| Age | Commit message (Collapse) | Author |
|---|
|
When running test-case gdb.dwarf2/fission-loclists.exp, I noticed:
...
warning: Section .debug_aranges in fission-loclists has duplicate \
debug_info_offset 0x8f, ignoring .debug_aranges.^M
...
Fix this by removing the duplicate .debug_aranges entry.
Tested on x86_64-linux.
|
|
In PR23888 an error is reported:
...
ERROR: tcl error sourcing watchpoint-unaligned.exp.
ERROR: expected boolean value but got ""
while executing
"if {$wpnum} {
...
This presumably happens when:
- skip_hw_watchpoint_tests returns 0 meaning hw watchpoints are supported
- gdb fails to set a hw watchpoint and instead sets a sw watchpoint
That particular situation is handled for arm:
...
-re "Watchpoint (\[0-9\]+): .*\r\n$gdb_prompt $" {
if {[istarget "arm*-*-*"]} {
untested $test
set wpnum 0
}
}
...
but not for any other targets so wpnum remains "", triggering the ERROR.
Possibly this has been fixed for powerpc by commit 8d4e4d13afb ("gdb Power 9
add test for HW watchpoint support."), but it's still possible for other
targets.
Fix this by:
- initializing wpnum to 0 instead of ""
- signalling the failure to set a hw watchpoint by a fail
Tested on x86_64-linux, also by adding:
...
gdb_test_no_output "set can-use-hw-watchpoints 0"
...
and verifying that it triggers the fail.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23888
|
|
On aarch64, I run into:
...
FAIL: gdb.base/large-frame.exp: optimize=-O0: backtrace
...
The problem is that the architecture-specific prologue analyzer fails to
handle the first two insns in the prologue properly:
...
0000000000400610 <func>:
400610: d2880210 mov x16, #0x4010
400614: cb3063ff sub sp, sp, x16
400618: a9007bfd stp x29, x30, [sp]
40061c: 910003fd mov x29, sp
400620: 910043a0 add x0, x29, #0x10
400624: 97fffff0 bl 4005e4 <blah>
...
so we get:
...
$ gdb -q -batch ./outputs/gdb.base/large-frame/large-frame-O0 -ex "b func"
Breakpoint 1 at 0x400614
...
Fix this by:
- fixing the support for the first insn to extract the immediate operand, and
- adding support for the second insn,
such that we have:
...
Breakpoint 1 at 0x400624
...
Note that we're overshooting by one insn (0x400620 is the first insn after the
prologue), but that's a pre-existing problem.
Tested on aarch64-linux.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29408
|
|
BFD_VMA_FMT can't be used in format strings that need to be
translated, because the translation won't work when the type of
bfd_vma differs from the machine used to compile .pot files. We've
known about this for a long time, but patches slip through review.
So just get rid of BFD_VMA_FMT, instead using the appropriate PRId64,
PRIu64, PRIx64 or PRIo64 and SCN variants for scanf. The patch is
mostly mechanical, the only thing requiring any thought is casts
needed to preserve PRId64 output from bfd_vma values, or to preserve
one of the unsigned output formats from bfd_signed_vma values.
|
|
Commit f493c2174e messed the formatting in linker map files,
particularly for 32-bit builds where a number of tests using map files
regressed. I should have noticed the BFD64 conditional printing of
spaces to line up output due to the original %V printing hex vmas with
16 digits when BFD64 and 8 digits when not. Besides that, it is nicer
to print 32-bit vmas for 32-bit targets. So change %V back to be
target dependent, now using bfd_sprintf_vma. Since minfo doesn't
return the number of chars printed, that means some places that
currently use %V must instead sprintf to a buffer in order to find the
length printed.
* ldmisc.h (print_spaces): Declare.
(print_space): Change to a macro.
* ldmisc.c (vfinfo): Use bfd_sprintf_vma for %V. Tidy %W case.
(print_space): Delete.
(print_spaces): New function.
* emultempl/aix.em (print_symbol): Use print_spaces.
* ldctor.c (ldctor_build_sets): Likewise.
* ldmain.c (add_archive_element): Likewise.
* ldlang.c (print_one_symbol, lang_print_asneeded): Likewise.
(print_output_section_statement, print_data_statement): Likewise.
(print_reloc_statement, print_padding_statement): Likewise.
(print_assignment): Likewise. Also replace %V printing of vmas
with printing to a buffer in order to properly format output.
(print_input_section, lang_one_common): Likewise.
|
|
R_MIPS_REL16 isn't a pc-relative reloc as the name might indicate.
* elf64-mips.c (mips_reloc_map): Map BFD_RELOC_16 to R_MIPS_REL16.
* elfn32-mips.c (mips_reloc_map): Likewise.
|
|
|
|
Reset alignment for each PT_LOAD segment to avoid using alignment from
the previous PT_LOAD segment.
bfd/
PR ld/29435
* elf.c (assign_file_positions_for_load_sections): Reset
alignment for each PT_LOAD segment.
ld/
PR ld/29435
* testsuite/ld-elf/pr29435.d: New file.
* testsuite/ld-elf/pr29435.s: Likewise.
|
|
In some cases, the objfile owns the per-bfd object. This is yet
another object that can sometimes be destroyed before the registry is
destroyed, possibly reslting in a use-after-free. Also, I noticed
that the condition for deleting the object is not the same as the
condition used to create it -- so it could possibly result in a memory
leak in some situations. This patch fixes the problem by introducing
a new unique_ptr that holds this object when necessary.
|
|
This changes objfile to use an auto_obstack. This helps prevent
use-after-free bugs, because it ensures that anything allocated on the
objfile obstack will live past the point at which the registry object
is destroyed.
|
|
This changes struct objfile to use a gdb_bfd_ref_ptr. In addition to
removing some manual memory management, this fixes a use-after-free
that was introduced by the registry rewrite series. The issue there
was that, in some cases, registry shutdown could refer to memory that
had already been freed. This help fix the bug by delaying the
destruction of the BFD reference (and thus the per-bfd object) until
after the registry has been shut down.
|
|
gprofng/Changelog:
2022-08-02 Ruud van der Pas <ruud.vanderpas@oracle.com>
PR gprofng/29410
* gp-display-html/gp-display-html.in: Remove non-breaking spaces.
|
|
libraries and the BFD library's file caching mechanism.
PR 29389
bfd * bfd.c (BFD_CLOSED_BY_CACHE): New bfd flag.
* cache.c (bfd_cache_delete): Set BFD_CLOSED_BY_DELETE on the
closed bfd.
(bfd_cache_lookup_worker): Clear BFD_CLOSED_BY_DELETE on the newly
reopened bfd.
* opncls.c (bfd_set_filename): Refuse to change the name of a bfd
that has been closed by bfd_cache_delete. Mark changed bfds as
uncacheable.
* bfd-in2.h: Regenerate.
ld * ldlang.h (lang_input_statement_struct): Add sort_key field.
* emultempl/pe.em (after_open): If multiple import libraries refer
to the same bfd, store their names in the sort_key field.
* emultempl/pep.em (after_open): Likewise.
* ldlang.c (sort_filename): New function. Returns the filename to
be used when sorting input files.
(wild_sort): Use the sort_filename function.
|
|
When building with clang 15, I got this,
CXX amd64-tdep.o
amd64-tdep.c:1410:13: error: variable 'insn' set but not used[-Werror,-Wunused-but-set-variable]
gdb_byte *insn = insn_details->raw_insn + modrm_offset;
^
1 error generated.
The function that uses this variable has been removed in this commit,
commit 870f88f7551b0f2d6aaaa36fb684b5ff8f468107
Date: Mon Apr 18 13:16:27 2016 -0400
remove trivialy unused variables
Fix this by removing unused variable.
Tested by rebuilding on x86_64-linux with clang 15 and gcc 12.
|
|
Commit bc20e562ec0 "Fix use after free in varobj" introduced a
regression. This commit makes sure that the varobj object does not
keeps stale references to object being freed when we unload an objfile.
This includes the "valid_block" field which is reset to nullptr if the
pointed to block is tied to an objfile being freed.
However, at some point varobj_invalidate_iter might try to recreate
varobjs tracking either floating or globals. Varobj tracking globals
are identified as having the "valid_block" field set nullptr, but as
bc20e562ec0 might clear this field, we have lost the ability to
distinguish between varobj referring to globals and non globals.
Fix this by introducing a "global" flag which tracks if a given varobj
was initially created as tracking a global.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29426
|
|
All of these buffer overrun tests are better written as a comparison
against size remaining, due to ISO C 9899 standard 6.5.2 para 8
regarding adding a constant to a pointer:
"If both the pointer operand and the result point to elements of the
same array object, or one past the last element of the array object,
the evaluation shall not produce an overflow; otherwise, the behavior
is undefined."
So "ex_dta + 4" might be undefined behaviour, if you interpret "the
array object" in this case to be the malloc'd section contents!
* pei-x86_64.c (pex64_get_unwind_info): Tidy sanity checks.
(pex64_xdata_print_uwd_codes): Likewise.
|
|
- Drop the rounding type check: We're past template matching, and none
of the involved insns support embedded rounding.
- Drop the extension opcode check: None of the involved opcodes have
variants with it being other than None.
- Instead check opcode space, even if just to be on the safe side going
forward.
- Reduce the number of comparisons by folding two groups.
|
|
Just like all Size64 insns are marked Cpu64, all Size32 insns ought to
be marked Cpu386.
|
|
First of all rename the meanwhile misleading Opcode_SIMD_FloatD, as it
has also been used for KMOV* and BNDMOV. Then simplify the condition
selecting which form if "reversing" to use - except for the MOV to/from
control/debug/test registers all extended opcode space insns use bit 0
(rather than bit 1) to indicate the direction (from/to memory) of an
operation. With that, D can simply be set on the first of the two
templates, while the other can be dropped.
|
|
|
|
include/elf/
* common.h: Add ELFCOMPRESS_ZSTD.
|
|
During the development of 40c23d880386d6e8202567eaa2a6b041feb1a652,
the return value of fbsd_nat_target::have_regset was changed from a
simple boolean to returning the size of the register set. The
comments and callers were all updated for this change, but the actual
return type was accidentally left as a bool. This change fixes the
return type to be a size_t.
Current callers of this only checked the value against 0 and thus
still worked correctly.
|
|
Even when there are no symbols (e.g. all relocations being against
absolute values), a symbol table (with just the first placeholder entry)
needs to be emitted. Otherwise tools like objdump won't properly process
the relocations. The respective checks in assign_section_numbers() and
_bfd_elf_compute_section_file_positions() support also this view. Oddly
enough so far HAS_RELOC was only set when reading in an object file, but
not when generating one anew; the flag would only have been cleared when
no relocations were found (anymore).
While there also amend the affected function's leading comment to also
mention gas.
|
|
In aarch64_tls_transition_without_check and elfNN_aarch64_tls_relax we
choose whether to perform a relaxation to an IE access model or an LE
access model based on whether the symbol itself is marked as local (i.e.
`h == NULL`).
This is problematic in two ways. The first is that sometimes a global
dynamic access can be relaxed to an initial exec access when creating a
shared library, and if that happens on a local symbol then we currently
relax it to a local exec access instead. This usually does not happen
since we only relax an access if aarch64_can_relax_tls returns true and
aarch64_can_relax_tls does not have the same problem. However, it can
happen when we have seen both an IE and GD access on the same symbol.
This case is exercised in the newly added testcase tls-relax-gd-ie-2.
The second problem is that deciding based on whether the symbol is local
misses the case when the symbol is global but is still non-interposable
and known to be located in the executable. This happens on all global
symbols in executables.
This case is exercised in the newly added testcase tls-relax-ie-le-4.
Here we adjust the condition we base our relaxation on so that we relax
to local-exec if we are creating an executable and the relevant symbol
we're accessing is stored inside that executable.
-- Updating tests for new relaxation criteria
Many of the tests added to check our relaxation to IE were implemented
by taking advantage of the fact that we did not relax a global symbol
defined in an executable.
Since a global symbol defined in an executable is still not
interposable, we know that a TLS version of such a symbol will be in the
main TLS block. This means that we can perform a stronger relaxation on
such symbols and relax their accesses to a local-exec access.
Hence we have to update all tests that relied on the older suboptimal
decision making.
The two cases when we still would want to relax a general dynamic access
to an initial exec one are:
1) When in a shared library and accessing a symbol which we have already
seen accessed with an initial exec access sequence.
2) When in an executable and accessing a symbol defined in a shared
library.
Both of these require shared library support, which means that these
tests are now only available on targets with that.
I have chosen to switch the existing testcases from a plain executable
to one dynamically linked to a shared object as that doesn't require
changing the testcases quite so much (just requires accessing a
different variable rather than requiring adding another code sequence).
The tls-relax-all testcase was an outlier to the above approach, since
it included a general dynamic access to both a local and global symbol
and inspected for the difference accordingly.
|
|
The updates are to ensure that the .data section exists. This means
that we always have a data section. That means that we don't create a
RWX segment and avoid the corresponding warning.
We get this warning when testing aarch64-none-elf with -mcmodel=tiny.
N.b. this changes quite a few testcases from fail to pass.
|
|
This patch extends assembler support for the use of register names to
allow for pseudo-registers, e.g. ra_auth_code register.
This is done particularly with CFI directives in mind, allowing for
expressions of the type:
.cfi_register ra_auth_code, 12
gas/Changelog:
* config/tc-arm.c (tc_arm_regname_to_dw2regnum): Add
REG_TYPE_PSEUDO handling.
* testsuite/gas/arm/cfi-pacbti-m-readelf.d: New.
* testsuite/gas/arm/cfi-pacbti-m.s: New.
|
|
This patch modifies the internal `struct reg_entry' numbering of DWARF
pseudo-registers to match values assigned in DWARF standards (see "4.1
DWARF register names" in [1])so ra_auth_code goes from 12 to 143 and
amends the unwinder .save directive-processing code to correctly handle
mixed register-type save directives.
The mechanism for splitting the register list is also re-written to
comply with register ordering on push statements, being that registers
are stored on the stack in numerical order, with the lowest numbered
register at the lowest address [2].
Consequently, the parsing of the hypothetical directive
.save{r4-r7, r10, ra_auth_core, lr}
has been changed such as rather than producing
.save{r4-r7, r10}
.save{ra_auth_code}
.save{lr}
as was the case with previous implementation, now produces:
.save{lr}
.save{ra_auth_code}
.save{r4-r7, r10}
[1] <https://github.com/ARM-software/abi-aa/blob/main/aadwarf32/aadwarf32.rst>
[2] <https://developer.arm.com/documentation/dui0473/j/arm-and-thumb-instructions/push>
gas/Changelog:
* config/tc-arm.c (REG_RA_AUTH_CODE): New.
(parse_dot_save): Likewise.
(parse_reg_list): Remove obsolete code.
(reg_names): Set ra_auth_code to 143.
(s_arm_unwind_save): Handle core and pseudo-register lists via
parse_dot_save.
(s_arm_unwind_save_mixed): Deleted.
(s_arm_unwind_save_pseudo): Handle one register at a time.
* testsuite/gas/arm/unwind-pacbti-m-readelf.d: Fix test.
* testsuite/gas/arm/unwind-pacbti-m.d: Likewise.
|
|
objdump -x on PE executables produces lots of "xdata section corrupt"
and "corrupt unwind data" warnings, and refuses to dump that info. It
turns out that the sanity checks were bad, not the data. Fix them.
* pei-x86_64.c (pex64_get_unwind_info): Correct buffer overrun
sanity checks.
(pex64_xdata_print_uwd_codes): Similarly.
|
|
By mistake it was permitted to be used from the very introduction of XOP
support.
|
|
|
|
EXCLUDE_SYMBOLS keyword
This was requested in review.
|
|
directives
This maps to the same as ld's --exclude-symbols command line option,
but allowing specifying the option via directives embedded in the
object files instead of passed manually on the command line.
|
|
The function read_addrmap_from_aranges contains code to issue a warning:
...
if (!insertpair.second)
{
warning (_("Section .debug_aranges in %s has duplicate "
"debug_info_offset %s, ignoring .debug_aranges."),
objfile_name (objfile), sect_offset_str (per_cu->sect_off));
return false;
}
...
but the warning is in fact activated when all_comp_units has duplicate
entries, which is very misleading.
Fix this by:
- adding a test-case that should trigger the warning,
- replacing the current implementation of the warning with an
assert that all_comp_units should not contain duplicates, and
- properly re-implementing the warning, such that it is triggered
by the test-case.
Tested on x86_64-linux.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29381
|
|
Without it in 16-bit mode a pointless operand size prefix would be
emitted.
|
|
This syntactic sugar is present in both classical and emerging
architectures, like Alpha, SPARC and RISC-V, and assembler macros
doing the same thing can already be found in the wild e.g. [1], proving
the feature's popularity. It's better to provide support directly in the
assembler so downstream users wouldn't have to re-invent this over and
over again.
[1]: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/loongarch/sysdep.h;h=c586df819cd90;hb=HEAD#l28
|
|
b{lt/ge}[u] forms
Also re-order the jump/branch opcodes while at it, so that insns are
sorted in ascending order according to opcodes, and the label form
preceding the real definition.
|
|
These two macros print either a 16 digit hex number or an 8 digit
hex number. Unfortunately they depend on both target and host, which
means that the output for 32-bit targets may be either 8 or 16 hex
digits.
Replace them in most cases with code that prints a bfd_vma using
PRIx64. In some cases, deliberately lose the leading zeros.
This change some output, notably in base/offset fields of m68k
disassembly which I think looks better that way, and in error
messages. I've kept leading zeros in symbol dumps (objdump -t)
and in PE header dumps.
bfd/
* bfd-in.h (fprintf_vma, sprintf_vma, printf_vma): Delete.
* bfd-in2.h: Regenerate.
* bfd.c (bfd_sprintf_vma): Don't use sprintf_vma.
(bfd_fprintf_vma): Don't use fprintf_vma.
* coff-rs6000.c (xcoff_reloc_type_tls): Don't use sprintf_vma.
Instead use PRIx64 to print bfd_vma values.
(xcoff_ppc_relocate_section): Likewise.
* cofflink.c (_bfd_coff_write_global_sym): Likewise.
* mmo.c (mmo_write_symbols_and_terminator): Likewise.
* srec.c (srec_write_symbols): Likewise.
* elf32-xtensa.c (print_r_reloc): Similarly for fprintf_vma.
* pei-x86_64.c (pex64_dump_xdata): Likewise.
(pex64_bfd_print_pdata_section): Likewise.
* som.c (som_print_symbol): Likewise.
* ecoff.c (_bfd_ecoff_print_symbol): Use bfd_fprintf_vma.
opcodes/
* dis-buf.c (perror_memory, generic_print_address): Don't use
sprintf_vma. Instead use PRIx64 to print bfd_vma values.
* i386-dis.c (print_operand_value, print_displacement): Likewise.
* m68k-dis.c (print_base, print_indexed): Likewise.
* ns32k-dis.c (print_insn_arg): Likewise.
* ia64-gen.c (_opcode_int64_low, _opcode_int64_high): Delete.
(opcode_fprintf_vma): Delete.
(print_main_table): Use PRIx64 to print opcode.
binutils/
* od-macho.c: Replace all uses of printf_vma with bfd_printf_vma.
* objcopy.c (copy_object): Don't use sprintf_vma. Instead use
PRIx64 to print bfd_vma values.
(copy_main): Likewise.
* readelf.c (CHECK_ENTSIZE_VALUES): Likewise.
(dynamic_section_mips_val): Likewise.
(print_vma): Don't use printf_vma. Instead use PRIx64 to print
bfd_vma values.
(dump_ia64_vms_dynamic_fixups): Likewise.
(process_version_sections): Likewise.
* rddbg.c (stab_context): Likewise.
gas/
* config/tc-i386.c (offset_in_range): Don't use sprintf_vma.
Instead use PRIx64 to print bfd_vma values.
(md_assemble): Likewise.
* config/tc-mips.c (load_register, macro): Likewise.
* messages.c (as_internal_value_out_of_range): Likewise.
* read.c (emit_expr_with_reloc): Likewise.
* config/tc-ia64.c (note_register_values): Don't use fprintf_vma.
Instead use PRIx64 to print bfd_vma values.
(print_dependency): Likewise.
* listing.c (list_symbol_table): Use bfd_sprintf_vma.
* symbols.c (print_symbol_value_1): Use %p to print pointers.
(print_binary): Likewise.
(print_expr_1): Use PRIx64 to print bfd_vma values.
* write.c (print_fixup): Use %p to print pointers. Don't use
fprintf_vma.
* testsuite/gas/all/overflow.l: Update expected output.
* testsuite/gas/m68k/mcf-mov3q.d: Likewise.
* testsuite/gas/m68k/operands.d: Likewise.
* testsuite/gas/s12z/truncated.d: Likewise.
ld/
* deffilep.y (def_file_print): Don't use fprintf_vma. Instead
use PRIx64 to print bfd_vma values.
* emultempl/armelf.em (gld${EMULATION_NAME}_finish): Don't use
sprintf_vma. Instead use PRIx64 to print bfd_vma values.
* emultempl/pe.em (gld${EMULATION_NAME}_finish): Likewise.
* ldlang.c (lang_map): Use %V to print region origin.
(lang_one_common): Don't use sprintf_vma.
* ldmisc.c (vfinfo): Don't use fprintf_vma or sprintf_vma.
* pe-dll.c (pe_dll_generate_def_file): Likewise.
gdb/
* remote.c (remote_target::trace_set_readonly_regions): Replace
uses of sprintf_vma with bfd_sprintf_vma.
|
|
|
|
* ctf-link.c (ctf_link_add_ctf_internal): Don't free uninitialised
pointers.
|
|
There is a problem with my commit 0e3c1eebb2, which replaced
bfd_uint64_t with uint64_t: Some hosts typedef int64_t to long long
even when long is the same size as long long. That confuses the code
choosing one of "l", "ll", or "I64" for BFD_VMA_FMT, and results in
warnings.
Write a direct configure test for the printf int64_t style instead.
This removes the last use of BFD_HOST_64BIT_LONG, so delete that.
Note that the changes to configure.com are pure guesswork.
PR 29348
* bfd-in.h (BFD_HOST_64BIT_LONG): Don't define.
(BFD_VMA_FMT): Define using BFD_INT64_FMT when 64-bit.
(bfd_vma, bfd_signed_vma): Move comments to 64-bit typedefs.
* configure.ac (BFD_HOST_64BIT_LONG): Delete.
(BFD_INT64_FMT): New config test.
* configure.com: Update similarly.
* Makefile.in: Regenerate.
* bfd-in2.h: Regenerate.
* configure: Regenerate.
|
|
|
|
|
|
On aarch64-linux, I run into:
...
(gdb) print 16#ffffffffffffffff#^M
$7 = 18446744073709551615^M
(gdb) FAIL: gdb.ada/literals.exp: print 16#ffffffffffffffff#
...
while on x86_64-linux instead, I get:
...
(gdb) print 16#ffffffffffffffff#^M
$7 = -1^M
(gdb) PASS: gdb.ada/literals.exp: print 16#ffffffffffffffff#
...
We can easily reproduce this on x86_64-linux using:
...
$ gdb -q -batch -ex "set lang ada" -ex "set arch i386" \
-ex "print 16#ffffffffffffffff#"
$1 = -1
$ gdb -q -batch -ex "set lang ada" -ex "set arch aarch64" \
-ex "print 16#ffffffffffffffff#"
$1 = 18446744073709551615
...
With i386, we have:
...
(gdb) p int_bits
$3 = 32
(gdb) p long_bits
$4 = 32
(gdb) p long_long_bits
$5 = 64
...
and so in processInt we hit the fits-in-unsigned-long-long case where we use
as type long long:
...
/* Note: Interprets ULLONG_MAX as -1. */
yylval.typed_val.type = type_long_long (par_state);
...
With aarch64, we have instead:
...
(gdb) p int_bits
$1 = 32
(gdb) p long_bits
$2 = 64
(gdb) p long_long_bits
$3 = 64
...
and so in processInt we hit the fits-in-unsigned-long case where we use
as type unsigned long:
...
yylval.typed_val.type
= builtin_type (par_state->gdbarch ())->builtin_unsigned_long;
...
It's not clear why for ada we're using long long for the
fits-in-unsigned-long-long case.
Fix this by using unsigned long long for the fits-in-unsigned-long-long case,
meaning the new reference output is 18446744073709551615 instead of -1.
Tested on x86_64-linux.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29416
|
|
Using different ways of passing source file paths to compilers results n
different file and directory paths in the line header. For example:
- gcc foo.c
- gcc ./foo.c
- gcc ../cwd/foo.c
- gcc $PWD/foo.c
Because of this, GDB sometimes failed to look up macros. The previous
patch fixed that as much as possible. This patch adds the corresponding
tests.
Add both a DWARF assembler-based test and a regular test. The DWARF
assembled-based one tests some hard-coded debug info based on what I
have observed some specific versions of gcc and clang generate. We want
to make sure that GDB keeps handling all these cases correctly, even if
it's not always clear whether they are really valid DWARF. Also, they
will be tested no matter what the current target compiler is for a given
test run.
The regular test is compiled using the target compiler, so it may help
find bugs when testing against some other toolchains than what was used
to generate the DWARF assembler-based test.
For the DWARF assembler-based test, add to testsuite/lib/dwarf.exp the
necessary code to generate a DWARF5 .debug_macro section. The design of
the new procs is based on what was done for rnglists and loclists.
To test against a specific compiler one can use this command, for
example:
$ make check TESTS="gdb.base/macro-source-path.exp" RUNTESTFLAGS="CC_FOR_TARGET=clang --target_board unix/gdb:debug_flags=-gdwarf-5"
Change-Id: Iab8da498e57d10cc2a3d09ea136685d9278cfcf6
|
|
The bit of code removed by this patch was introduced to fix the same
kind of problem that the previous patch fixes. That is, to try to match
existing subfiles when different name forms are used to refer to a same
file.
The thread for the patch that introduced this code is:
https://pi.simark.ca/gdb-patches/45F8CBDF.9090501@hq.tensilica.com/
The important bits are that the compiler produced a compilation unit
with:
DW_AT_name : test.c
DW_AT_comp_dir : /home/maxim/W/BadgerPass/PR_14999
and DWARF v2 line table with:
The Directory Table:
/home/maxim/W/BadgerPass/PR_14999
The File Name Table:
Entry Dir Time Size Name
1 1 1173897037 152 test.c
Because the main symtab was created with only DW_AT_name, it was named
"test.c". And because the path built from the line header contained the
"directory" part, it was "/home/maxim/W/BadgerPass/PR_14999/test.c".
Because of this mismatch, thing didn't work, so they added this code to
prepend the compilation directory to the existing subfile names, so that
this specific case would work.
With the changes done earlier in this series, where subfiles are
identified using the "most complete path possible", this case would be
handled. The main subfile's would be
"/home/maxim/W/BadgerPass/PR_14999/test.c" from the start
(DW_AT_comp_dir + DW_AT_name). It's not so different from some DWARF 5
cases actually, which make the compilation directory explicit in the
line table header.
I therefore think that this code is no longer needed. It does feel like
a quick hack to make one specific case work, and we have a more general
solution now. Also, this code was introduced to work around a problem
in the DWARF debug info or the DWARF debug info reader. In general, I
think it's preferable for these hacks to be located in the specific
debug info reader code, rather than in the common code.
Even though this code was added to work around a DWARF reader problem,
it's possible that some other debug info reader has started taking
advantage of this code in the mean time. It's very difficult to
know or verify, but I think the likelyhood is quite small, so I'm
proposing to get rid of it to simplify things a little bit.
Change-Id: I710b8ec0d449d1b110d67ddf9fcbdb2b37108306
|
|
Printing macros defined in the main source file doesn't work reliably
using various toolchains, especially when DWARF 5 is used. For example,
using the binaries produced by either of these commands:
$ gcc --version
gcc (GCC) 11.2.0
$ ld --version
GNU ld (GNU Binutils) 2.38
$ gcc test.c -g3 -gdwarf-5
$ clang --version
clang version 13.0.1
$ clang test.c -gdwarf-5 -fdebug-macro
I get:
$ ./gdb -nx -q --data-directory=data-directory a.out
(gdb) start
Temporary breakpoint 1 at 0x111d: file test.c, line 6.
Starting program: /home/simark/build/binutils-gdb-one-target/gdb/a.out
Temporary breakpoint 1, main () at test.c:6
6 return ZERO;
(gdb) p ZERO
No symbol "ZERO" in current context.
When starting to investigate this (taking the gcc-compiled binary as an
example), we see that GDB fails to look up the appropriate macro scope
when evaluating the expression. While stopped in
macro_lookup_inclusion:
(top-gdb) p name
$1 = 0x62100011a980 "test.c"
(top-gdb) p source.filename
$2 = 0x62100011a9a0 "/home/simark/build/binutils-gdb-one-target/gdb/test.c"
`source` is the macro_source_file that we would expect GDB to find.
`name` comes from the symtab::filename field of the symtab we are
stopped in. GDB doesn't find the appropriate macro_source_file because
the name of the macro_source_file doesn't match exactly the name of the
symtab.
The name of the main symtab comes from the compilation unit's
DW_AT_name, passed to the buildsym_compunit's constructor:
https://gitlab.com/gnutools/binutils-gdb/-/blob/4815d6125ec580cc02a1094d61b8c9d1cc83c0a1/gdb/dwarf2/read.c#L10627-10630
The contents of DW_AT_name, in this case, is "test.c". It is typically
(what I witnessed all compilers do) the same string that was passed to
the compiler on the command-line.
The name of the macro_source_file comes from the line number program
header's file table, from the call to the line_header::file_file_name
method:
https://gitlab.com/gnutools/binutils-gdb/-/blob/4815d6125ec580cc02a1094d61b8c9d1cc83c0a1/gdb/dwarf2/macro.c#L54-65
line_header::file_file_name prepends the directory path that the file
entry refers to, in the file table (if the file name is not already
absolute). In this case, the file name is "test.c", appended to the
directory "/home/simark/build/binutils-gdb-one-target/gdb".
Because the symtab's name is not created the same way as the
macro_source_file's name is created, we get this mismatch. GDB fails to
find the appropriate macro scope for the symtab, and we can't print
macros when stopped in that symtab.
To make this work, we must ensure that paths produced in these two ways
end up identical. This can be tricky because of the different ways a
path can be passed to the compiler by the user.
Another thing to consider is that while the main symtab's name (or
subfile, before it becomes a symtab) is created using DW_AT_name, the
main symtab is also referred to using its entry in the line table
header's file table, when processing the line table. We must therefore
ensure that the same name is produced in both cases, so that a call to
"start_subfile" for the main subfile will correctly find the
already-created subfile, created by buildsym_compunit's constructor. If
we fail to do that, things still often work, because of a fallback: the
watch_main_source_file_lossage method. This method determines that if
the main subfile has no symbols but there exists another subfile with
the same basename (e.g. "test.c") that does have symbols, it's probably
because there was some filename mismatch. So it replaces the main
subfile with that other subfile. I think that heuristic is useful as a
last effort to work around any bug or bad debug info, but I don't think
we should design things such as to rely on it. It's a heuristic, it can
get things wrong. So in my search for a fix, it is important that given
some good debug info, we don't end up relying on that for things to
work.
A first attempt at fixing this was to try to prepend the compilation
directory here or not prepend it there. In practice, because of all the
possible combinations of debug info the compilers produce, it was not
possible to get something that would produce reliable, consistent paths.
Another attempt at fixing this was to make both macro_source_file
objects and symtab objects use the most complete form of path possible.
That means to prepend directories at least until we get an absolute
path. In theory, we should end up with the same path in all cases.
This generally worked, but because it changed the symtab names, it
resulted in user-visible changes (for example, paths to source files in
Breakpoint hit messages becoming always absolute). I didn't find this
very good, first because there is a "set filename-display" setting that
lets the user control how they want the paths to be displayed, and that
would suddenly make this setting completely ineffective (although even
today, it is a bit dependent on the debug info). Second, it would
require a good amount of testsuite tweaks to make tests accept these
suddenly absolute paths.
This new patch is a slight variation of that: it adds a new field called
"filename_for_id" in struct symtab and struct subfile, next to the
existing filename field. The goal is to separate the internal ids used
for finding objects from the names used for presentation. This field is
used for identifying subfiles, symtabs and macro_source_files
internally. For DWARF symtabs, this new field is meant to contain the
"most complete possible" path, as discussed above. So for a given file,
it must always be in the same form, everywhere. The existing
symtab::filename field remains the one used for printing to the user, so
there shouldn't be any change in how paths are printed.
Changes in the core symtab files are:
- Add "name_for_id" and "filename_for_id" fields to "struct subfile"
and "struct symtab", next to existing "name" and "filename" fields.
- Make buildsym_compunit::buildsym_compunit and
buildsym_compunit::start_subfile accept a "name_for_id" parameter
next to the existing "name" ones.
- Make buildsym_compunit::start_subfile use "name_for_id" for looking
up existing subfiles. This is the key thing for making calls
to start_subfile for the main source file look up the existing
subfile successfully, and avoid relying on
watch_main_source_file_lossage.
- Make sal_macro_scope pass "filename_for_id", rather than "filename",
to macro_lookup_inclusion. This is the key thing to making the
lookup work and macro printing work.
Changes in the DWARF files are:
- Make line_header::file_file_name return the "most complete possible"
name. The only pre-existing user of this method is the macro code,
to give the macro_source_file objects their name. And we now want
them to have this "most complete possible" name, which will match the
corresponding symtab's "filename_for_id".
- Make dwarf2_cu::start_compunit_symtab pass the "most complete
possible" name for the main symtab's "filename_for_id". In this
context, where the info comes from the compilation unit's DW_AT_name
/ DW_AT_comp_dir, it means prepending DW_AT_comp_dir to DW_AT_name if
DW_AT_name is not already absolute.
- Change dwarf2_start_subfile to build a name_for_id for the subfile
being started. The simplest way is to re-use
line_header::file_file_name, since the callers always have a
file_entry handy. This ensures that it will get the exact same path
representation as the macro code does, for the same file (since it
also uses line_header::file_file_name).
- Update calls to allocate_symtab to pass the "name_for_id" from the
subfile.
Tests exercising all this are added by the following patch.
Of all the cases I tried, the only one I found that ends up relying on
watch_main_source_file_lossage is the following one:
$ clang --version
clang version 13.0.1
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin
$ clang ./test.c -g3 -O0 -gdwarf-4
$ ./gdb -nx --data-directory=data-directory -q -readnow -iex "set debug symtab-create 1" a.out
...
[symtab-create] start_subfile: name = test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/test.c
[symtab-create] start_subfile: name = ./test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/./test.c
[symtab-create] start_subfile: name = ./test.c, name_for_id = /home/simark/build/binutils-gdb-one-target/gdb/./test.c
[symtab-create] start_subfile: found existing symtab with name_for_id /home/simark/build/binutils-gdb-one-target/gdb/./test.c (/home/simark/build/binutils-gdb-one-target/gdb/./test.c)
[symtab-create] watch_main_source_file_lossage: using subfile ./test.c as the main subfile
As we can see, there are two forms used for "test.c", one with a "." and
one without. This comes from the fact that the compilation unit DIE
contains:
DW_AT_name ("test.c")
DW_AT_comp_dir ("/home/simark/build/binutils-gdb-one-target/gdb")
without a ".", and the line table for that file contains:
include_directories[ 1] = "."
file_names[ 1]:
name: "test.c"
dir_index: 1
When assembling the filename from that entry, we get a ".".
It is a bit unexpected that the main filename resulting from the line
table header does not match exactly the name in the compilation unit.
For instance, gcc uses "./test.c" for the DW_AT_name, which gives
identical paths in the compilation unit and in the line table header.
Similarly, with DWARF 5:
$ clang ./test.c -g3 -O0 -gdwarf-5
clang create two entries that refer to the same file but are of in a different
form.
include_directories[ 0] = "/home/simark/build/binutils-gdb-one-target/gdb"
include_directories[ 1] = "."
file_names[ 0]:
name: "test.c"
dir_index: 0
file_names[ 1]:
name: "test.c"
dir_index: 1
The first file name produces a path without a "." while the second does.
This is not caught by watch_main_source_file_lossage, because of
dwarf_decode_lines that creates a symtab for each file entry in the line
table. It therefore appears as "non-empty" to
watch_main_source_file_lossage. This results in two symtabs:
(gdb) maintenance info symtabs
{ objfile /home/simark/build/binutils-gdb-one-target/gdb/a.out ((struct objfile *) 0x613000005d00)
{ ((struct compunit_symtab *) 0x62100011aca0)
debugformat DWARF 5
producer clang version 13.0.1
name test.c
dirname /home/simark/build/binutils-gdb-one-target/gdb
blockvector ((struct blockvector *) 0x621000129ec0)
user ((struct compunit_symtab *) (null))
{ symtab test.c ((struct symtab *) 0x62100011ad20)
fullname (null)
linetable ((struct linetable *) 0x0)
}
{ symtab ./test.c ((struct symtab *) 0x62100011ad60)
fullname (null)
linetable ((struct linetable *) 0x621000129ef0)
}
}
}
I am not sure what is the consequence of this, but this is also what
happens before my patch, so I think its acceptable to leave it as-is.
To handle these two cases nicely, I think we will need a function that
removes the unnecessary "." from path names, something that can be done
later.
Finally, I made a change in find_file_and_directory is necessary to
avoid breaking test
gdb.dwarf2/dw2-compdir-oldgcc.exp: info source gcc42
Without that change, we would get:
(gdb) info source
Current source file is /dir/d/dw2-compdir-oldgcc42.S
Compilation directory is /dir/d
whereas the expected result is:
(gdb) info source
Current source file is dw2-compdir-oldgcc42.S
Compilation directory is /dir/d
This test was added here:
https://sourceware.org/pipermail/gdb-patches/2012-November/098144.html
Long story short, GCC <= 4.2 apparently had a bug where it would
generate a DW_AT_name with a full path ("/dir/d/dw2-compdir-oldgcc42.S")
and no DW_AT_comp_dir. The line table has one entry with filename
"dw2-compdir-oldgcc42.S", which refers to directory 0. Directory 0
normally refers to the compilation unit's comp dir, but it is
non-existent in this case.
This caused some symtab lookup problems, and to work around them, some
workaround was added, which today reads as:
if (res.get_comp_dir () == nullptr
&& producer_is_gcc_lt_4_3 (cu)
&& res.get_name () != nullptr
&& IS_ABSOLUTE_PATH (res.get_name ()))
res.set_comp_dir (ldirname (res.get_name ()));
Source: https://gitlab.com/gnutools/binutils-gdb/-/blob/6577f365ebdee7dda71cb996efa29d3714cbccd0/gdb/dwarf2/read.c#L9428-9432
It extracts an artificial DW_AT_comp_dir from DW_AT_name, if there is no
DW_AT_comp_dir and DW_AT_name is absolute.
Prior to my patch, a subfile would get created with filename
"/dir/d/dw2-compdir-oldgcc42.S", from DW_AT_name, and another would get
created with filename "dw2-compdir-oldgcc42.S" from the line table's
file table. Then watch_main_source_file_lossage would kick in and merge
them, keeping only the "dw2-compdir-oldgcc42.S" one:
[symtab-create] start_subfile: name = /dir/d/dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: found existing symtab with name dw2-compdir-oldgcc42.S (dw2-compdir-oldgcc42.S)
[symtab-create] watch_main_source_file_lossage: using subfile dw2-compdir-oldgcc42.S as the main subfile
And so "info source" would show "dw2-compdir-oldgcc42.S" as the
filename.
With my patch applied, but without the change in
find_file_and_directory, both DW_AT_name and the line table would try to
start a subfile with the same filename_for_id, and there was no need for
watch_main_source_file_lossage - which is what we want:
[symtab-create] start_subfile: name = /dir/d/dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: found existing symtab with name_for_id /dir/d/dw2-compdir-oldgcc42.S (/dir/d/dw2-compdir-oldgcc42.S)
[symtab-create] start_subfile: name = dw2-compdir-oldgcc42.S, name_for_id = /dir/d/dw2-compdir-oldgcc42.S
[symtab-create] start_subfile: found existing symtab with name_for_id /dir/d/dw2-compdir-oldgcc42.S (/dir/d/dw2-compdir-oldgcc42.S)
But since the one with name == "/dir/d/dw2-compdir-oldgcc42.S", coming
from DW_AT_name, gets created first, it wins, and the symtab ends up
with "/dir/d/dw2-compdir-oldgcc42.S" as the name, "info source" shows
"/dir/d/dw2-compdir-oldgcc42.S" and the test breaks.
This is not wrong per-se, after all DW_AT_name is
"/dir/d/dw2-compdir-oldgcc42.S", so it wouldn't be wrong to report the
current source file as "/dir/d/dw2-compdir-oldgcc42.S". If you compile
a file passing "/an/absolute/path.c", DW_AT_name typically contains (at
least with GCC) "/an/absolute/path.c" and GDB tells you that the source
file is "/an/absolute/path.c". But we can also keep the existing
behavior fairly easily with a little change in find_file_and_directory.
When extracting an artificial DW_AT_comp_dir from DW_AT_name, we now
modify the name to just keep the file part. The result is coherent with
what compilers do when you compile a file by just passing its filename
("gcc path.c -g"):
DW_AT_name ("path.c")
DW_AT_comp_dir ("/home/simark/build/binutils-gdb-one-target/gdb")
With this change, filename_for_id is still the full name,
"/dir/d/dw2-compdir-oldgcc42.S", but the filename of the subfile /
symtab (what ends up shown by "info source") is just
"dw2-compdir-oldgcc42.S", and that makes the test happy.
Change-Id: I8b5cc4bb3052afdb172ee815c051187290566307
|
|
In the following patch, there will be some callers of file_file_name
that will already have access to the file_entry object for which they
want the file name. It would be inefficient to have them pass an index,
only for line_header::file_file_name to re-lookup the same file_entry
object. Change line_header::file_file_name to accept a file_entry
object reference, instead of an index to look up.
I think this change makes sense in any case. Callers that have an index
can first obtain a file_entry using line_header::file_name_at or
line_header::file_names.
When passing a file_entry object, we can assume that the file_entry's
index is valid, unlike when passing an index. So, push the special case
about an invalid index to the sole current caller of file_file_name,
macro_start_file. I think that error belongs there anyway, since it
specifically talks about "bad file number in macro information".
This requires recording the file index in the file_entry structure, so
add that.
Change-Id: Ic6e44c407539d92b7863d7ba82405ade17f384ad
|
|
The following patch changes line_header::file_file_name to prepend the
compilation directory to the file name, if needed. For that, the line
header needs to know about the compilation directory. Prepare for that
by adding a constructor that takes it as a parameter, and passing the
value down everywhere needed. Add a second constructor for the special
case of building a line_header for doing a hash table lookup, since that
case doesn't require a compilation directory value.
Change-Id: Iba3ba0293e4e2d13a64b257cf9a3094684d54330
|
|
Add a few debug prints in buildsym.c that were helpful to me in writing
this series.
Change-Id: If10a818feaee3ce1b78a2a254013b62dd578002b
|
|
Introduce symtab_create_debug_printf and symtab_create_debug_printf_v,
to print the debug messages enabled by "set debug symtab-create".
Change-Id: I442500903f72d4635c2dd9eaef770111f317dc04
|
