Fix collection and processing of autoprofile data for target libs
cc1, cc1plus, and lto built during STAGEautoprofile need to be built with
debug info since they are used to build target libs. -gtoggle was
turning off debug info for this stage.
create_gcov should be passed prev-gcc/cc1, prev-gcc/cc1plus, and prev-gcc/lto
instead of stage1-gcc/cc1, stage1-gcc/cc1plus, and stage1-gcc/lto when
processing profile data collected while building target libraries.
Tested on x86_64-pc-linux-gnu.
* Makefile.tpl: Remove -gtoggle for STAGEautoprofile
Collect both user and kernel events for autofdo tests and autoprofiledbootstrap
When we collect just user events for autofdo with lbr we get some events where branch
sources are kernel addresses and branch targets are user addresses. Without kernel MMAP
events create_gcov can't make sense of kernel addresses. Currently create_gcov fails if
it can't map at least 95% of events. We sometimes get below this threshold with just
user events. The change is to collect both user events and kernel events.
Tested on x86_64-pc-linux-gnu.
* Makefile.tpl: Collect both kernel and user events for autofdo
The D front-end is now itself written in D, in order to build GDC, you
will need a working GDC compiler (GCC version 9.1 or later).
GCC changes:
- Add support for bootstrapping the D front-end.
These add the required components in order to have a D front-end written
in D itself. Because the compiler front-end only depends on the core
runtime modules, only libdruntime is built for the bootstrap stages.
D front-end changes:
- Import dmd v2.098.0-beta.1.
Druntime changes:
- Import druntime v2.098.0-beta.1.
Phobos changes:
- Import phobos v2.098.0-beta.1.
The jump from v2.076.1 to v2.098.0 covers nearly 4 years worth of
development on the D programming language and run-time libraries.
* Makefile.def: Add bootstrap to libbacktrace, libphobos, zlib, and
libatomic.
* Makefile.tpl (POSTSTAGE1_HOST_EXPORTS): Fix command for GDC.
(STAGE1_CONFIGURE_FLAGS): Add --with-libphobos-druntime-only if
target-libphobos-bootstrap.
(STAGE2_CONFIGURE_FLAGS): Likewise.
Olivier Hainque [Mon, 7 Aug 2023 11:07:33 +0000 (13:07 +0200)]
Generic configury support for shared libs on VxWorks
This change adds the configury bits to activate the build of
shared libs on VxWorks ports configured with --enable-shared,
for libraries variants where this is generally supported (rtp,
code model !large - currently not compatible with -fPIC).
Set lt_cv_deplibs_check_method in libtool.m4, so the build of
libraries know how to establish dependencies. This is useful in
configurations such as aarch64 where proper support of LSE relies
on accurate dependency information between libstdc++ and libgcc_s
to begin with.
* libtool.m4 (*vxworks*): When enable_shared, set dynamic_linker
and friends for rtp !large. Assume the linker has the required
abilities and set lt_cv_deplibs_check_method.
Martin Storsjö [Mon, 7 Aug 2023 11:07:25 +0000 (13:07 +0200)]
mh-mingw: Set __USE_MINGW_ACCESS in missed C++ flags variables
This is similar to what was done in eea4e2ff0a3f5e7f37df204c070cc5d9ef339e6e (where it was added to
STAGE*_CXXFLAGS), but this adds the flag to the CXXFLAGS and
BOOT_CXXFLAGS variables too (as it's already added to CFLAGS and
BOOT_CFLAGS).
2021-04-09 Martin Storsjö <martin@martin.st>
config/
* mh-mingw: Set __USE_MINGW_ACCESS in missed C++ flags
variables
Iain Sandoe [Mon, 7 Aug 2023 11:07:24 +0000 (13:07 +0200)]
configure: Allow host fragments to react to --enable-host-shared.
This makes the host_shared value available to host makefile
fragments.
It uses this to adjust Darwin's mdynamic-no-pic in the case that
shared host resources are required.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
config/
* mh-darwin: Require a non-shared host configuration to
enable mdynamic-no-pic where that is supported.
Iain Sandoe [Mon, 7 Aug 2023 11:07:23 +0000 (13:07 +0200)]
Darwin, config: Revise host config fragment.
There were two uses for the Darwin host config fragment:
The first is to arrange for targets that support mdynamic-no-pic
to be built with that enabled (since it makes a significant
difference to the compiler performance). We can be more specific
in the application of this, since it only applies to 32b hosts
plus powerpc64-darwin9.
The second was to work around a tool bug where -fno-PIE was not
propagated to the link stage. This second use is redundant,
since the buggy toolchain cannot bootstrap current GCC sources
anyway.
This makes the host fragment more specific and reduces the number
of toolchains for which it is included which reduces clutter in
configure lines.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
config/
* mh-darwin: Make this specific to handling the
mdynamic-no-pic case.
Alexandre Oliva [Mon, 7 Aug 2023 11:07:18 +0000 (13:07 +0200)]
Add TFLAGS to gcc's GCC_FOR_TARGET
When the GCC build runs GCC_FOR_TARGET, e.g. for selftests or for
dumping specs, it doesn't use TFLAGS in non-bootstrap scenarios. This
patch arranges for TFLAGS to be passed from the top level down to gcc
in GCC_FOR_TARGET in this case.
* Makefile.tpl (HOST_EXPORTS): Add TFLAGS to GCC_FOR_TARGET.
(EXTRA_GCC_FLAGS): Likewise.
David Seifert [Mon, 7 Aug 2023 11:07:16 +0000 (13:07 +0200)]
configure: use OBJDUMP determined by libtool [PR95648]
$ac_cv_prog_OBJDUMP contains the --host OBJDUMP that
libtool has inferred. Current config/gcc-plugin.m4 does
not respect the user's choice for OBJDUMP.
config/
* gcc-plugin.m4: Use libtool's $ac_cv_prog_OBJDUMP.
Iain Sandoe [Mon, 7 Aug 2023 11:07:14 +0000 (13:07 +0200)]
configure, Darwin: Ensure overrides to host-pie are passed to gcc configure.
The latest versions of Darwin on the Aarch64 platform mandate PIE executables.
On x86_64 it remains optional, but produces tool warnings after Darwin20, so
we default to PIE executables there too.
All (non-PowerPC) 64b Darwin platforms mandate PIC code and therefore force
host_shared on (we issue a diagnostic if the user tries to configure them
non-shared).
However, this also means we cannot test the host_shared setting independently
of the host_pie setting so that the logic for setting PICFLAG must be amended
for Darwin.
For Darwin versions required to have PIE executables, in the event that the
user tries to configure these as --disable-host-pie, we issue a warning and
override the setting. These versions must also switch host_pie on even if it
is not given in the configure line. To cater for this we pass the current
value of host_pie, as determined by top-level configure, to the GCC configure.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
* Makefile.def: Pass the enable-host-pie value to GCC configure.
* configure.ac: Adjust the logic for shared and PIE host flags to
ensure that PIE is passed for hosts that require it.
Marek Polacek [Mon, 7 Aug 2023 11:07:12 +0000 (13:07 +0200)]
configure: Implement --enable-host-pie
This patch implements the --enable-host-pie configure option which
makes the compiler executables PIE. This can be used to enhance
protection against ROP attacks, and can be viewed as part of a wider
trend to harden binaries.
Co-Authored by: Iain Sandoe <iain@sandoe.co.uk>
* configure.ac (--enable-host-pie): New check. Set PICFLAG after this
check.
intl/
* configure.ac (--enable-host-shared): Don't set PICFLAG here.
(--enable-host-pie): New check. Set PICFLAG after this check.
libdecnumber/
* configure.ac (--enable-host-shared): Don't set PICFLAG here.
(--enable-host-pie): New check. Set PICFLAG after this check.
zlib/
* configure.ac (--enable-host-shared): Don't set PICFLAG here.
(--enable-host-pie): New check. Set PICFLAG after this check.
Iain Sandoe [Mon, 7 Aug 2023 11:07:11 +0000 (13:07 +0200)]
configure: When host-shared, pass --with-pic to in-tree lib configs.
If we are building PIC/PIE host executables, and we are building dependent
libs (e.g. GMP) in-tree those libs need to be configured to generate PIC code.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
* Makefile.def: Pass host_libs_picflag to host dependent library
configures.
* configure.ac (host_libs_picflag): New configure variable set to
'--with-pic' when building 'host_shared'.
Iain Sandoe [Mon, 7 Aug 2023 11:07:10 +0000 (13:07 +0200)]
configure: Do not build the ununsed libffi shared library.
We do not use the shared libffi libraray, nor do we install it.
However, on at least Darwin, the shared version will be picked
up for testing, so it is preferrable not to build it.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
* Makefile.def: Do not build shared libffi.
Iain Sandoe [Mon, 7 Aug 2023 11:07:09 +0000 (13:07 +0200)]
Darwin : Update libtool and dependencies for Darwin20 [PR97865]
The change in major version (and the increment from Darwin19 to 20)
caused libtool tests to fail which resulted in incorrect build settings
for shared libraries.
PR target/97865
* libtool.m4: Update handling of Darwin platform link flags
for Darwin20.
LoongArch always support clz and ctz instructions, so we can always use
__builtin_{clz,ctz} for count_{leading,trailing}_zeros. This improves
the code of libgcc, and also benefits Glibc once we merge longlong.h
there.
Bootstrapped and regtested on loongarch64-linux-gnu.
Lulu Cheng [Mon, 7 Aug 2023 11:07:05 +0000 (13:07 +0200)]
Libvtv: Add loongarch support.
The loongarch64 specification permits page sizes of 4KiB, 16KiB and 64KiB,
but only 16KiB pages are supported for now.
Co-Authored-By: qijingwen <qijingwen@loongson.cn>
include/
* vtv-change-permission.h (defined): Determines whether the macro
__loongarch_lp64 is defined
(VTV_PAGE_SIZE): Set VTV_PAGE_SIZE to 16KiB for loongarch64.
Carl Love [Fri, 11 Aug 2023 21:08:37 +0000 (17:08 -0400)]
gdb.ada/mi_var_access.exp
The NUMCHILD value for the "A_String_Access" test differs for X86 and
PowerPC. The patch substitutes $decimal instead of "1" to match the value
of NUMCHILD.
The test "-var-update A_String_Access" generates different output depending
on the value of VAROBJ_UPDATE_RESULT.TYPE_CHANGED. If the value is true,
the strings "new_type" and "new_num_children" are printed along with their
values.
The VAROBJ_UPDATE_RESULT.TYPE_CHANGED value is true on PowerPC which
produces the output:
Tom Tromey [Fri, 11 Aug 2023 16:04:49 +0000 (10:04 -0600)]
Test GNAT encodings in arr_acc_idx_w_gap.exp
While working on a GNAT bug, I wanted to also test
arr_acc_idx_w_gap.exp using GNAT encodings. When the GNAT change is
ready, I plan to add a new case here.
Tested on x86-64 Fedora 36. I am checking this in.
Tsukasa OI [Fri, 11 Aug 2023 03:12:52 +0000 (03:12 +0000)]
RISC-V: Reflect actual range of vlen for hashing
Before actual vlen handling, fix the riscv_gdbarch_features hashing
function based on the actual valid range of vlen. In bytes, vlen is 0,
or 4 <= xlen <= 8192.
Tsukasa OI [Fri, 11 Aug 2023 03:09:58 +0000 (03:09 +0000)]
RISC-V: Add reference to Zve32*
Before actual vlen handling, this commit fixes its description to allow vlen
less than 16 (but 4 or greater), to support vector subset extensions for
embedded environment ('Zve32*').
Alan Modra [Wed, 9 Aug 2023 00:28:36 +0000 (09:58 +0930)]
gdb: warn unused result for bfd IO functions
This fixes the compilation warnings introduced by my bfdio.c patch.
The removed bfd_seeks in coff_symfile_read date back to 1994, commit 7f4c859520, prior to which the file used stdio rather than bfd to read
symbols. Since it now uses bfd to read the file there should be no
need to synchronise to bfd's idea of the file position. I also fixed
a potential uninitialised memory access.
Jan Beulich [Fri, 11 Aug 2023 08:05:32 +0000 (10:05 +0200)]
gas: purge md_elf_section_word()
It's not documented anyway, and having it makes no sense anymore with
obj_elf_section_word() now being TC_SPARC-only. In any event the x86
backing function was dead code.
Jan Beulich [Fri, 11 Aug 2023 08:04:49 +0000 (10:04 +0200)]
x86: pack CPU flags in opcode table
The table constantly growing in two dimensions (number of table entries
times number of ISA extension flags) doesn't scale very well. Use a more
compact representation: Only identifiers which need to combine with
other identifiers retain individual flag bits. All others are combined
into an enum, with a new helper added to transform the table entries
into the original i386_cpu_flags layout. This way the table in the final
binary shrinks by almost a third (the generated source code shrinks by
about half), and isn't likely to grow again in that dimension any time
soon.
While moving the 3DNow! fields, drop the stray inner 'a' from their
names.
Alan Modra [Mon, 7 Aug 2023 11:29:08 +0000 (20:59 +0930)]
warn unused result for bfd IO functions
This patch fixes all the warnings I found in bfd, binutils and ld,
plus some bitrotted COFF_GO32 code that tried to allocate -168ul
bytes. When the malloc fail was reported these testsuite fails
resulted:
i386-go32 +FAIL: go32 stub
i386-go32 +ERROR: tcl error sourcing /home/alan/src/binutils-gdb/ld/testsuite/ld-i386/i386.exp.
i386-go32 +ERROR: couldn't open "tmpdir/go32stub": no such file or directory
i386-go32 +FAIL: ld-scripts/sane1
i386-go32 +FAIL: ld-scripts/assign-loc
i386-go32 +FAIL: ld-scripts/pr18963
This does result in some warnings in gdb which are fixed in a followup
patch.
bfd/
* bfdio.c (bfd_read, bfd_write): Add ATTRIBUTE_WARN_UNUSED_RESULT.
(bfd_tell, bfd_stat, bfd_seek, bfd_mmap): Likewise.
* bfd-in2.h: Regenerate.
* coff-rs6000.c (xcoff_write_armap_big) Don't ignore bfd_write
return value.
(xcoff_generate_rtinit): Likewise. Also free data_buffer and
string_table before returning.
* coff64-rs6000.c (xcoff64_generate_rtinit): Likewise.
* coff-stgo32.c (go32exe_check_format): Don't ignore bfd_seek
return value.
* coffcode.h (coff_apply_checksum): Don't ignore bfd_write return.
(coff_write_object_contents <COFF_GO32>): Likewise, and bfd_malloc.
Fix bitrotted code to look for first section with non-zero filepos.
* elf64-ia64-vms.c (elf64_vms_write_shdrs_and_ehdr): Don't ignore
bfd_seek or bfd_write return values.
* pef.c (bfd_pef_scan_section): Likewise.
(bfd_pef_read_header, bfd_pef_xlib_read_header): Likewise.
* vms-misc.c (_bfd_vms_output_end): Likewise. Return status.
* vms.h (_bfd_vms_output_end): Update prototype.
* vms-alpha.c: Pass _bfd_vms_output_end status up call chains.
* wasm-module.c (wasm_compute_custom_section_file_position): Don't
ignore bfd_seek or bfd_write return values.
(wasm_compute_section_file_positions): Likewise.
* xsym.c (bfd_sym_scan): Don't ignore bfd_seek return value.
(bfd_sym_read_name_table): Likewise.
binutils/
* ar.c (print_contents, extract_file): Don't ignore bfd_seek
return value.
ld/
* pdb.c (create_section_contrib_substream): Don't ignore bfd_seek
return value.
(create_section_header_stream): Likewise.
* pe-dll.c (pe_get16, pe_get32): Add fail param to return results
from bfd_seek and bfd_read.
(pe_implied_import_dll): Handle these fails, and other bfd_seek
and bfd_read return values.
Tsukasa OI [Sun, 6 Aug 2023 01:37:05 +0000 (01:37 +0000)]
RISC-V: Fix opcode entries of "vmsge{,u}.vx"
Their check_func should be "match_never", not "match_opcode". The reasons
this error did not cause any disassembler errors are:
1. The problem will not reproduce if "no-aliases" is specified
(because macro instructions are handled as aliases).
2. If not, all affected compressed instructions or their aliases
precede before "vmsge{,u}.vx" macro instructions.
However, it'll easily break if we reorder opcode entries. This commit
fixes this issue before the *accident* occurs.
opcodes/ChangeLog:
* riscv-opc.c (riscv_opcodes): Make sure that we never match to
vmsge{,u}.vx instructions unless specified in the assembler.
Tsukasa OI [Thu, 3 Aug 2023 01:01:23 +0000 (01:01 +0000)]
RISC-V: Remove support for non-existing 'Zve32d'
Since this "extension" does not exist (on the other hand, 'Zve64d' exists)
and it's not useful if we keep it (as other code portions just ignore
"zve32d"), this commit just removes it.
bfd/ChangeLog:
* elfxx-riscv.c (riscv_supported_std_z_ext): Remove 'Zve32d'
extension from the list.
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)
By following the parent links in the text, we can conclude that the qualified
name of DIE 0x25 is foo::foo (which is incorrect, that's PR symtab/30739).
But it's not evident, and also hard to verify in a test-case.
Add dumping of the qualified name, such that we have:
...
[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)
Carl Love [Thu, 3 Aug 2023 17:55:51 +0000 (13:55 -0400)]
Fix gdb.ada/O2_float_param.exp for PowerPC
The frame command on Power pc prints the address in hex between the
#0 and in calle.increment. For example
(gdb) frame
#0 0x0000000010010a88 in callee.increment (val=val@entry=99.0, msg=...)
at /home/.../gdb/testsuite/gdb.ada/O2_float_param/callee.adb:19
19 procedure Increment (Val : in out Float; Msg: String) is
The printing of the address for the frame is done by function
print_frame in gdb/stack.c. If SAL.IS_stmt is false for the frame,
function frame_show_address returns true and print_frame prints the
address. Currently, SAL.IS is false on PowerPC and true on X86-64.
Update the set re string to accept the hex address if it exits.
This behavior is a bit unfortunate -- I would think that it would be
better to arrange for such notification if something on the gdb side
is interested.
Meanwhile, this patch fixes py-thread-exited.exp to work around this
problem.
Tom de Vries [Thu, 10 Aug 2023 13:46:30 +0000 (15:46 +0200)]
[gdb/testsuite] Fix gdb.dwarf2/enqueued-cu-base-addr.exp with cc-with-gnu-debuglink
When running test-case gdb.dwarf2/enqueued-cu-base-addr.exp with target board
cc-with-gnu-debuglink, I run into:
...
(gdb) PASS: gdb.dwarf2/enqueued-cu-base-addr.exp: ptype foo
maint print symbols -objfile enqueued-cu-base-addr^M
(gdb) FAIL: gdb.dwarf2/enqueued-cu-base-addr.exp: CU addr found
...
The problem is that the CU we're trying to print is in objfile
enqueued-cu-base-addr.debug instead of enqueued-cu-base-addr.
Fix this by replacing "-objfile enqueued-cu-base-addr" with "-source cu2".
Tom de Vries [Thu, 10 Aug 2023 13:46:30 +0000 (15:46 +0200)]
[gdb/testsuite] Improve failure mode in gdb.dwarf2/enqueued-cu-base-addr.exp
I ran test-case gdb.dwarf2/enqueued-cu-base-addr.exp with target board
cc-with-debug-names, and ran into:
...
FAIL: gdb.dwarf2/enqueued-cu-base-addr.exp: ptype foo (GDB internal error)
FAIL: gdb.dwarf2/enqueued-cu-base-addr.exp: CU addr found
...
The first FAIL is a known issue, PR symtab/29572.
The following FAIL is a consequence of the first FAIL, so require for the
second test that the first test passes.
Tested on x86_64-linux, with target boards unix and cc-with-debug-names.
Tom de Vries [Thu, 10 Aug 2023 13:46:30 +0000 (15:46 +0200)]
[gdb/symtab] Fix assertion in write_debug_names
When running test-case gdb.dwarf2/pr13961.exp with target-board
cc-with-debug-names, I run into:
...
Running gdb.dwarf2/pr13961.exp ...
gdb compile failed, gdb/dwarf2/index-write.c:1305: internal-error: \
write_debug_names: Assertion `counter == per_bfd->all_units.size ()' failed.
...
This is a regression since commit 542a33e348a ("Only use the per-BFD object to
write a DWARF index"), which did:
...
- gdb_assert (counter == per_objfile->per_bfd->all_comp_units.size ());
+ gdb_assert (counter == per_bfd->all_units.size ());
...
Fix this by reverting to using all_comp_units:
...
gdb_assert (counter == per_bfd->all_comp_units.size ());
...
Tested on x86_64-linux, using target boards unix and cc-with-debug-names.
David Faust [Wed, 9 Aug 2023 18:44:38 +0000 (11:44 -0700)]
bpf: use w regs in 32-bit non-fetch atomic pseudo-c
The 32-bit non-fetching atomic instructions treat the source register as
32-bits, which means in the pseudo-c syntax the "w" registers should be
used rather than the "r" registers.
opcodes/
* bpf-opc-c (bpf_opcodes): Use %sw for AAD32, AOR32, AAND32
and AXOR32 pseudo-c dialect asm templates.
gas/
* testsuite/gas/bpf/atomic-be-pseudoc.d: Use "w" for source reg
in non-fetching 32-bit atomic instructions.
* testsuite/gas/bpf/atomic-pseudoc.d: Likewise.
* testsuite/gas/bpf/atomic-pseudoc.s: Likewise.
This patch introduces new debugging information that prints
breakpoint location insertion and removal flow.
The debug output looks like:
~~~
(gdb) set debug breakpoint on
(gdb) disassemble main
Dump of assembler code for function main:
0x0000555555555129 <+0>: endbr64
0x000055555555512d <+4>: push %rbp
0x000055555555512e <+5>: mov %rsp,%rbp
=> 0x0000555555555131 <+8>: mov $0x0,%eax
0x0000555555555136 <+13>: pop %rbp
0x0000555555555137 <+14>: ret
End of assembler dump.
(gdb) break *0x0000555555555137
Breakpoint 2 at 0x555555555137: file main.c, line 4.
[breakpoint] update_global_location_list: insert_mode = UGLL_MAY_INSERT
(gdb) c
Continuing.
[breakpoint] update_global_location_list: insert_mode = UGLL_INSERT
[breakpoint] insert_bp_location: Breakpoint 2 (0x5565daddb1e0) at address 0x555555555137 in main at main.c:4
[breakpoint] insert_bp_location: Breakpoint -2 (0x5565dab51c10) at address 0x7ffff7fd37b5
[breakpoint] insert_bp_location: Breakpoint -5 (0x5565dab68f30) at address 0x7ffff7fe509e
[breakpoint] insert_bp_location: Breakpoint -7 (0x5565dab694f0) at address 0x7ffff7fe63f4
[breakpoint] remove_breakpoint_1: Breakpoint 2 (0x5565daddb1e0) at address 0x555555555137 in main at main.c:4 due to regular remove
[breakpoint] remove_breakpoint_1: Breakpoint -2 (0x5565dab51c10) at address 0x7ffff7fd37b5 due to regular remove
[breakpoint] remove_breakpoint_1: Breakpoint -5 (0x5565dab68f30) at address 0x7ffff7fe509e due to regular remove
[breakpoint] remove_breakpoint_1: Breakpoint -7 (0x5565dab694f0) at address 0x7ffff7fe63f4 due to regular remove
Breakpoint 2, 0x0000555555555137 in main () at main.c:4
4 }
~~~
Co-Authored-By: Christina Schimpe <christina.schimpe@intel.com>
Alan Modra [Mon, 7 Aug 2023 05:10:35 +0000 (14:40 +0930)]
Rename bfd_bread and bfd_bwrite
These were renamed from bfd_read and bfd_write back in 2001 when they
lost an unnecessary parameter. Rename them back, and get rid of a few
casts that are only needed without prototyped functions (K&R C).
Alan Modra [Sun, 6 Aug 2023 22:58:55 +0000 (08:28 +0930)]
PR30724, cygwin ld performance regression since 014a602b86
According to the reporter of this bug the newlib fseek implementation
is likely slowed down by locking and fflush, only attempting to
optimise seeks when the file is opened read-only. Thus when writing
the output we get a dramatic slowdown due to commit 014a602b86.
PR 30724
* bfd.c (enum bfd_last_io): New.
(struct bfd): Add last_io field.
* bfd-in2.h: Regenerate.
* bfd-io.c (bfd_bread, bfd_bwrite): Force seek if last_io is
opposite direction.
(bfd_seek): Reinstate optimisation for seek to same position.
Rainer Orth [Tue, 8 Aug 2023 08:13:58 +0000 (10:13 +0200)]
Guard against killing unrelated processes in amd64-disp-step.exp
When testing current gdb trunk on Solaris/amd64, the whole session was
reliably terminated by make check. I could trace this to the following
entry in gdb.arch/amd64-disp-step/gdb.log:
FAIL: gdb.arch/amd64-disp-step.exp: add into rcx: send_signal=on: get
inferior pid
Executing on target: kill -ALRM -1 (timeout = 300)
builtin_spawn -ignore SIGHUP kill -ALRM -1
If $inferior_pid doesn't refer to a single process for some reason, this
kill would terminate either a process group or the whole session.
This patch avoids this by ensuring that the pid arg is positive.
Tested on amd64-pc-solaris2.11 and x86_64-pc-linux-gnu.
Co-Authored-By: Andrew Burgess <aburgess@redhat.com> Approved-By: Tom Tromey <tom@tromey.com>
Tom de Vries [Tue, 8 Aug 2023 08:08:04 +0000 (10:08 +0200)]
[gdb/build] Fix build breaker with -std=c++11
When building with -std=c++11 I run into:
...
gdb/dwarf2/cooked-index.c: In member function \
‘void cooked_index::start_writing_index(dwarf2_per_bfd*)’:
gdb/dwarf2/cooked-index.c:469:10: error: lambda capture initializers only \
available with -std=c++14 or -std=gnu++14 [-Werror]
ctx = std::move (ctx)] ()
^~~
...
Fix this by capturing a copy instead when using -std=c++11:
...
= gdb::thread_pool::g_thread_pool->post_task ([this, per_bfd, ctx] ()
...
Tested by building with and without -stdc=++11 on x86_64-linux.
Reported-By: Tom Tromey <tom@tromey.com> Reviewed-by: John Baldwin <jhb@FreeBSD.org>
Jan Beulich [Mon, 7 Aug 2023 09:56:50 +0000 (11:56 +0200)]
RISC-V: move comment describing rules for riscv_opcodes[]
It makes little sense to have this comment meanwhile over a hundred
lines ahead of the array. In fact until spotting the comment, I was
wondering why those pretty important aspects aren't spelled out
anywhere.
Richard Bunt [Mon, 7 Aug 2023 08:12:53 +0000 (09:12 +0100)]
gdb/fortran: Align intrinsic/variable precedence
Fortran allows variables and function to be named after language defined
intrinsics as they are not reserved keywords. For example, the abs maths
intrinsic can be hidden by a user declaring a variable called abs.
The behavior before this patch was to favour the intrinsic, which meant
that any variables named, for example "allocated", could not be
inspected by GDB.
This patch inverts this priority to bring GDB's behaviour closer to the
Fortran language, where the user defined symbol can hide the intrinsic.
Special care was need to prevent any C symbols from overriding either
Fortran intrinsics or user defined variables. This was observed to be
the case when GDB has access to symbols for abs from libm. This was
solved by only allowing symbols not marked with language_fortran to be
overridden.
In total this brings the order of precedence to the following (highest
first):
1. User defined Fortran variable or function.
2. Fortran intrinsic.
3. Symbols from languages other than Fortran.
The sizeof intrinsic is now case insensitive. This is closer to the
Fortran language. I believe this change is safe enough as it increases
the acceptance of the grammar, rather than restricts it. I.e. it should
not break any existing scripts which rely on it. Unless of course they
rely on SIZEOF being rejected.
GDB built with GCC 13.
No test suite regressions detected. Compilers: GCC, ACfL, Intel, Intel
LLVM, NVHPC; Platforms: x86_64, aarch64.
Existing tests in gdb.fortran cover the invocation of intrinsics
including: intrinsics.exp, shape.exp, rank.exp, lbound-ubound.exp.
Tom de Vries [Sat, 5 Aug 2023 15:57:13 +0000 (17:57 +0200)]
[gdb/symtab] Find main language without symtab expansion
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 d3214198119 ("[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 7f4307436fd ("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.
Tom Tromey [Fri, 4 Aug 2023 11:58:35 +0000 (05:58 -0600)]
Consolidate calls to bfd_set_cacheable
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.
Nick Clifton [Fri, 4 Aug 2023 13:19:28 +0000 (14:19 +0100)]
Fix potential infinite loop in bfd_cache_close_all.
PR 15545 * cache.c (bfd_cache_close_all): Extend description to note that all files will be closed, even those that are not cacheable. Add code to prevent a possible infinite loop.
The race happens when issuing the "file $exec" command.
The race is between:
- a worker thread writing the index cache, and in the process reading
dwarf2_per_cu_data::is_debug_type, and
- the main thread writing to dwarf2_per_cu_data::files_read.
The two bitfields dwarf2_per_cu_data::files_read and
dwarf2_per_cu_data::is_debug_type share the same bitfield container.
Fix this by making dwarf2_per_cu_data::files_read a packed<bool, 1>.
The race happens when issuing the "file $exec" command.
The race is between:
- a worker thread writing the index cache, and in the process reading
dwarf2_per_cu_data::is_debug_type, and
- the main thread writing to dwarf2_per_cu_data::mark.
The two bitfields dwarf2_per_cu_data::mark and
dwarf2_per_cu_data::is_debug_type share the same bitfield container.
Fix this by making dwarf2_per_cu_data::mark a packed<unsigned int, 1>.
Tom de Vries [Fri, 4 Aug 2023 13:02:43 +0000 (15:02 +0200)]
[gdb/testsuite] Extend gdb.base/index-cache.exp
The test-case gdb.base/index-cache.exp uses only one source file, which
contains main.
While doing "file $exec", in set_initial_language a symbol lookup of "main" is
done, causing the symtab containing main to be expanded.
Handling of main is special, and a future optimization may skip the lookup and
expansion.
Reliably exercise:
- the lookup of main, expanding the symtab containing main, by doing
"ptype main", and
- the lookup of another symbol, expanding a symtab not containing main, by:
- adding another source file containing function foo, and
- doing "ptype foo".
This triggered a segfault with target board native-extended-gdbserver, filed
as PR symtab/30712, but that seems to be fixed by a previous commit in this
series.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30712
The race happens when issuing the "file $exec" command.
The race is between:
- a worker thread writing the index cache, and in the process reading
dwarf2_per_cu_data::is_debug_type, and
- the main thread writing to dwarf2_per_cu_data::m_header_read_in.
The two bitfields dwarf2_per_cu_data::m_header_read_in and
dwarf2_per_cu_data::is_debug_type share the same bitfield container.
Fix this by making dwarf2_per_cu_data::m_header_read_in a packed<bool, 1>.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/30392
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30392
The race happens when issuing the "file $exec" command.
The race is between:
- a worker thread writing the index cache, and in the process reading
dwarf2_per_cu_data::is_debug_type, and
- the main thread expanding the CU containing main, and in the process setting
dwarf2_per_cu_data::queued.
The two bitfields dwarf2_per_cu_data::queue and
dwarf2_per_cu_data::is_debug_type share the same bitfield container.
Fix this by making dwarf2_per_cu_data::queued a packed<bool, 1>.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/30392
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30392
The race happens when issuing the "file $exec" command.
The race is between:
- a worker thread getting the build id while writing the index cache, and in
the process reading bfd::format, and
- the main thread calling find_main_name, and in the process setting
bfd::cacheable.
The two bitfields bfd::cacheable and bfd::format share the same bitfield
container.
Fix this by capturing the build id in the main thread, and using the captured
value in the worker thread.
Likewise for the dwz build id, which likely suffers from the same issue.
While we're at it, also move the creation of the cache directory to
the index_cache_store_context constructor, to:
- make sure there's no race between subsequent file commands, and
- issue any related warning or error messages during the file command.
Tested on x86_64-linux.
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/30392
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30392
Location is global 'global_index_cache' of size 48 at 0x00000321f520 (gdb+0x00000321f540)
...
SUMMARY: ThreadSanitizer: data race gdb/dwarf2/index-cache.c:76 in index_cache::enable()
...
The race happens when issuing a "file $exec" command followed by a
"set index-cache enabled on" command.
The race is between:
- a worker thread reading index_cache::m_enabled to determine whether an
index-cache entry for $exec needs to be written
(due to command "file $exec"), and
- the main thread setting index_cache::m_enabled
(due to command "set index-cache enabled on").
Fix this by capturing the value of index_cache::m_enabled in the main thread,
and using the captured value in the worker thread.
Alan Modra [Fri, 4 Aug 2023 08:07:12 +0000 (17:37 +0930)]
ppc: sanity check writing relocs
Check for output buffer overruns.
* elf32-ppc.c (swap_reloc_out, count_and_swap_reloc_out): New
functions. Use throughout file.
* elf64-ppc.c (swap_reloc_out, count_and_swap_reloc_out): Likewise.
Alan Modra [Fri, 4 Aug 2023 05:39:53 +0000 (15:09 +0930)]
PR30697, ppc32 mix of local-dynamic and global-dynamic TLS
This fixes miscounting of dynamic relocations on GOT entries when
a) there are both local-dynamic and global-dynamic tls accesss for a
given symbol, and
b) the symbol is global with non-default visibility, and
c) the __tls_get_addr calls aren't optimised away.
PR 30697
bfd/
* elf32-ppc.c (allocate_dynrelocs): Correct local-dynamic
reloc count.
ld/
* testsuite/ld-powerpc/tls32ldgd.d,
* testsuite/ld-powerpc/tls32ldgd.s: New test.
* testsuite/ld-powerpc/powerpc.exp: Run it.
Bruno Larsen [Fri, 14 Jul 2023 08:06:03 +0000 (10:06 +0200)]
gdb/testsuite: Disable gdb.compile when testing with clang
Attempting to test the gdb.compile with clang as the compiler results in
over 300 unexpected errors, due to a segmentation fault and several
handshake failures. Since the whole feature is designed around a gcc
plugin, and even the gcc testing is shaky at best, this commit restricts
those tests to only running under gcc. If that gets fixed, this commit
can be reverted.