One test name in gdb.base/dlmopen.exp changes from run to run
since it includes a process id:
PASS: gdb.base/dlmopen.exp: attach 3442682
This is not convenient do diff gdb.sum files to compare test runs.
Fix by using gdb_attach helper function to handle attaching to the
process as it produce a constant test name.
While at it also check gdb_attach's return value to only run the
rest of the test if the attach was successful.
Approved-By: Simon Marchi <simon.marchi@efficios.com>
The mnemonics for the pmxvf16ger*, pmxvf32ger*,pmxvf64ger*, pmxvi4ger8*,
pmxvi8ger4*, and pmxvi16ger2* instructions were officially changed to
pmdmxbf16ger*, pmdmxvf32ger*, pmdmxvf64ger*, pmdmxvi4ger8*, pmdmxvi8ger4*,
pmdmxvi16ger* respectively. The old mnemonics are still supported by the
assembler as extended mnemonics. The disassembler generates the new
mnemonics. The name changes occurred in commit:
commit bb98553cad4e017f1851153fa5de91f2cee98fb2
Author: Peter Bergner <bergner@linux.ibm.com>
Date: Sat Oct 8 16:19:51 2022 -0500
PowerPC: Add support for RFC02658 - MMA+ Outer-Product Instructions
gas/
* config/tc-ppc.c (md_assemble): Only check for prefix opcodes.
* testsuite/gas/ppc/rfc02658.s: New test.
* testsuite/gas/ppc/rfc02658.d: Likewise.
* testsuite/gas/ppc/ppc.exp: Run it.
opcodes/
* ppc-opc.c (XMSK8, P_GERX4_MASK, P_GERX2_MASK, XX3GERX_MASK): New.
(powerpc_opcodes): Add dmxvi8gerx4pp, dmxvi8gerx4, dmxvf16gerx2pp,
dmxvf16gerx2, dmxvbf16gerx2pp, dmxvf16gerx2np, dmxvbf16gerx2,
dmxvi8gerx4spp, dmxvbf16gerx2np, dmxvf16gerx2pn, dmxvbf16gerx2pn,
dmxvf16gerx2nn, dmxvbf16gerx2nn, pmdmxvi8gerx4pp, pmdmxvi8gerx4,
pmdmxvf16gerx2pp, pmdmxvf16gerx2, pmdmxvbf16gerx2pp, pmdmxvf16gerx2np,
pmdmxvbf16gerx2, pmdmxvi8gerx4spp, pmdmxvbf16gerx2np, pmdmxvf16gerx2pn,
pmdmxvbf16gerx2pn, pmdmxvf16gerx2nn, pmdmxvbf16gerx2nn.
This patch updates the comments in the various gdb files to reflect the
name changes. There are no functional changes made by this patch.
The older instruction names are still used in the test
gdb.reverse/ppc_record_test_isa_3_1.exp for backwards compatibility.
Patch has been tested on Power 10 with no regressions.
The mnemonics for the pmxvf16ger*, pmxvf32ger*,pmxvf64ger*, pmxvi4ger8*,
pmxvi8ger4*, pmxvi16ger2* instructions were officially changed to
pmdmxvf16ger*, pmdmxvf32ger*, pmdmxvf64ger*, pmdmxvi4ger8*, pmdmxvi8ger4*,
pmdmxvi16ger* respectively. The old mnemonics are still supported by the
assembler as extended mnemonics. The disassembler generates the new
mnemonics. The name changes occurred in commit:
commit bb98553cad4e017f1851153fa5de91f2cee98fb2
Author: Peter Bergner <bergner@linux.ibm.com>
Date: Sat Oct 8 16:19:51 2022 -0500
PowerPC: Add support for RFC02658 - MMA+ Outer-Product Instructions
gas/
* config/tc-ppc.c (md_assemble): Only check for prefix opcodes.
* testsuite/gas/ppc/rfc02658.s: New test.
* testsuite/gas/ppc/rfc02658.d: Likewise.
* testsuite/gas/ppc/ppc.exp: Run it.
opcodes/
* ppc-opc.c (XMSK8, P_GERX4_MASK, P_GERX2_MASK, XX3GERX_MASK): New.
(powerpc_opcodes): Add dmxvi8gerx4pp, dmxvi8gerx4, dmxvf16gerx2pp,
dmxvf16gerx2, dmxvbf16gerx2pp, dmxvf16gerx2np, dmxvbf16gerx2,
dmxvi8gerx4spp, dmxvbf16gerx2np, dmxvf16gerx2pn, dmxvbf16gerx2pn,
dmxvf16gerx2nn, dmxvbf16gerx2nn, pmdmxvi8gerx4pp, pmdmxvi8gerx4,
pmdmxvf16gerx2pp, pmdmxvf16gerx2, pmdmxvbf16gerx2pp, pmdmxvf16gerx2np,
pmdmxvbf16gerx2, pmdmxvi8gerx4spp, pmdmxvbf16gerx2np, pmdmxvf16gerx2pn,
pmdmxvbf16gerx2pn, pmdmxvf16gerx2nn, pmdmxvbf16gerx2nn.
The above commit results in about 224 failures on Power 10 since the
disassembled names do not match the expected names in the test. This
patch updates the expected names in the test to match the values produced
by the disassembler.
This patch updates file gdb.arch/powerpc-power10.exp with the new expected
values to the instructions. The comment giving the name of the instruction
for each binary value in the file gdb.arch/powerpc-power10.c is updated
with the new name. There are no functional changes in file
gdb.arch/powerpc-power10.c.
The test disassembles function foo and searches for the line
"End of assembler dump" to determing the last address in the function. The
assumption is the last instruction will be given right before the line
"End of assembler dump". This assumption fails on PowerPC.
The PowerPC disassembly of the function foo looks like:
Dump of assembler code for function foo:
# => 0x00000000100006dc <+0>: std r31,-8(r1)
# 0x00000000100006e0 <+4>: stdu r1,-48(r1)
# 0x00000000100006e4 <+8>: mr r31,r1
# 0x00000000100006e8 <+12>: nop
# 0x00000000100006ec <+16>: addi r1,r31,48
# 0x00000000100006f0 <+20>: ld r31,-8(r1)
# 0x00000000100006f4 <+24>: blr
# 0x00000000100006f8 <+28>: .long 0x0
# 0x00000000100006fc <+32>: .long 0x0
# 0x0000000010000700 <+36>: .long 0x1000180
# End of assembler dump.
The blr instruction is the last instruction in function foo. The lines
with .long following the blr instruction need to be ignored.
This patch adds a new condition to the gdb_test_multiple "disassemble foo"
test to ignore the lines with the .long.
The patch has been tested on PowerPC and Intel X86-64.
Recent changes to gdb.reverse/step-reverse.exp revealed the latent bug
PR record/29745, where we can't skip one funcion forward if we're using
native-gdbserver. This commit just adds kfails to the test.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29745
Approved-By: Simon Marchi <simon.marchi@efficios.com>
After this commit:
commit 6576bffe6cbbb53c5756b2fccd2593ba69b74cdf
Date: Thu Jul 7 13:43:45 2022 +0100
opcodes/arm: add disassembler styling for arm
Some people were seeing their builds failing with complaints about a
possible uninitialized variable usage. I previously fixed an instance
of this issue in this commit:
commit 2df82cd4b459fbc32120e0ad1ce19e26349506fe
Date: Tue Nov 1 10:36:59 2022 +0000
opcodes/arm: silence compiler warning about uninitialized variable use
which did fix the build problems that the sourceware buildbot was
hitting, however, an additional instance of the same problem was
brought to my attention, and that is fixed in this commit.
Where commit 2df82cd4b4 fixed the uninitialized variable problem in
print_mve_unpredictable, this commit fixes the same problem in
print_mve_undefined.
As with the previous commit, I don't believe we could really ever get
an uninitialized variable usage, based on the current usage of the
function, so I have just initialized the reason variable to "??".
These manual settings were necessary when we weren't doing automatic
header dependency tracking. That was changed a while ago, and we use
automake now to do it all for us. As a result, many of these vars
aren't even referenced anymore.
Further, some of the source file generation (e.g. .c files, or igen,
or cgen outputs) were moved to the common automake build, and it takes
care of dependency tracking for us with the object files.
This patch is based on MULTIPLE_FRAME_SECTIONS and EH_FRAME_LINKONCE,
it allows backend to enable this feature and use '--gc-sections' simply.
* gas/dw2gencfi.h (TARGET_MULTIPLE_EH_FRAME_SECTIONS): New.
(MULTIPLE_FRAME_SECTIONS): Add TARGET_MULTIPLE_EH_FRAME_SECTIONS.
* gas/dw2gencfi.c (EH_FRAME_LINKONCE): Add TARGET_MULTIPLE_EH_FRAME_SECTIONS.
(is_now_linkonce_segment): Likewise.
(get_cfi_seg): Create relocation info between .eh_frame.* and .text.* section.
* bfd/elf-bfd.h (elf_backend_can_make_multiple_eh_frame): New.
* bfd/elfxx-target.h (elf_backend_can_make_multiple_eh_frame): Likewise.
* bfd/elflink.c (_bfd_elf_default_action_discarded): Add checking for
elf_backend_can_make_multiple_eh_frame.
We have configure tests for this in the top-level configure script
to link this when necessary, so we don't need to explicitly list it
for specific ports.
With more C libraries moving functions entirely into the main -lc,
change the AC_CHECK_LIB calls to AC_SEARCH_LIBS so we look in there
first and avoid extra linkage when possible.
COMMON_LIBS is set to $(LIBS), and CONFIG_LIBS is set to that plus
@LIBS@. This leds to the values being used twice. Inline the
CONFIG_LIBS variable without @LIBS@ since it's used only once.
The top-level already sets up a libtool script for the host, so use
that when linking rather than invoking CC directly. This will also
happen when we (someday) move the building to pure automake.
This test uses the test itself as an input to stating regular files.
This gets funky though: when we run check in parallel, the output
object dir is the subdir that matches the .exp file. When we run
with -j1, the output object dir is the sim builddir itself.
The old test would append argv[0] to find the file, while the new
test uses basename on it. Each method works in only one of the
aforementioned build scenarios. Rather than complicate this any
more, switch to a different file that we know will always exist:
the Makefile.
This test assumes that /bin/sh will never be a CRIS ELF by way of
assuming that the current bfd cannot load it (since a basic cris
cross-compiler only understands CRIS ELFs). In a multi-target
build though, bfd understands just about every ELF out there, so
we're able to read the /bin/sh format before failing at a diff
point in the cris code.
Let's switch to using / instead since it'll fail for a similar
reason (at least similar enough for what this test is testing).
This is used to allow for dangling \ in object lists, but these are the
only ports that do it, and it isn't really necessary. Punt it to keep
the various makefiles harmonized.
The erc32 sim does a lot itself, including handling of the CLI. It
used to provide a run-compatible interface in the pre-nrun days, but
it was dropped when the old run interface was punted. Since the old
commit 465fb143c87076b6416a8d0d5dd79bb016060fe3 ("sim: make nrun the
default run program"), the erc32 run & sis programs have been the
same, and erc32 hasn't provide a real run-compatible interface.
Simplify this by linking the two programs via ln/cp instead of running
the linking phase twice to produce the same result. If/when we fix up
the erc32 port to have a proper run interface, it should be easy to
split these back apart into real programs.
Note: the interf.o reference in here is a bit of a misdirect. Since
that object is placed into libsim.a, it's never been linked into the
programs since the linker ignores objects that aren't referenced, and
only gdb uses those symbols.
The v850 port uses -DDEBUG to control whether to enable internal tracing.
We already have such options via the common trace framework, and those
can be controlled at build time via configure flags (which the v850 code
currently cannot). So switch it over to WITH_TRACE_ANY_P to simplify the
v850 build code even if it doesn't (yet) respect any other trace options.
Clang up to version 15 (current) adds macros that were defined in the
command line or by "other means", according to the Dwarf specification,
after the last DW_MACRO_end_file, instead of before the first
DW_MACRO_start_file, as the specification dictates. When GDB reads the
macros after the last file is closed, the macros never end up "in scope"
and so we can't print them. This has been submitted as a bug to Clang
developers (https://github.com/llvm/llvm-project/issues/54506), and PR
macros/29034 was opened for GDB to keep track of this.
Seeing as there is no expected date for it to be fixed, add a workaround
for all current versions of Clang. The workaround detects when
the main file would be closed and if the producer is Clang, and turns
that operation into a noop, so we keep a reference to the current_file
as those macros are read.
A test case was added to confirm the functionality, and the KFAIL for
running gdb.base/macro-source-path when using clang.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29034
Approved-By: Simon Marchi <simon.marchi@efficios.com>
PR 29741
* scripttempl/avr.sc (__DATA_REGION_ORIGIN__): Define. If a value
has not been provided on the command line then use DATA_ORIGIN.
(MEMORY): Use __DATA_REGION_ORIGIN__ as the start of the data region.
PR 29471
* scripttempl/avr.sc (__DATA_REGION_ORIGIN__): Define. If a value
has not been provided on the command line then use DATA_ORIGIN.
(MEMORY): Use __DATA_REGION_ORIGIN__ as the start of the data region.
Since all host files we compile use these settings, move them out of
libcommon.a and into the default AM_CPPFLAGS. This has the effect of
dropping the custom per-target automake rules. Currently it saves us
~150 lines, but since it's about ~8 lines per object, the overhead
will increase quite a bit as we merge more files into a single build.
This also changes the object output names, so we have to tweak the
rules that were pulling in the common objects when linking.
A bunch of these paths don't include any headers, and most likely
never will, so there's no real need to keep them. This will let
us harmonize paths with the top-level Makefile more easily, which
will in turn make it easier to move more compile steps there.
';' does not always indicate the start of a comment, and commit
8cb6e17571f3fb66ccd4fa19f881602542cd06fc incorrectly replaced 3
instances of ';' with '@' in expected diagnostics, leading to tests
failures.
This patch restores the original ';' as needed in these testcases.
Fixes bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29739
In order to merge more common/ files into the top-level, we need to
add more host flags to CPPFLAGS, and that conflicts with our current
use with build-time tools. So split them apart like we do with all
other build flags to avoid the issue.
When reading/writing arbitrary data to the system's memory, the unsigned
char pointer type doesn't make that much sense. Switch it to void so we
align a bit with standard C library read/write functions, and to avoid
having to sprinkle casts everywhere.
Absent _UNICODE being defined (which gdb's Makefile doesn't do),
windows.h will always define STARTUPINFO is as STARTUPINFOA, so this
cast isn't correct when create_process expects a STARTUPINFOW
parameter (i.e. in a Cygwin build).
Instead write this as &info_ex.StartupInfo (which is always of the
correct type).
Prior to commit 1cb0ab18ad24 ("x86/Intel: restrict suffix derivation")
the Tbyte modifier on the FLDT and FSTPT templates was pointless, as
No_ldSuf would have prevented it being accepted. Due to the special
nature of LONG_DOUBLE_MNEM_SUFFIX said commit, however, has led to these
insns being accepted in Intel syntax mode even when "tbyte ptr" was
present. Restore original behavior by dropping Tbyte there. (Note that
these insns in principle should by marked AT&T syntax only, but since
they haven't been so far we probably shouldn't change that.)
Comparing the sum of the relevant .imm<N> fields against a constant imo
makes more obvious what is actually meant. It allows dropping of two
static variables, with a 3rd drop requiring two more minor adjustments
elsewhere, utilizing that "i" is zeroed first thing in md_assemble().
This also increases the chances of the compiler doing the calculations
all in registers.
Consider the case,
.option arch, rv32i
.option norelax
.option arch, +c
.byte 1
.align 2
addi a0, zero, 1
Assembler adds $d for the odd .byte, and then adds $x+arch for the
alignment. Since norelax, riscv_add_odd_padding_symbol will add the
$d and $x for the odd alignment, but accidently remove the $x+arch because
it has the same address as $d. Therefore, we will get the unexpected result
before applying this patch,
.byte 1 # $d
.align 2 # odd alignment, $xrv32ic replaced by $d + $x
After this patch, the expected result should be,
.byte 1 # $d
.align 2 # odd alignment, $xrv32ic replaced by $d + $xrv32ic
gas/
* config/tc-riscv.c (make_mapping_symbol): If we are adding mapping symbol
for odd alignment, then we probably will remove the $x+arch by accidently
when it has the same address of $d. Try to add the removed $x+arch back
after the $d rather than just $x.
(riscv_mapping_state): Updated since parameters of make_mapping_symbol are
changed.
(riscv_add_odd_padding_symbol): Likewise.
(riscv_remove_mapping_symbol): Removed and moved the code into the
riscv_check_mapping_symbols.
(riscv_check_mapping_symbols): Updated.
* testsuite/gas/riscv/mapping-dis.d: Updated and added new testcase.
* testsuite/gas/riscv/mapping-symbols.d: Likewise.
* testsuite/gas/riscv/mapping.s: Likewise.
