.
The BFD file cache is a bit odd. BFDs that are opened locally are
unconditionally registered with the cache, and their underlying file
descriptor will always be closed when bfd_cache_close_all is called.
However, only "cacheable" BFDs will be eligible for reopening when
needed -- and by default BFD decides that if a file descriptor is
passed in, then it should not be cacheable. If a non-cacheable BFD's
file descriptor is closed, there is no offical way to reopen it.
gdb needs to call bfd_cache_close_all, because some systems cannot
start an executable when it has an open file descriptor referencing
it.
However, gdb also will sometimes passes an open file descriptor to the
various BFD open functions. And, due to lazy DWARF reading, gdb may
also need to reopen these BFDs.
Rather than having all the callers figure out when exactly to set the
cacheable flag, I think it makes sense to consolidate this logic into
the gdb_bfd.c wrapper functions. It is ok to do this because gdb
always passes a filename to these open functions, so reopening should
work ok.
Regression tested on x86-64 Fedora 38.
Reviewed-by: John Baldwin <jhb@FreeBSD.org>
Alexandre Oliva found a bug in gdb's handling of fixed-point -- a
certain Ada fixed-point type would be misintepreted. The bug was that
the DW_AT_small looked like:
<1><13cd>: Abbrev Number: 16 (DW_TAG_constant)
<13ce> DW_AT_GNU_numerator: 1
<13cf> DW_AT_GNU_denominator: 0x8000000000000000
... but gdb interpreted the denominator as a negative value.
