The original reproducer for PR28030 required use of a specific
compiler version - gcc-c++-11.1.1-3.fc34 is mentioned in the PR,
though it seems probable that other gcc versions might also be able to
reproduce the bug as well. This commit introduces a test case which,
using the DWARF assembler, provides a reproducer which is independent
of the compiler version. (Well, it'll work with whatever compilers
the DWARF assembler works with.)
To the best of my knowledge, it's also the first test case which uses
the DWARF assembler to provide debug info for a shared object. That
being the case, I provided more than the usual commentary which should
allow this case to be used as a template when a combo shared
library / DWARF assembler test case is required in the future.
I provide some details regarding the bug in a comment near the
beginning of locexpr-dml.exp.
This problem was difficult to reproduce; I found myself constantly
referring to the backtrace while trying to figure out what (else) I
might be missing while trying to create a reproducer. Below is a
partial backtrace which I include for posterity.
#0 internal_error (
file=0xc50110 "/ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c", line=5575,
fmt=0xc520c0 "Unexpected type field location kind: %d")
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdbsupport/errors.cc:51
#1 0x00000000006ef0c5 in copy_type_recursive (objfile=0x1635930,
type=0x274c260, copied_types=0x30bb290)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c:5575
#2 0x00000000006ef382 in copy_type_recursive (objfile=0x1635930,
type=0x274ca10, copied_types=0x30bb290)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/gdbtypes.c:5602
#3 0x0000000000a7409a in preserve_one_value (value=0x24269f0,
objfile=0x1635930, copied_types=0x30bb290)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/value.c:2529
#4 0x000000000072012a in gdbscm_preserve_values (
extlang=0xc55720 <extension_language_guile>, objfile=0x1635930,
copied_types=0x30bb290)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/guile/scm-value.c:94
#5 0x00000000006a3f82 in preserve_ext_lang_values (objfile=0x1635930,
copied_types=0x30bb290)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/extension.c:568
#6 0x0000000000a7428d in preserve_values (objfile=0x1635930)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/value.c:2579
#7 0x000000000082d514 in objfile::~objfile (this=0x1635930,
__in_chrg=<optimized out>)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:549
#8 0x0000000000831cc8 in std::_Sp_counted_ptr<objfile*, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x1654580)
at /usr/include/c++/11/bits/shared_ptr_base.h:348
#9 0x00000000004e6617 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release (this=0x1654580) at /usr/include/c++/11/bits/shared_ptr_base.h:168
#10 0x00000000004e1d2f in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x190bb88, __in_chrg=<optimized out>)
at /usr/include/c++/11/bits/shared_ptr_base.h:705
#11 0x000000000082feee in std::__shared_ptr<objfile, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x190bb80, __in_chrg=<optimized out>)
at /usr/include/c++/11/bits/shared_ptr_base.h:1154
#12 0x000000000082ff0a in std::shared_ptr<objfile>::~shared_ptr (
this=0x190bb80, __in_chrg=<optimized out>)
at /usr/include/c++/11/bits/shared_ptr.h:122
#13 0x000000000085ed7e in __gnu_cxx::new_allocator<std::_List_node<std::shared_ptr<objfile> > >::destroy<std::shared_ptr<objfile> > (this=0x114bc00,
__p=0x190bb80) at /usr/include/c++/11/ext/new_allocator.h:168
#14 0x000000000085e88d in std::allocator_traits<std::allocator<std::_List_node<std::shared_ptr<objfile> > > >::destroy<std::shared_ptr<objfile> > (__a=...,
__p=0x190bb80) at /usr/include/c++/11/bits/alloc_traits.h:531
#15 0x000000000085e50c in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::_M_erase (this=0x114bc00, __position=
std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x1635930})
at /usr/include/c++/11/bits/stl_list.h:1925
#16 0x000000000085df0e in std::__cxx11::list<std::shared_ptr<objfile>, std::allocator<std::shared_ptr<objfile> > >::erase (this=0x114bc00, __position=
std::shared_ptr<objfile> (expired, weak count 1) = {get() = 0x1635930})
at /usr/include/c++/11/bits/list.tcc:158
#17 0x000000000085c748 in program_space::remove_objfile (this=0x114bbc0,
objfile=0x1635930)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/progspace.c:210
#18 0x000000000082d3ae in objfile::unlink (this=0x1635930)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:487
#19 0x000000000082e68c in objfile_purge_solibs ()
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/objfiles.c:875
#20 0x000000000092dd37 in no_shared_libraries (ignored=0x0, from_tty=1)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/solib.c:1236
#21 0x00000000009a37fe in target_pre_inferior (from_tty=1)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/target.c:2496
#22 0x00000000007454d6 in run_command_1 (args=0x0, from_tty=1,
run_how=RUN_NORMAL)
at /ironwood1/sourceware-git/f34-pr28030/bld/../../worktree-pr28030/gdb/infcmd.c:437
I'll note a few points regarding this backtrace:
Frame #1 is where the internal error occurs. It's caused by an
unhandled case for FIELD_LOC_KIND_DWARF_BLOCK. The fix for this bug
adds support for this case.
Frame #22 - it's a partial backtrace - shows that GDB is attempting to
(re)run the program. You can see the exact command sequence that was
used for reproducing this problem in the PR (at
https://sourceware.org/bugzilla/show_bug.cgi?id=28030), but in a
nutshell, after starting the program and advancing to the appropriate
source line, GDB was asked to step into libstdc++; a "finish" command
was issued, returning a value. The fact that a value was returned is
very important. GDB was then used to step back into libstdc++. A
breakpoint was set on a source line in the library after which a "run"
command was issued.
Frame #19 shows a call to objfile_purge_solibs. It's aptly named.
Frame #7 is a call to the destructor for one of the objfile solibs; it
turned out to be the one for libstdc++.
Frames #6 thru #3 show various value preservation frames. If you look
at preserve_values() in gdb/value.c, the value history is preserved
first, followed by internal variables, followed by values for the
extension languages (python and guile).
When building gdb with "-Wall -O2 -g -flto=auto", I run into:
...
(gdb) call clear_complaints()^M
No symbol "clear_complaints" in current context.^M
(gdb) FAIL: gdb.gdb/complaints.exp: clear complaints
...
The problem is that lto has optimized away clear_complaints, and consequently
the selftests cannot run.
Fix this by:
- using info function to detect presence of clear_complaints
- handling the absence of clear_complaints by calling untested
...
(gdb) UNTESTED: gdb.gdb/complaints.exp: \
Cannot find clear_complaints, skipping test
...
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-09-03 Tom de Vries <tdevries@suse.de>
* gdb.gdb/complaints.exp: Use untested if clear_complaints cannot
be found.
Values of type _Float16 and _Float16 _Complex can now be used on CPUs with
AVX512-FP16 support. Return values of those types are located in XMM0.
Compiler support for gcc and clang is in progress, see e.g.:
https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574117.html
gdb/ChangeLog:
2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* amd64-tdep.c (amd64_classify): Classify _Float16 and
_Float16 _Complex as AMD64_SSE.
* i386-tdep.c (i386_extract_return_value): Read _Float16 and
_Float16 _Complex from xmm0.
gdb/testsuite/ChangeLog:
2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* gdb.arch/x86-avx512fp16-abi.c: New file.
* gdb.arch/x86-avx512fp16-abi.exp: New file.
This adds support for the half datatype, FP16, to the AVX512 register printing.
gdb/ChangeLog:
2020-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* i386-tdep.c (i386_zmm_type) <v32_half>: New field.
(i386_ymm_type) <v16_half>: New field.
(i386_gdbarch_init): Add set_gdbarch_half_format.
* features/i386/64bit-avx512.xml: Add half type.
* features/i386/64bit-avx512.c: Regenerated.
* features/i386/64bit-sse.xml: Add half type.
* features/i386/64bit-sse.c: Regenerated.
gdb/testsuite/ChangeLog:
2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* gdb.arch/x86-avx512fp16.c: New file.
* gdb.arch/x86-avx512fp16.exp: New file.
* lib/gdb.exp (skip_avx512fp16_tests): New function.
Values of type bfloat16 can also be used on 32-bit targets, which was missed
in the original enablement. This also adjusts the testcase to pass with
"unix/-m32", where only the lower 8 AVX registers are available.
gdb/ChangeLog:
2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* features/i386/32bit-sse.xml: Add bfloat16 type.
* features/i386/32bit-sse.c: Regenerated.
gdb/testsuite/ChangeLog:
2021-07-21 Felix Willgerodt <Felix.Willgerodt@intel.com>
* gdb.arch/x86-avx512bf16.exp: Only use x/z/ymm 0-7.
When building gdb with "-Wall -O2 -g -flto=auto", I run into:
...
FAIL: gdb.gdb/python-helper.exp: breakpoint in captured_main \
(got interactive prompt)
FAIL: gdb.gdb/python-helper.exp: run until breakpoint at captured_main
WARNING: Couldn't test self
...
and similar in gdb.gdb/selftest.exp.
The first FAIL in more detail:
...
(gdb) break captured_main^M
Function "captured_main" not defined.^M
Make breakpoint pending on future shared library load? (y or [n]) n^M
(gdb) FAIL: gdb.gdb/python-helper.exp: breakpoint in captured_main \
(got interactive prompt)
...
The problem is that lto has optimized away the captured_main function
and consequently the selftests dependent on that cannot run.
Fix this by:
- using gdb_breakpoint to detect failure to set the breakpoint
- handling the failure to set a breakpoint by calling untested
- not emitting the warning if we've already got untested
such that we have:
...
(gdb) UNTESTED: gdb.gdb/python-helper.exp: Cannot set breakpoint at \
captured_main, skipping testcase.
...
gdb/testsuite/ChangeLog:
2021-09-02 Tom de Vries <tdevries@suse.de>
* lib/selftest-support.exp: Emit untested when not being able to set
breakpoint.
Fixes a few issues:
1) If you use "-fsanitize=address,undefined" in CFLAGS, the Makefile
attempt to trim off -fsanitize options left us with ",undefined".
2) ld_compile adds CFLAGS_FOR_TARGET itself, no need to pass it.
3) CFLAGS might be needed linking bootstrap test.
* Makefile.am (CFLAGS_FOR_TARGET, CXXFLAGS_FOR_TARGET): Trim off
all -fsanitize=*.
* Makefile.in: Regenerate.
* testsuite/ld-bootstrap/bootstrap.exp: Use CFLAGS when linking.
* testsuite/ld-cdtest/cdtest.exp: Use CFLAGS_FOR_TARGET when
linking.
* testsuite/ld-auto-import/auto-import.exp: Don't pass
CFLAGS_FOR_TARGET to ld_compile.
* testsuite/ld-cygwin/exe-export.exp: Likewise.
* testsuite/ld-elfvers/vers.exp: Likewise.
* testsuite/ld-elfvsb/elfvsb.exp: Likewise.
* testsuite/ld-elfweak/elfweak.exp: Likewise.
* testsuite/ld-gc/gc.exp: Likewise.
* testsuite/ld-pe/pe-compile.exp: Likewise.
* testsuite/ld-pe/pe-run.exp: Likewise.
* testsuite/ld-pe/pe-run2.exp: Likewise.
* testsuite/ld-plugin/plugin.exp: Likewise.
* testsuite/ld-shared/shared.exp: Likewise.
* testsuite/ld-elfcomm/elfcomm.exp: Likewise, and don't allow
nios2 testing to trash CFLAGS_FOR_TARGET.
* testsuite/ld-scripts/crossref.exp: Don't pass options in
CC_FOR_TARGET, do so in CFLAGS_FOR_TARGET instead.
* testsuite/ld-srec/srec.exp: Likewise, and for CXX.
The top level Makefile, the ld Makefile and others, define
CC_FOR_TARGET to be a compiler for the binutils target machine. This
is the compiler that should be used for almost all tests with C
source. There are _FOR_TARGET versions of CFLAGS, CXX, and CXXFLAGS
too. This was all supposed to work with the testsuite .exp files
using CC for the target compiler, and CC_FOR_HOST for the host
compiler, with the makefiles passing CC=$CC_FOR_TARGET and
CC_FOR_HOST=$CC to the runtest invocation.
One exception to the rule of using CC_FOR_TARGET is the native-only ld
bootstrap test, which uses the newly built ld to link a copy of
itself. Since the files being linked were created with the host
compiler, the boostrap test should use CC and CFLAGS, in case some
host compiler option provides needed libraries automatically.
However, bootstrap.exp used CC where it should have used CC_FOR_HOST.
I set about fixing that problem, then decided that playing games in
the makefiles with CC was a bad idea. Not only is it confusing, but
other dejagnu code knows about CC_FOR_TARGET. See dejagnu/target.exp.
So this patch gets rid of the makefile variable renaming and changes
all the .exp files to use the correct _FOR_TARGET variables.
CC_FOR_HOST and CFLAGS_FOR_HOST disappear. A followup patch will
correct bootstrap.exp to use CFLAGS, and a number of other things I
noticed.
binutils/
* testsuite/lib/binutils-common.exp (run_dump_test): Use
CC_FOR_TARGET and CFLAGS_FOR_TARGET rather than CC and CFLAGS.
ld/
* Makefile.am (check-DEJAGNU): Don't set CC to CC_FOR_TARGET
and similar. Pass variables with unchanged names. Don't set
CC_FOR_HOST or CFLAGS_FOR_HOST.
* Makefile.in: Regenerate.
* testsuite/config/default.exp: Update default CC and similar.
(compiler_supports, plug_opt): Use CC_FOR_TARGET.
* testsuite/ld-cdtest/cdtest.exp: Replace all uses of CC with
CC_FOR_TARGET, and similarly for CFLAGS, CXX and CXXFLAGS.
* testsuite/ld-auto-import/auto-import.exp: Likewise.
* testsuite/ld-cygwin/exe-export.exp: Likewise.
* testsuite/ld-elf/dwarf.exp: Likewise.
* testsuite/ld-elf/indirect.exp: Likewise.
* testsuite/ld-elf/shared.exp: Likewise.
* testsuite/ld-elfcomm/elfcomm.exp: Likewise.
* testsuite/ld-elfvers/vers.exp: Likewise.
* testsuite/ld-elfvsb/elfvsb.exp: Likewise.
* testsuite/ld-elfweak/elfweak.exp: Likewise.
* testsuite/ld-gc/gc.exp: Likewise.
* testsuite/ld-ifunc/ifunc.exp: Likewise.
* testsuite/ld-mn10300/mn10300.exp: Likewise.
* testsuite/ld-pe/pe-compile.exp: Likewise.
* testsuite/ld-pe/pe-run.exp: Likewise.
* testsuite/ld-pe/pe-run2.exp: Likewise.
* testsuite/ld-pie/pie.exp: Likewise.
* testsuite/ld-plugin/lto.exp: Likewise.
* testsuite/ld-plugin/plugin.exp: Likewise.
* testsuite/ld-scripts/crossref.exp: Likewise.
* testsuite/ld-selective/selective.exp: Likewise.
* testsuite/ld-sh/sh.exp: Likewise.
* testsuite/ld-shared/shared.exp: Likewise.
* testsuite/ld-srec/srec.exp: Likewise.
* testsuite/ld-undefined/undefined.exp: Likewise.
* testsuite/ld-unique/unique.exp: Likewise.
* testsuite/ld-x86-64/tls.exp: Likewise.
* testsuite/lib/ld-lib.exp: Likewise.
libctf/
* Makefile.am (check-DEJAGNU): Don't set CC to CC_FOR_TARGET.
Pass CC and CC_FOR_TARGET. Don't set CC_FOR_HOST.
* Makefile.in: Regenerate.
* testsuite/config/default.exp: Update default CC and similar.
* testsuite/lib/ctf-lib.exp (run_native_host_cmd): Use CC rather
than CC_FOR_HOST.
(run_lookup_test): Use CC_FOR_TARGET and CFLAGS_FOR_TARGET.
--enable-maintainer-mode showed a number of files needing to be
regenerated, and in the case of ld/Makefile.in that the file was
regenerated by hand. Nothing to see here really.
ld/
* Makefile.am (ALL_64_EMULATION_SOURCES): Sort haiku entry.
* Makefile.in: Regenerate.
* po/BLD-POTFILES.in: Regenerate.
libctf/
* configure: Regenerate.
zlib/
* configure: Regenerate.
to match GNU ld.
gold/
* archive.cc (Library_base::should_include_member): Don't handle
--export-dynamic-symbol.
* symtab.cc (Symbol_table::do_add_undefined_symbols_from_command_line):
Likewise.
PR 28292
gas * config/tc-v850.c (handle_lo16): Also accept
BFD_RELOC_V850_LO16_SPLIT_OFFSET.
* testsuite/gas/v850/split-lo16.s: Add extra line.
* testsuite/gas/v850/split-lo16.d: Update expected disassembly.
opcodes * v850-opc.c (D16): Use BFD_RELOC_V850_LO16_SPLIT_OFFSET in place
of BFD_RELOC_16.
.symtab_shndx section contents is an array, one entry for each symbol
in .symtab, present when the number of symbols exceeds a little less
than 64k. Since the mapping is 1-1 with symbols there is no need to
keep both dest_index and destshndx_index in elf_sym_strtab. Instead,
just make sure that the shndx pointers to the swap functions are kept
NULL when .symtab_shndx does not exist. Also, strtabcount in the
linker's elf hash table is incremented in lock-step with the output
symcount, so that can disappear too.
This defines a couple of macros used to avoid ubsan complaints about
calculations involving NULL pointers. PTR_ADD should be used in the
case where it is known that the offset is always zero with a NULL
pointer, and you'd like to know if a non-zero offset is ever used.
NPTR_ADD should be rarely used, but is defined for cases where a
non-zero offset is expected and should be ignored if the pointer is
NULL.
bfd/
* bfd-in.h (PTR_ADD, NPTR_ADD): Define.
* bfd-in2.h: Regenerate.
* elf-eh-frame.c (adjust_eh_frame_local_symbols): Avoid NULL
pointer calculations.
* elflink.c (_bfd_elf_strip_zero_sized_dynamic_sections): Likewise.
(bfd_elf_add_dt_needed_tag, elf_finalize_dynstr): Likewise.
(elf_link_add_object_symbols, elf_link_input_bfd): Likewise.
(bfd_elf_final_link, bfd_elf_gc_record_vtinherit): Likewise.
binutils/
* objdump.c (disassemble_section): Use PTR_ADD for rel_ppend.
Current ubsan complains on every use of __PTR_ALIGN (when ptrdiff_t is
as large as a pointer), due to making calculations relative to a NULL
pointer. This patch avoids the problem by extracting out and
simplifying __BPTR_ALIGN for the usual case. I've continued to use
ptrdiff_t here, where it might be better to throw away __BPTR_ALIGN
entirely and just assume uintptr_t exists.
* obstack.h (__PTR_ALIGN): Expand and simplify __BPTR_ALIGN
rather than calculating relative to a NULL pointer.
[ Using $build for /home/vries/gdb_versions/devel/build to make things a bit
more readable. ]
When using make check// to run test-case gdb.dwarf2/fission-base.exp:
...
( cd $build/gdb; make check//unix RUNTESTFLAGS="fission-base.exp" )
...
we run into:
...
(gdb) file \
$build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base^M
Reading symbols from \
$build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base...^M
warning: Could not find DWO CU \
$build/gdb/testsuite.1/outputs/gdb.dwarf2/fission-base/fission-base.dwo \
(0x807060504030201) referenced by CU at offset 0xc7 [in module \
$build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base]^M
...
The problem is that the executable refers to the dwo file using path name
$build/gdb/testsuite.1/outputs/gdb.dwarf2/fission-base/fission-base.dwo,
while the actual dwo file is at
$build/gdb/testsuite.unix/outputs/gdb.dwarf2/fission-base/fission-base.dwo.
This is caused by this trick in fission-base.S:
...
#define XSTR(s) STR(s)
#define STR(s) #s
...
.asciz XSTR(DWO) # DW_AT_GNU_dwo_name
...
and:
...
$ echo | gcc -E -dD - | grep "define unix"
...
I used this trick to avoid doing additional_flags=-DDWO=\"$dwo\", since I was
concerned that there could be quoting issues.
However, I've found other uses of this pattern, f.i. in
gdb/testsuite/gdb.base/corefile-buildid.exp:
...
additional_flags=-DSHLIB_NAME=\"$dlopen_lib\"]
...
So, fix this by:
- using additional_flags=-DDWO=\"$dwo\" and
- using plain DWO instead of XSTR(DWO)
Likewise in other gdb.dwarf2/fission*.exp test-cases.
Tested on x86_64-linux, using make check//unix.
gdb/testsuite/ChangeLog:
2021-09-01 Tom de Vries <tdevries@suse.de>
PR testsuite/28298
* gdb.dwarf2/fission-base.S: Use DWO instead of XSTR(DWO).
* gdb.dwarf2/fission-loclists-pie.S: Same.
* gdb.dwarf2/fission-loclists.S: Same.
* gdb.dwarf2/fission-reread.S: Same.
* gdb.dwarf2/fission-base.exp: Use additional_flags=-DDWO=\"$dwo\".
* gdb.dwarf2/fission-loclists-pie.exp: Same.
* gdb.dwarf2/fission-loclists.exp: Same.
* gdb.dwarf2/fission-reread.exp: Same.
On openSUSE Tumbleweed I ran into:
...
(gdb) ptype outstring_func.part^M
No symbol "outstring_func" in current context.^M
(gdb) FAIL: gdb.fortran/call-no-debug.exp: ptype outstring_func.part
...
while on openSUSE Leap 15.2 I have instead:
...
(gdb) ptype string_func_^M
type = <unknown return type> ()^M
(gdb) PASS: gdb.fortran/call-no-debug.exp: ptype string_func_
...
The difference is caused by the result for "info function string_func", which
is this for the latter:
...
(gdb) info function string_func^M
All functions matching regular expression "string_func":^M
^M
Non-debugging symbols:^M
0x000000000040089c string_func_^M
...
but this for the former:
...
(gdb) info function string_func^M
All functions matching regular expression "string_func":^M
^M
Non-debugging symbols:^M
0x00000000004012bb string_func_^M
0x00007ffff7bac5b0 outstring_func.part^M
0x00007ffff7bb1a00 outstring_func.part^M
...
The extra symbols are part of glibc:
...
$ nm /lib64/libc.so.6 | grep string_func
00000000000695b0 t outstring_func.part.0
000000000006ea00 t outstring_func.part.0
...
If glibc debug info is installed, we get instead:
...
(gdb) info function string_func^M
All functions matching regular expression "string_func":^M
^M
File /usr/src/debug/glibc-2.33-9.1.x86_64/stdio-common/vfprintf-internal.c:^M
236: static int outstring_func(int, size_t, const unsigned int *, FILE *);^M
^M
File vfprintf-internal.c:^M
236: static int outstring_func(int, size_t, const unsigned char *, FILE *);^M
^M
Non-debugging symbols:^M
0x00000000004012bb string_func_^M
...
and the FAIL doesn't trigger.
Fix this by calling "info function string_func" before starting the exec, such
that only symbols of the exec are taken into account.
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-09-01 Tom de Vries <tdevries@suse.de>
* gdb.fortran/call-no-debug.exp: Avoid shared lib symbols for
find_mangled_name calls.
Remove breakpoint_find_if, replace its sole usage with using
all_breakpoints directly instead. At the same time, change return
types to use bool.
Change-Id: I9ec392236b4804b362d16ab563330b9c07311106
The .insn directive can let users use their own instructions, or
some new instruction, which haven't supported in the old binutils.
For example, if users want to use sifive cache instruction, they
cannot just write "cflush.d1.l1" in the assembly code, they should
use ".insn i SYSTEM, 0, x0, x10, -0x40". But the .insn directive
may not easy to use for some cases, and not so friendly to users.
Therefore, I believe most of the users will use ".word 0xfc050073",
to encode the instructions directly, rather than use .insn. But
once we have supported the mapping symbols, the .word directives
are marked as data, so disassembler won't dump them as instructions
as usual. I have discussed this with Kito many times, we all think
extend the .insn direcitve to support the hardcode encoding, is the
easiest way to resolve the problem. Therefore, there are two more
.insn formats are proposed as follows,
(original) .insn <type>, <operand1>, <operand2>, ...
.insn <insn-length>, <value>
.insn <value>
The <type> is string, and the <insn-length> and <value> are constants.
gas/
* config/tc-riscv.c (riscv_ip_hardcode): Similar to riscv_ip,
but assembles an instruction according to the hardcode values
of .insn directive.
* doc/c-riscv.texi: Document two new .insn formats.
* testsuite/gas/riscv/insn-fail.d: New testcases.
* testsuite/gas/riscv/insn-fail.l: Likewise.
* testsuite/gas/riscv/insn-fail.s: Likewise.
* testsuite/gas/riscv/insn.d: Updated.
* testsuite/gas/riscv/insn.s: Likewise.
gdbfmt was already used for printf_filtered, so using it for
vprintf_filtered is more consistent.
As a result, all callers of vfprintf_maybe_filtered now use gdbfmt, so
the function can be simplified to assume the gdbfmt case and remove
the associated bool argument. Similary, vprintf_filtered is now a
simple wrapper around vfprintf_filtered.
The handler for 'info proc status' for native processes on FreeBSD
uses the 'j' size modifier along with uintmax_t / intmax_t casts to
output integer values for types such as off_t that are not aliases of
a basic C type such as 'int' or 'long'. printf_filtered does not
support the 'j' modifer, so this resulted in runtime errors in
practice:
(gdb) info proc stat
process 8674
Name: ls
State: T (stopped)
Parent process: 8673
Process group: 8674
Session id: 2779
Unrecognized format specifier 'j' in printf
Instead, use plongest and pulongest to generate the output strings of
these integer values.
We get this error when building GDB on some platforms. I get it using
g++-10 on Ubuntu 20.04 (installed using the distro package). It was
also reported by John Baldwin, using a clang that uses libc++.
CXX unittests/parallel-for-selftests.o
cc1plus: warning: command line option '-Wmissing-prototypes' is valid for C/ObjC but not for C++
/home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c: In function 'void selftests::parallel_for::test(int)':
/home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c:53:30: error: use of deleted function 'std::atomic<int>::atomic(const std::atomic<int>&)'
53 | std::atomic<int> counter = 0;
| ^
In file included from /usr/include/c++/9/future:42,
from /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/thread-pool.h:29,
from /home/smarchi/src/binutils-gdb/gdb/../gdbsupport/parallel-for.h:26,
from /home/smarchi/src/binutils-gdb/gdb/unittests/parallel-for-selftests.c:22:
/usr/include/c++/9/atomic:755:7: note: declared here
755 | atomic(const atomic&) = delete;
| ^~~~~~
/usr/include/c++/9/atomic:759:17: note: after user-defined conversion: 'constexpr std::atomic<int>::atomic(std::atomic<int>::__integral_type)'
759 | constexpr atomic(__integral_type __i) noexcept : __base_type(__i) { }
| ^~~~~~
I haven't dug to know why it does not happen everywhere, but this patch
fixes it by using the constructor to initialize the variable, rather
than the assignment operator.
Change-Id: I6b27958171bf6187f6a875657395fd10441db7e6
* According to PR28291, we get the following unexpected gdb behavior,
(gdb) disassemble 0x0,+4
Dump of assembler code from 0x0 to 0x4:
0x0000000000000000:
0x0000000000000001:
0x0000000000000002:
0x0000000000000003:
End of assembler dump.
* This patch should fix it to the right behavior,
(gdb) disassemble 0x0,+4
Dump of assembler code from 0x0 to 0x4:
0x0000000000000000: Cannot access memory at address 0x0
opcodes/
pr 28291
* riscv-dis.c (print_insn_riscv): Return STATUS if it is not zero.
Tom de Vries noticed that a patch in the DWARF scanner rewrite series
caused a regression in parallel_for_each -- it started crashing in the
case where the number of threads is 0 (there was an unchecked use of
"n-1" that was used to size an array).
He also pointed out that there were no tests of parallel_for_each.
This adds a few tests of parallel_for_each, primarily testing that
different settings for the number of threads will work. This test
catches the bug that he found in that series.
I wanted to see how many threads gdb thought it was using, but
"maint show worker-threads" only reported "unlimited". This patch
adds a show function so that it will now report the number of threads
gdb has started.
Regression tested on x86-64 Fedora 34.
With current gdb we run into:
...
$ gdb -batch '' ''
: No such file or directory.
pathstuff.cc:132: internal-error: \
gdb::unique_xmalloc_ptr<char> gdb_abspath(const char*): \
Assertion `path != NULL && path[0] != '\0'' failed.
...
Fix this by skipping the call to gdb_abspath in core_target_open in the
empty-string case, such that we have instead:
...
$ gdb -batch '' ''
: No such file or directory.
: No such file or directory.
$
...
Tested on x86_64-linux.
gdb/ChangeLog:
2021-08-30 Tom de Vries <tdevries@suse.de>
PR cli/28290
* gdb/corelow.c (core_target_open): Skip call to gdb_abspath in the
empty-string case.
gdb/testsuite/ChangeLog:
2021-08-30 Tom de Vries <tdevries@suse.de>
PR cli/28290
* gdb.base/batch-exit-status.exp: Add gdb '' and gdb '' '' tests.
Similar to ARM/AARCH64, we add mapping symbols in the symbol table,
to mark the start addresses of data and instructions. The $d means
data, and the $x means instruction. Then the disassembler uses these
symbols to decide whether we should dump data or instruction.
Consider the mapping-04 test case,
$ cat tmp.s
.text
.option norelax
.option norvc
.fill 2, 4, 0x1001
.byte 1
.word 0
.balign 8
add a0, a0, a0
.fill 5, 2, 0x2002
add a1, a1, a1
.data
.word 0x1 # No need to add mapping symbols.
.word 0x2
$ riscv64-unknown-elf-as tmp.s -o tmp.o
$ riscv64-unknown-elf-objdump -d tmp.o
Disassembly of section .text:
0000000000000000 <.text>:
0: 00001001 .word 0x00001001 # Marked $d, .fill directive.
4: 00001001 .word 0x00001001
8: 00000001 .word 0x00000001 # .byte + part of .word.
c: 00 .byte 0x00 # remaining .word.
d: 00 .byte 0x00 # Marked $d, odd byte of alignment.
e: 0001 nop # Marked $x, nops for alignment.
10: 00a50533 add a0,a0,a0
14: 20022002 .word 0x20022002 # Marked $d, .fill directive.
18: 20022002 .word 0x20022002
1c: 2002 .short 0x2002
1e: 00b585b3 add a1,a1,a1 # Marked $x.
22: 0001 nop # Section tail alignment.
24: 00000013 nop
* Use $d and $x to mark the distribution of data and instructions.
Alignments of code are recognized as instructions, since we usually
fill nops for them.
* If the alignment have odd bytes, then we cannot just fill the nops
into the spaces. We always fill an odd byte 0x00 at the start of
the spaces. Therefore, add a $d mapping symbol for the odd byte,
to tell disassembler that it isn't an instruction. The behavior
is same as Arm and Aarch64.
The elf/linux toolchain regressions all passed. Besides, I also
disable the mapping symbols internally, but use the new objudmp, the
regressions passed, too. Therefore, the new objudmp should dump
the objects corretly, even if they don't have any mapping symbols.
bfd/
pr 27916
* cpu-riscv.c (riscv_elf_is_mapping_symbols): Define mapping symbols.
* cpu-riscv.h: extern riscv_elf_is_mapping_symbols.
* elfnn-riscv.c (riscv_maybe_function_sym): Do not choose mapping
symbols as a function name.
(riscv_elf_is_target_special_symbol): Add mapping symbols.
binutils/
pr 27916
* testsuite/binutils-all/readelf.s: Updated.
* testsuite/binutils-all/readelf.s-64: Likewise.
* testsuite/binutils-all/readelf.s-64-unused: Likewise.
* testsuite/binutils-all/readelf.ss: Likewise.
* testsuite/binutils-all/readelf.ss-64: Likewise.
* testsuite/binutils-all/readelf.ss-64-unused: Likewise.
gas/
pr 27916
* config/tc-riscv.c (make_mapping_symbol): Create a new mapping symbol.
(riscv_mapping_state): Decide whether to create mapping symbol for
frag_now. Only add the mapping symbols to text sections.
(riscv_add_odd_padding_symbol): Add the mapping symbols for the
riscv_handle_align, which have odd bytes spaces.
(riscv_check_mapping_symbols): Remove any excess mapping symbols.
(md_assemble): Marked as MAP_INSN.
(riscv_frag_align_code): Marked as MAP_INSN.
(riscv_init_frag): Add mapping symbols for frag, it usually called
by frag_var. Marked as MAP_DATA for rs_align and rs_fill, and
marked as MAP_INSN for rs_align_code.
(s_riscv_insn): Marked as MAP_INSN.
(riscv_adjust_symtab): Call riscv_check_mapping_symbols.
* config/tc-riscv.h (md_cons_align): Defined to riscv_mapping_state
with MAP_DATA.
(TC_SEGMENT_INFO_TYPE): Record mapping state for each segment.
(TC_FRAG_TYPE): Record the first and last mapping symbols for the
fragments. The first mapping symbol must be placed at the start
of the fragment.
(TC_FRAG_INIT): Defined to riscv_init_frag.
* testsuite/gas/riscv/mapping-01.s: New testcase.
* testsuite/gas/riscv/mapping-01a.d: Likewise.
* testsuite/gas/riscv/mapping-01b.d: Likewise.
* testsuite/gas/riscv/mapping-02.s: Likewise.
* testsuite/gas/riscv/mapping-02a.d: Likewise.
* testsuite/gas/riscv/mapping-02b.d: Likewise.
* testsuite/gas/riscv/mapping-03.s: Likewise.
* testsuite/gas/riscv/mapping-03a.d: Likewise.
* testsuite/gas/riscv/mapping-03b.d: Likewise.
* testsuite/gas/riscv/mapping-04.s: Likewise.
* testsuite/gas/riscv/mapping-04a.d: Likewise.
* testsuite/gas/riscv/mapping-04b.d: Likewise.
* testsuite/gas/riscv/mapping-norelax-04a.d: Likewise.
* testsuite/gas/riscv/mapping-norelax-04b.d: Likewise.
* testsuite/gas/riscv/no-relax-align.d: Updated.
* testsuite/gas/riscv/no-relax-align-2.d: Likewise.
include/
pr 27916
* opcode/riscv.h (enum riscv_seg_mstate): Added.
opcodes/
pr 27916
* riscv-dis.c (last_map_symbol, last_stop_offset, last_map_state):
Added to dump sections with mapping symbols.
(riscv_get_map_state): Get the mapping state from the symbol.
(riscv_search_mapping_symbol): Check the sorted symbol table, and
then find the suitable mapping symbol.
(riscv_data_length): Decide which data size we should print.
(riscv_disassemble_data): Dump the data contents.
(print_insn_riscv): Handle the mapping symbols.
(riscv_symbol_is_valid): Marked mapping symbols as invalid.
The current syntax of proc arange is:
...
proc arange { arange_start arange_length {comment ""} {seg_sel ""} } {
...
and a typical call looks like:
...
arange $start $len
...
This style is somewhat annoying because if you want to specify the last
parameter, you need to give the default values of all the other optional ones
before as well:
...
arange $start $len "" $seg_sel
...
Update the syntax to:
...
proc arange { options arange_start arange_length } {
parse_options {
{ comment "" }
{ seg_sel "" }
}
...
such that a typical call looks like:
...
arange {} $start $len
...
and a call using seg_sel looks like:
...
arange {
seg_sel $seg_sel
} $start $len
...
Also update proc aranges, which already has an options argument, to use the
new proc parse_options.
Tested on x86_64-linux.
Co-Authored-By: Simon Marchi <simon.marchi@polymtl.ca>
When comparing the sizes of the index files generated for shlib
outputs/gdb.dwarf2/dw2-zero-range/shr1.sl, I noticed a large difference
between .debug_names:
...
$ gdb -q -batch $shlib -ex "save gdb-index -dwarf-5 ."
$ du -b -h shr1.sl.debug_names shr1.sl.debug_str
61 shr1.sl.debug_names
0 shr1.sl.debug_str
...
and .gdb_index:
...
$ gdb -q -batch $shlib -ex "save gdb-index ."
$ du -b -h shr1.sl.gdb-index
8.2K shr1.sl.gdb-index
...
The problem is that the .gdb_index contains a non-empty symbol table with only
empty entries.
Fix this by making the symbol table empty, such that we have instead:
...
$ du -b -h shr1.sl.gdb-index
184 shr1.sl.gdb-index
...
Tested on x86_64-linux.
Before commit 5ef670d81f "[gdb/testsuite] Add dummy start and end CUs in
dwarf assembly" we had in exec outputs/gdb.dlang/watch-loc/watch-loc a D
compilation unit at offset 0xc7:
...
Compilation Unit @ offset 0xc7:
Length: 0x4c (32-bit)
Version: 4
Abbrev Offset: 0x64
Pointer Size: 8
<0><d2>: Abbrev Number: 2 (DW_TAG_compile_unit)
<d3> DW_AT_language : 19 (D)
...
with a corresponding .debug_aranges entry:
...
Offset into .debug_info: 0xc7
Pointer Size: 4
Segment Size: 0
Address Length
004004a7 0000000b
00000000 00000000
...
After that commit we have a dummy CU at offset 0xc7 and the D compilation unit
at offset 0xd2:
...
Compilation Unit @ offset 0xc7:
Length: 0x7 (32-bit)
Version: 4
Abbrev Offset: 0x64
Pointer Size: 8
Compilation Unit @ offset 0xd2:
Length: 0x4c (32-bit)
Version: 4
Abbrev Offset: 0x65
Pointer Size: 8
<0><dd>: Abbrev Number: 2 (DW_TAG_compile_unit)
<de> DW_AT_language : 19 (D)
...
while the .debug_aranges entry still points to 0xc7.
The problem is that the test-case uses a hack (quoting from
commit 75f06e9dc5):
...
[ Note: this is a non-trivial test-case. The file watch-loc-dw.S contains a
.debug_info section, but not an .debug_aranges section or any actual code.
The file watch-loc.c contains code and a .debug_aranges section, but no other
debug section. So, the intent for the .debug_aranges section in watch-loc.c
is to refer to a compilation unit in the .debug_info section in
watch-loc-dw.S. ]
...
and adding the dummy CU caused that hack to stop working.
Fix this by moving the generation of .debug_aranges from watch-loc.c to
watch-loc.exp, such that we have:
...
Offset into .debug_info: 0xd2
Pointer Size: 4
Segment Size: 0
Address Length
004004a7 0000000b
00000000 00000000
...
Tested on x86_64-linux.
A best practise for DWARF [1] is to generate .debug_aranges entries for CUs
even if they have no address range.
Generate .debug_arange entries for the dummy CUs added by the DWARF assembler.
Tested on x86_64-linux.
[1] http://wiki.dwarfstd.org/index.php?title=Best_Practices
A couple of test-cases fail when run with target board cc-with-debug-names due
to missing .debug_aranges entries for the CUs added by the dwarf assembler.
Add a .debug_aranges entry for those CUs.
Tested on x86_64-linux.
Add a proc aranges such that we can generate .debug_aranges sections in dwarf
assembly using:
...
cu { label cu_label } {
...
}
aranges {} cu_label {
arange $addr $len [<comment>] [$segment_selector]
}
...
Tested on x86_64-linux.
We can use current dwarf assembly infrastructure to declare a label that marks
the start of the CU header:
...
declare_labels header_start_cu_a
_section ".debug_info"
header_start_cu_a : cu {} {
}
_section ".debug_info"
header_start_cu_b : cu {} {
}
...
on the condition that we switch to the .debug_info section before, which makes
this style of use fragile.
Another way to achieve the same is to use the label as generated by the cu
proc itself:
...
variable _cu_label
cu {} {
}
set header_start_cu_a $_cu_label
cu {} {
}
set header_start_cu_b $_cu_label
...
but again that seems fragile given that adding a new CU inbetween will
silently result in the wrong value for the label.
Add a label option to proc cu such that we can simply do:
...
cu { label header_start_cu_a } {
}
cu { label header_start_cu_b } {
}
...
Tested on x86_64-linux.
