When using cpychecker, we run into:
...
gdb/python/py-progspace.c: \
In function ‘PyObject* pspy_solib_name(PyObject*, PyObject*)’:
gdb/python/py-progspace.c:370:25: error: Mismatching type in call to \
PyArg_ParseTuple with format code "K" [-Werror]
if (!PyArg_ParseTuple (args, GDB_PY_LLU_ARG, &pc))
argument 3 ("&pc") had type
"gdb_py_longest *" (pointing to 64 bits)
but was expecting
"long long unsigned int *" (pointing to 64 bits)
for format code "K"
...
Fix this fixing the type of the variable.
Build and reg-tested on x86_64-linux.
2018-10-05 Tom de Vries <tdevries@suse.de>
* python/py-progspace.c (pspy_solib_name): Fix type mismatch in
PyArg_ParseTuple call.
When using cpychecker, we run into this error:
...
gdb/python/py-record-btrace.c: \
In function ‘PyObject* recpy_bt_goto(PyObject*, PyObject*)’:
gdb/python/py-record-btrace.c:783:25: error: Mismatching type in call to \
PyArg_ParseTuple with format code "O" [-Werror]
if (!PyArg_ParseTuple (args, "O", &obj))
argument 3 ("&obj") had type
"const struct recpy_element_object * *"
but was expecting
"struct PyObject * *"
for format code "O"
...
Fix this by using a new variable of the expected type instead.
Build and reg-tested on x86_64-linux.
2018-10-05 Tom de Vries <tdevries@suse.de>
* python/py-record-btrace.c (recpy_bt_goto): Fix type mismatch in
PyArg_ParseTuple call.
This is just to make sure one of the parameters doesn't end past
the 80 characters limit. And while at it, since all parameters were
on their own line except the first two that were listed on the same
line, it felf more consistent to just split them to have each and
every parameter on their own line.
gdb/ChangeLog:
* psymtab.c (recursively_search_psymtabs): Reformat parameters
to avoid exceeding 80 characters per line limit.
Tested by rebuilding GDB on x86_64-linux.
This removes the SHORT_FIRST_MESSAGE case from complaints, leaving
only a single case. This allows for the removal of the last argument
to clear_complaints, and also simplifies complaint_internal, removing
an extra allocation in the process.
After this, the "./gdb -iex 'set complaint 1' -nx ./gdb" example will
show:
Reading symbols from ./gdb...
During symbol reading: .debug_ranges entry has start address of zero [in module /home/tromey/gdb/build/gdb/gdb]
During symbol reading: DW_AT_low_pc 0x0 is zero for DIE at 0x17116c1 [in module /home/tromey/gdb/build/gdb/gdb]
During symbol reading: .debug_line address at offset 0xa22f5 is 0 [in module /home/tromey/gdb/build/gdb/gdb]
During symbol reading: unsupported tag: 'DW_TAG_unspecified_type'
During symbol reading: const value length mismatch for 'std::ratio<1, 1000000000>::num', got 8, expected 0
This is a bit wordier but, I think, a bit more clear, as the form of
the message no longer depends on precisely when it was emitted. In
particular if you compare to the output from the 'Clean up "Reading
symbols" output' patch, you can see that earlier gdb would switch from
the prefix-less form to the "During symbol reading" form at a point
that is meaningless to the user (specifically, after psymtab reading
is done and gdb tries to expand a CU).
2018-10-04 Tom Tromey <tom@tromey.com>
* symfile.c (syms_from_objfile_1, finish_new_objfile)
(reread_symbols): Update.
* complaints.h (clear_complaints): Remove argument.
* complaints.c (enum complaint_series): Remove.
(series): Remove global.
(complaint_internal): Update.
(clear_complaints): Remove argument.
gdb/testsuite/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* gdb.cp/maint.exp (test_invalid_name): Update expected output.
* gdb.gdb/complaints.exp (test_short_complaints): Remove.
(test_initial_complaints, test_empty_complaints): Update.
* gdb.dwarf2/dw2-stack-boundary.exp: Update.
The "no debugging symbols" message can be confusing in some cases, for
example when gdb finds separate debug info for an objfile, but the
separate debug info does not contain symbols.
For example:
(gdb) file /bin/ls
Reading symbols from /bin/ls...
Reading symbols from .gnu_debugdata for /usr/bin/ls...
(No debugging symbols found in .gnu_debugdata for /usr/bin/ls)
(No debugging symbols found in /bin/ls)
Here, I think the second "no debugging symbols" message is redundant
and confusing.
This patch changes gdb to only emit this message when the objfile in
question does not have a separate debug file. So, in the example
above, the output would now read:
(gdb) file /bin/ls
Reading symbols from /bin/ls...
Reading symbols from .gnu_debugdata for /usr/bin/ls...
(No debugging symbols found in .gnu_debugdata for /usr/bin/ls)
2018-10-04 Tom Tromey <tom@tromey.com>
* symfile.c (symbol_file_add_with_addrs): Do not print "no
debugging symbols" message if there is a separate debug objfile.
This patch is another attempt to fix PR cli/19551. Unlike my previous
attempt, it doesn't print progress. Instead, it just changes some
messages and adds newlines to make the output a bit nicer.
It also removes the "done." text that was previously emitted. The
idea here is that it is obvious when gdb is done reading debug info,
as it starts then doing something else; and that while this message
did not provide much benefit to users, it did make it harder to make
the output clean.
After this change the output from "./gdb -iex 'set complaint 1' -nx ./gdb"
reads:
Reading symbols from ./gdb...
.debug_ranges entry has start address of zero [in module /home/tromey/gdb/build/gdb/gdb]
DW_AT_low_pc 0x0 is zero for DIE at 0x17116c1 [in module /home/tromey/gdb/build/gdb/gdb]
.debug_line address at offset 0xa22f5 is 0 [in module /home/tromey/gdb/build/gdb/gdb]
During symbol reading, unsupported tag: 'DW_TAG_unspecified_type'.
During symbol reading, const value length mismatch for 'std::ratio<1, 1000000000>::num', got 8, expected 0.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
PR cli/19551:
* symfile.c (symbol_file_add_with_addrs): Update output.
* psymtab.c (require_partial_symbols): Update output.
gdb/testsuite/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
PR cli/19551:
* lib/mi-support.exp (mi_gdb_file_cmd): Update.
* lib/gdb.exp (gdb_file_cmd): Update.
* gdb.stabs/weird.exp (print_weird_var): Update.
* gdb.server/solib-list.exp: Update.
* gdb.multi/remove-inferiors.exp (test_remove_inferiors): Update.
* gdb.mi/mi-cli.exp: Update.
* gdb.linespec/linespec.exp: Update.
* gdb.dwarf2/dw2-stack-boundary.exp: Update.
* gdb.dwarf2/dw2-objfile-overlap.exp: Update.
* gdb.cp/cp-relocate.exp: Update.
* gdb.base/sym-file.exp: Update.
* gdb.base/relocate.exp: Update.
* gdb.base/readnever.exp: Update.
* gdb.base/print-symbol-loading.exp (test_load_core): Update.
* gdb.base/kill-detach-inferiors-cmd.exp: Update.
* gdb.base/dbx.exp (gdb_file_cmd): Update.
* gdb.base/code_elim.exp: Update.
* gdb.base/break-unload-file.exp (test_break): Update.
* gdb.base/break-interp.exp (test_attach_gdb): Update.
* gdb.base/break-idempotent.exp (force_breakpoint_re_set):
Update.
* gdb.base/attach.exp (do_attach_tests): Update.
* gdb.base/sepdebug.exp: Update.
* gdb.python/py-section-script.exp: Update.
Currently complaints are not always printed with a newline. For
example, when I run gdb on itself, I see output like:
(gdb) set complaints 5
(gdb) file ./gdb/gdb
Reading symbols from ./gdb/gdb...DW_AT_low_pc 0x0 is zero for DIE at 0x437dd4 [in module /home/tromey/gdb/build/gdb/gdb]....debug_line address at offset 0x21bf9 is 0 [in module /home/tromey/gdb/build/gdb/gdb]...DW_AT_low_pc 0x0 is zero for DIE at 0x5a85dd [in module /home/tromey/gdb/build/gdb/gdb]....debug_line address at offset 0x2dc2d is 0 [in module /home/tromey/gdb/build/gdb/gdb]...DW_AT_low_pc 0x0 is zero for DIE at 0xab6033 [in module /home/tromey/gdb/build/gdb/gdb]....debug_line address at offset 0x4f683 is 0 [in module /home/tromey/gdb/build/gdb/gdb]...DW_AT_low_pc 0x0 is zero for DIE at 0x10028f0 [in module /home/tromey/gdb/build/gdb/gdb]....debug_line address at offset 0x75edf is 0 [in module /home/tromey/gdb/build/gdb/gdb]...DW_AT_low_pc 0x0 is zero for DIE at 0x1021364 [in module /home/tromey/gdb/build/gdb/gdb]....debug_line address at offset 0x76f62 is 0 [in module /home/tromey/gdb/build/gdb/gdb]...done.
That's one very long line. I find it quite difficult to read that,
and I thought it would be better with some newlines, which is what
this patch does. Now the output looks like:
(gdb) file ./gdb
Reading symbols from ./gdb...
DW_AT_low_pc 0x0 is zero for DIE at 0x437dd4 [in module /home/tromey/gdb/build/gdb/gdb]
.debug_line address at offset 0x21bf9 is 0 [in module /home/tromey/gdb/build/gdb/gdb]
DW_AT_low_pc 0x0 is zero for DIE at 0x5a85dd [in module /home/tromey/gdb/build/gdb/gdb]
.debug_line address at offset 0x2dc2d is 0 [in module /home/tromey/gdb/build/gdb/gdb]
done.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
PR cli/22234:
* complaints.c: Emit \n.
gdb/testsuite/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
PR cli/22234:
* gdb.dwarf2/dw2-stack-boundary.exp: Update expected output.
* gdb.gdb/complaints.exp (test_short_complaints): Update expected
output.
While working on this series, I found some unfiltered prints that
didn't make sense -- many things, like complaints, are filtered, while
their context (in this case the "Reading symbols ..." messages) may
not be.
This patch changes some spots in symbol reading to use filtered
prints. I think this is preferable overall; in fact I'd go farther
and say that unfiltered printing should just be removed.
One of these spots was not strictly related (the change to
dump_psymtab), but I left it in partly because it makes psymtab.c
"unfiltered-clean", and partly because it is an example of filtered
and unfiltered printing being mixed in the same function.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* symfile.c (symbol_file_add_with_addrs, symbol_file_clear)
(separate_debug_file_exists, find_separate_debug_file)
(add_symbol_file_command, reread_symbols, allocate_symtab)
(allocate_compunit_symtab): Use filtered printing, not
unfiltered.
* psymtab.c (require_partial_symbols, dump_psymtab)
(allocate_psymtab): Use filtered printing, not unfiltered.
complaint_internal had an off-by-one error, where it would allow one
extra complaint to be issued.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* complaints.c (complaint_internal): Correctly check complaint
count.
There were a couple of leftovers from earlier patches in
complaints.[ch]. This removes them.
gdb/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* complaints.h (struct complaints): Remove declaration.
* complaints.c (clear_complaints): Remove an unused variable.
Since ft32-* and xgate-* use generic linker, SHT_NOTE sections aren't
grouped nor sorted.
PR ld/23658
* testsuite/ld-elf/pr23658-1.d: Also xfail ft32-* and xgate-*.
While re-testing the complaint series, I saw some unresolved tests in
attach.exp. In particular, the tests were failing because the pager
was active.
This is partly a new problem, introduced because that series changes
some prints from unfiltered to filtered. However, it is also a latent
bug, which you can see by shrinking your window very small and then
running the test.
This patch avoids the problem by passing -quiet to gdb and arranging
to set the window height and width in one other test.
Arguably instead of -quiet we should disable the pager during gdb's
welcome message. I can do that if it seems desirable; but meanwhile
this patch is safe.
gdb/testsuite/ChangeLog
2018-10-04 Tom Tromey <tom@tromey.com>
* gdb.base/attach.exp (test_command_line_attach_run): Use -quiet;
set width and height to 0.
(test_command_line_attach_run): Use -quiet.
Joel asked me to add myself to gdb/MAINTAINERS. Done as follows, fixing two
incorrectly indented lines on the way. Installed on master.
* MAINTAINERS (Write After Approval): Add self.
The SPARC ELF BFD backend uses a hack in order to accomodate the
STT_REGISTER symbols mandated by the SPARC V9 ABI for 64-bit objects.
The hack works as follows:
- Early in `size_dynamic_symbols', it adds the dynamic STT_REGISTER
symbols and the corresponding DT_SPARC_REGISTER tags if needed,
i.e. if the input object has been annotated by the assembler to use
any of the global registers requiring annotations by the ABI.
The STT_REGISTER symbols are not local, but nevertheless they are
added to the end of the dynlocal linked list (eek, yes) to be fixed
"later". This is done so the symbols are emitted in the symtab.
- Consequently, when the `sh_info' field of the .dynsym section is
calculated in `bfd_elf_final_link' to be `local_dynsymcount + 1', it
may have the wrong value, since the real first global symbol is the
first STT_REGISTER symbol.
- However, this temporary inconsistency is fixed in the
`elf64_sparc_output_arch_syms' backend hook: the sh_index is
adjusted to its rightful value. So all is well and good.
However the 2015 changeset
commit 8539e4e89e
Author: Alan Modra <amodra@gmail.com>
Date: Thu Jan 15 19:42:59 2015 +1030
Fix ARM fail of gap test
ld-elf/gap test was failing due to the ARM backend attempting to output
arch symbols when ld -s (strip all symbols) is in force. This patch
stops that happening and tidies the code a little.
made the `elf_backend_output_arch_syms' backend hook to not be called
when all symbols are to be stripped. This resulted in an incorrect
sh_index for .dynsym when a link is performed with -s (strip_all), in
64-bit sparc ELF objects.
This patch moves the sh_index adjusting code from the target
`output_arch_syms' to `finish_dynamic_sections'. It also removes the
strip_all check from `elf64_sparc_output_arch_syms', as the function
is no longer called in that case.
Tested in sparc64-linux-gnu and sparc-linux-gnu.
No regressions observed.
bfd/ChangeLog:
2018-10-04 Jose E. Marchesi <jose.marchesi@oracle.com>
* elf64-sparc.c (elf64_sparc_output_arch_syms): Do not correct the
impact of STT_REGISTER symbols in the dynsym sh_index here...
* elfxx-sparc.c (_bfd_sparc_elf_finish_dynamic_sections): ... but
do it here.
This avoids a couple of uninitialized warnings from gcc by
initializing the object in question. The one in coffread.c seems like
it could be a latent bug. The one in scm-value.c is harmless, but GCC
can't see that.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* guile/scm-value.c (gdbscm_value_to_string): Initialize
"buffer_contents".
* coffread.c (coff_symtab_read): Initialize "newobj".
I get the following error with gcc 6.3.0:
/home/simark/src/binutils-gdb/gdb/dwarf2read.c: In function 'void read_func_scope(die_info*, dwarf2_cu*)':
/home/simark/src/binutils-gdb/gdb/dwarf2read.c:13838:12: error: types may not be defined in a for-range-declaration [-Werror]
for (struct symbol *sym : template_args)
^~~~~~
Removing the struct keyword fixes it.
gdb/ChangeLog:
* dwarf2read.c (read_func_scope): Remove struct keyword in
range-based for.
This adds --enable-ubsan to gdb's configure. By default it is enabled
in development mode, and disabled otherwise. This passes both
-fsanitize=undefined and -fno-sanitize-recover=undefined to
compilations, so that undefined behavior violations will be sure to
cause test failures.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* README: Mention --enable-ubsan.
* NEWS: Mention --enable-ubsan.
* acinclude.m4: Include sanitize.m4.
* configure: Rebuild.
* configure.ac: Call AM_GDB_UBSAN.
* sanitize.m4: New file.
gdb/doc/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* gdb.texinfo (Configure Options): Document --enable-ubsan.
-fsanitize=undefined pointed out undefined behavior in
dump_raw_expression like:
runtime error: load of value 2887952, which is not a valid value for type 'exp_opcode'
dump_raw_expression will try to print the opcode for each element of
the expression, even when it is not valid. To allow this, but have it
avoid undefined behavior, this patch sets the underlying type of enum
exp_opcode, and arranges for op_name to handle invalid opcodes more
nicely.
Before this patch, debug-expr.exp shows:
Dump of expression @ 0x60f000007750, before conversion to prefix form:
Language c, 8 elements, 16 bytes each.
Index Opcode Hex Value String Value
0 OP_TYPE 89 Y...............
<unknown 3851920> 107820862850704 ..:..b..........
2 OP_TYPE 89 Y...............
3 OP_VAR_VALUE 40 (...............
4 <unknown 2807568> 107820861806352 ..*..b..........
5 <unknown 2806368> 107820861805152 `.*..b..........
6 OP_VAR_VALUE 40 (...............
7 UNOP_MEMVAL_TYPE 57 9...............
Afterward, the output is:
Dump of expression @ 0x4820f90, before conversion to prefix form:
Language c, 8 elements, 16 bytes each.
Index Opcode Hex Value String Value
0 OP_TYPE 89 Y...............
1 unknown opcode: 176 75444400 .0..............
2 OP_TYPE 89 Y...............
3 OP_VAR_VALUE 40 (...............
4 OP_BOOL 74616912 P.r.............
5 unknown opcode: 128 74615680 ..r.............
6 OP_VAR_VALUE 40 (...............
7 UNOP_MEMVAL_TYPE 57 9...............
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* expression.h (enum exp_opcode): Use uint8_t as base type.
* expprint.c (op_name): Handle invalid opcodes.
-fsanitize=undefined pointed out this error:
runtime error: load of value 2887952, which is not a valid value for type 'exp_opcode'
This happens in gdb.ada/complete.exp when processing "complete p
my_glob". This does not parse, so the Ada parser throws an exception;
but then the code in parse_exp_in_context_1 accepts the expression
anyway. However, as no elements have been written to the expression,
undefined behavior results.
The fix is to notice this case in parse_exp_in_context_1. This patch
also adds an assertion to prefixify_expression to enforce this
pre-existing constraint.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* parse.c (prefixify_expression): Add assert.
(parse_exp_in_context_1): Throw exception if the expression is
empty.
-fsanitize=undefined pointed out that read_signed_leb128 had an
undefined left-shift when processing the final byte of a 64-bit leb:
runtime error: left shift of 127 by 63 places cannot be represented in type 'long int'
and an undefined negation:
runtime error: negation of -9223372036854775808 cannot be represented in type 'long int'; cast to an unsigned type to negate this value to itself
Both of these problems are readily avoided by havinng
read_signed_leb128 work in an unsigned type, and then casting to the
signed type at the return.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* dwarf2read.c (read_signed_leb128): Work in ULONGEST.
-fsanitize=undefined pointed out that c-exp.y relied on undefined
behavior here:
if (c != 'l' && c != 'u')
n *= base;
...when a large hex constant "just fit" into a LONGEST, causing the
high bit to be set.
This fixes the problem by having the function work in an unsigned
type.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* c-exp.y (parse_number): Work in unsigned. Remove casts.
-fsanitize=undefined pointed out an undefined shift of a negative
value in read_subrange_type. The fix is to do the work in an unsigned
type, where this is defined.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* dwarf2read.c (read_subrange_type): Make "negative_mask"
unsigned.
-fsanitize=undefined showed that extract_integer could left-shift a
negative value, which is undefined. This patch fixes the problem by
doing all the work in an unsigned type. This relies on
implementation-defined behavior, but I tend to think we are on safe
ground there. (Also, if need be, violations of this could probably be
detected, either by configure or by a static_assert.)
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* findvar.c (extract_integer): Do work in an unsigned type.
-fsanitize=undefined complains about using operator~ on various enum
types that are used with DEF_ENUM_FLAGS_TYPE. This patch fixes these
problems by explicitly setting the base type for these enums to
unsigned. It also adds a static assert to enum_flags to ensure that
future enums used this way have an unsigned underlying type.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* common/enum-flags.h (enum_flags::operator~): Add static assert.
* symfile-add-flags.h (enum symfile_add_flag): Use unsigned as
base type.
* objfile-flags.h (enum objfile_flag): Use unsigned as base type.
* gdbtypes.h (enum type_instance_flag_value): Use unsigned as base
type.
* c-lang.h (enum c_string_type_values): Use unsigned as base
type.
* btrace.h (enum btrace_thread_flag): Use unsigned as base type.
This changes dwarf2_frame_state_reg_info::reg to be a std::vector.
This avoids passing NULL to memcpy in the copy constructor when the
original object does not have any registers.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* dwarf2-frame.h (dwarf2_frame_state_reg_info)
<~dwarf2_frame_state_reg_info>: Update.
<dwarf2_frame_state_reg_info>: Update.
<alloc_regs>: Add assertion. Update.
<reg>: Now a std::vector.
<num_regs>: Remove.
<swap>: Update.
* dwarf2-frame.c (dwarf2_restore_rule, execute_cfa_program)
(execute_cfa_program_test, dwarf2_frame_cache): Update.
-fsanitize=undefined pointed out a spot that passes NULL to memcpy,
which is undefined behavior according to the C standard.
gdb/ChangeLog
2018-10-03 Tom Tromey <tom@tromey.com>
* namespace.c (add_using_directive): Don't pass NULL to memcpy.
Add testcases to verify that all SHT_NOTE sections with the same section
alignment are placed in a single PT_NOTE segment.
PR ld/23658
* testsuite/ld-elf/pr23658-1.d: New file.
* testsuite/ld-elf/pr23658-1a.s: Likewise.
* testsuite/ld-elf/pr23658-1b.s: Likewise.
* testsuite/ld-elf/pr23658-1c.s: Likewise.
* testsuite/ld-elf/pr23658-1d.s: Likewise.
* testsuite/ld-elf/pr23658-2.rd: Likewise.
* testsuite/ld-elf/pr23658-3.d: Likewise.
* testsuite/ld-elf/pr23658-3.s: Likewise.
* testsuite/ld-elf/pr23658-3.t: Likewise.
* testsuite/ld-elf/shared.exp: Run PR ld/23658 tests.
To support putting all adjacent SHT_NOTE sections with the same section
alignment into a single PT_NOTE segment, lang_insert_orphan must group
and sort output note sections by section alignments in both output
section list as well as output section statement list.
PR ld/23658
* ldlang.c (lang_insert_orphan): Group and sort output note
sections by section alignments.
Alignments of SHT_NOTE sections can be 8 bytes for 64-bit ELF files. We
should put all adjacent SHT_NOTE sections with the same section alignment
into a single PT_NOTE segment even when the section alignment != 4 bytes.
Also check SHT_NOTE section type instead of section name.
PR ld/23658
* elf.c (get_program_header_size): Put all adjacent SHT_NOTE
sections with the same section alignment into a single PT_NOTE
segment. Check SHT_NOTE section type instead of section name.
(_bfd_elf_map_sections_to_segments): Likewise.
This patch adds the tests and expected output for each of the conditions where
the MOVPRFX constraint should apply.
The specific test cases are all documented to indicate what the expected
behavior should be.
gas/
* testsuite/gas/aarch64/sve-movprfx_1.d: New test.
* testsuite/gas/aarch64/sve-movprfx_1.s: New test.
* testsuite/gas/aarch64/sve-movprfx_10.d: New test.
* testsuite/gas/aarch64/sve-movprfx_10.l: New test.
* testsuite/gas/aarch64/sve-movprfx_10.s: New test.
* testsuite/gas/aarch64/sve-movprfx_11.d: New test.
* testsuite/gas/aarch64/sve-movprfx_11.s: New test.
* testsuite/gas/aarch64/sve-movprfx_12.d: New test.
* testsuite/gas/aarch64/sve-movprfx_12.s: New test.
* testsuite/gas/aarch64/sve-movprfx_13.d: New test.
* testsuite/gas/aarch64/sve-movprfx_13.l: New test.
* testsuite/gas/aarch64/sve-movprfx_13.s: New test.
* testsuite/gas/aarch64/sve-movprfx_14.d: New test.
* testsuite/gas/aarch64/sve-movprfx_14.l: New test.
* testsuite/gas/aarch64/sve-movprfx_14.s: New test.
* testsuite/gas/aarch64/sve-movprfx_15.d: New test.
* testsuite/gas/aarch64/sve-movprfx_15.l: New test.
* testsuite/gas/aarch64/sve-movprfx_15.s: New test.
* testsuite/gas/aarch64/sve-movprfx_16.d: New test.
* testsuite/gas/aarch64/sve-movprfx_16.s: New test.
* testsuite/gas/aarch64/sve-movprfx_17.d: New test.
* testsuite/gas/aarch64/sve-movprfx_17.l: New test.
* testsuite/gas/aarch64/sve-movprfx_17.s: New test.
* testsuite/gas/aarch64/sve-movprfx_18.d: New test.
* testsuite/gas/aarch64/sve-movprfx_18.l: New test.
* testsuite/gas/aarch64/sve-movprfx_18.s: New test.
* testsuite/gas/aarch64/sve-movprfx_19.d: New test.
* testsuite/gas/aarch64/sve-movprfx_19.s: New test.
* testsuite/gas/aarch64/sve-movprfx_2.d: New test.
* testsuite/gas/aarch64/sve-movprfx_2.l: New test.
* testsuite/gas/aarch64/sve-movprfx_2.s: New test.
* testsuite/gas/aarch64/sve-movprfx_20.d: New test.
* testsuite/gas/aarch64/sve-movprfx_20.l: New test.
* testsuite/gas/aarch64/sve-movprfx_20.s: New test.
* testsuite/gas/aarch64/sve-movprfx_21.d: New test.
* testsuite/gas/aarch64/sve-movprfx_21.s: New test.
* testsuite/gas/aarch64/sve-movprfx_22.d: New test.
* testsuite/gas/aarch64/sve-movprfx_22.l: New test.
* testsuite/gas/aarch64/sve-movprfx_22.s: New test.
* testsuite/gas/aarch64/sve-movprfx_23.d: New test.
* testsuite/gas/aarch64/sve-movprfx_23.l: New test.
* testsuite/gas/aarch64/sve-movprfx_23.s: New test.
* testsuite/gas/aarch64/sve-movprfx_24.d: New test.
* testsuite/gas/aarch64/sve-movprfx_24.l: New test.
* testsuite/gas/aarch64/sve-movprfx_24.s: New test.
* testsuite/gas/aarch64/sve-movprfx_25.d: New test.
* testsuite/gas/aarch64/sve-movprfx_25.l: New test.
* testsuite/gas/aarch64/sve-movprfx_25.s: New test.
* testsuite/gas/aarch64/sve-movprfx_26.d: New test.
* testsuite/gas/aarch64/sve-movprfx_26.l: New test.
* testsuite/gas/aarch64/sve-movprfx_26.s: New test.
* testsuite/gas/aarch64/sve-movprfx_3.d: New test.
* testsuite/gas/aarch64/sve-movprfx_3.l: New test.
* testsuite/gas/aarch64/sve-movprfx_3.s: New test.
* testsuite/gas/aarch64/sve-movprfx_4.d: New test.
* testsuite/gas/aarch64/sve-movprfx_4.l: New test.
* testsuite/gas/aarch64/sve-movprfx_4.s: New test.
* testsuite/gas/aarch64/sve-movprfx_5.d: New test.
* testsuite/gas/aarch64/sve-movprfx_5.s: New test.
* testsuite/gas/aarch64/sve-movprfx_6.d: New test.
* testsuite/gas/aarch64/sve-movprfx_6.l: New test.
* testsuite/gas/aarch64/sve-movprfx_6.s: New test.
* testsuite/gas/aarch64/sve-movprfx_7.d: New test.
* testsuite/gas/aarch64/sve-movprfx_7.l: New test.
* testsuite/gas/aarch64/sve-movprfx_7.s: New test.
* testsuite/gas/aarch64/sve-movprfx_8.d: New test.
* testsuite/gas/aarch64/sve-movprfx_8.l: New test.
* testsuite/gas/aarch64/sve-movprfx_8.s: New test.
* testsuite/gas/aarch64/sve-movprfx_9.d: New test.
* testsuite/gas/aarch64/sve-movprfx_9.l: New test.
* testsuite/gas/aarch64/sve-movprfx_9.s: New test.
This patch wires in the new constraint verifiers into the assembler and
disassembler. Because of this the MOVPRFX tests have to be split out from the
generic SVE tests into their own tests so warnings can be ignored.
These tests are only intended to test the encoding correctness and not the
constraints.
gas/
* testsuite/gas/aarch64/sve-movprfx.d: New test.
* testsuite/gas/aarch64/sve-movprfx.s: New test.
* testsuite/gas/aarch64/sve.d: Refactor.
* testsuite/gas/aarch64/sve.s: Refactor.
* testsuite/gas/aarch64/sysreg-diagnostic.d: Update.
opcodes/
* aarch64-asm.c (aarch64_opcode_encode): Apply constraint verifier.
* aarch64-dis.c (print_operands): Refactor to take notes.
(print_verifier_notes): New.
(print_aarch64_insn): Apply constraint verifier.
(print_insn_aarch64_word): Update call to print_aarch64_insn.
* aarch64-opc.c (aarch64_print_operand): Remove attribute, update notes format.
Any open sequence at the end of a section or assembly is considered an error.
This patch adds a check at the end to ensure that all sequences have been closed
and if not reports a warning.
During disassembly it's not possible to detect this condition in the back-end so
the warning is only emitted from the assembler for now.
gas/
* config/tc-aarch64.c (force_automatic_sequence_close,
aarch64_frob_section): New.
* config/tc-aarch64.h (tc_frob_section, aarch64_frob_section): New.
This patch adds the verification rules for move prefix constraints.
The Arm SVE instruction MOVPRFX introduces[1] constraints on the instruction at
PC+4. Particularly the following constraints are handled by this patch
* MOVPRFX must be followed by an instruction.
* MOVPRFX can only be followed by non-layout altering directives.
* MOVPRFX destination register MUST be used as the destination register in the
instruction at PC+4, and is not allowed to be used in any other position other than
destructive input. This includes registers that architecturally overlap. e.g. x1
should be treated as z1.
* MOVPRFX must be followed by a restricted set of SVE instructions.
* The size of the destination register of MOVPRFX must be equal to that of
the operation at PC+4.
* The predicate register and operation of MOVPRFX must match that of the instruction
at PC+4
* The predicated instruction at PC+4 must use the merging predicate.
* Architectural aliases and pseudo-instructions need to be supported as well.
* MOVPRFX cannot be the last instruction in a sequence
Any failure to adhere to any of these constrains will emit an assembly warning
and a disassembly note.
[1] https://developer.arm.com/docs/ddi0584/latest/arm-architecture-reference-manual-supplement-the-scalable-vector-extension-sve-for-armv8-a
include/
* opcode/aarch64.h (aarch64_inst): Remove.
(enum err_type): Add ERR_VFI.
(aarch64_is_destructive_by_operands): New.
(init_insn_sequence): New.
(aarch64_decode_insn): Remove param name.
opcodes/
* aarch64-opc.c (init_insn_block): New.
(verify_constraints, aarch64_is_destructive_by_operands): New.
* aarch64-opc.h (verify_constraints): New.
gas/
* config/tc-aarch64.c (output_operand_error_report): Order warnings.
The current verifiers only take an instruction description and encoded value as
arguments. This was enough when the verifiers only needed to do simple checking
but it's insufficient for the purposes of validating instruction sequences.
This patch adds the required arguments and also a flag to allow a verifier to
distinguish between whether it's being run during encoding or decoding. It also
allows for errors and warnings to be returned by a verifier instead of a simple
pass/fail.
include/
* opcode/aarch64.h (struct aarch64_opcode): Expand verifiers to take
more arguments.
opcodes/
* aarch64-dis.c (aarch64_opcode_decode): Update verifier call.
* aarch64-opc.c (verify_ldpsw): Update arguments.
Previously the ERR_ values were defined as different constants, to make this a
bit more type safe and so they can be more easily re-used I'm changing them into
an actual enum and updating any usages.
include/
* opcode/aarch64.h (enum err_type): New.
(aarch64_decode_insn): Use it.
opcodes/
* aarch64-dis.c (ERR_OK, ERR_UND, ERR_UNP, ERR_NYI): Remove.
(aarch64_decode_insn, print_insn_aarch64_word): Use err_type.
This patch introduces aarch64_instr_sequence which is a structure similar to IT
blocks on Arm in order to track instructions that introduce a constraint or
dependency on instruction 1..N positions away from the instruction that opened
the block.
The struct is also wired through to the locations that require it.
gas/
* config/tc-aarch64.c (now_instr_sequence):
(*insn_sequence, now_instr_sequence): New.
(output_operand_error_record, do_encode): Add insn_sequence.
(md_assemble): Update insn_sequence.
(try_to_encode_as_unscaled_ldst, fix_mov_imm_insn, fix_insn):
Pass insn_sequence.
* config/tc-aarch64.h (struct aarch64_segment_info_type):
Add insn_sequence.
include/
* opcode/aarch64.h (struct aarch64_instr_sequence): New.
(aarch64_opcode_encode): Use it.
opcodes/
* aarch64-asm.c (aarch64_opcode_encode): Add insn_sequence.
* aarch64-dis.c (insn_sequence): New.
This patch series is to allow certain instructions such as the SVE MOVPRFX
instruction to apply a constraint/dependency on the instruction at PC+4.
This patch starts this off by marking which instructions impose the constraint
and which instructions must adhere to the constraint. This is done in a
generic way by extending the verifiers.
* The constraint F_SCAN indicates that an instruction opens a sequence and imposes
a constraint on an instructions following it. The length of the sequence depends
on the instruction itself and it handled in the verifier code.
* The C_SCAN_MOVPRFX flag is used to indicate which constrain the instruction is
checked against. An instruction with both F_SCAN and C_SCAN_MOVPRFX starts a
block for the C_SCAN_MOVPRFX instruction, and one with only C_SCAN_MOVPRFX must
adhere to a previous block constraint is applicable.
The SVE instructions in this list have been marked according to the SVE
specification[1].
[1] https://developer.arm.com/docs/ddi0584/latest/arm-architecture-reference-manual-supplement-the-scalable-vector-extension-sve-for-armv8-a
include/
* opcode/aarch64.h (struct aarch64_opcode): Add constraints,
extend flags field size.
(F_SCAN, C_SCAN_MOVPRFX, C_MAX_ELEM): New.
opcodes/
* aarch64-tbl.h (CORE_INSN, __FP_INSN, SIMD_INSN, CRYP_INSN, _CRC_INSN,
_LSE_INSN, _LOR_INSN, RDMA_INSN, FF16_INSN, SF16_INSN, V8_2_INSN,
_SVE_INSN, V8_3_INSN, CNUM_INSN, RCPC_INSN, SHA2_INSN, AES_INSN,
V8_4_INSN, SHA3_INSN, SM4_INSN, FP16_V8_2_INSN, DOT_INSN): Initialize
constraints.
(_SVE_INSNC): New.
(struct aarch64_opcode): (fjcvtzs, ldpsw, ldpsw, esb, psb): Initialize
constraints.
(movprfx): Change _SVE_INSN into _SVE_INSNC, add C_SCAN_MOVPRFX and
F_SCAN flags.
(msb, mul, neg, not, orr, rbit, revb, revh, revw, sabd, scvtf,
sdiv, sdivr, sdot, smax, smin, smulh, splice, sqadd, sqdecd, sqdech,
sqdecp, sqdecw, sqincd, sqinch, sqincp, sqincw, sqsub, sub, subr, sxtb,
sxth, sxtw, uabd, ucvtf, udiv, udivr, udot, umax, umin, umulh, uqadd,
uqdecd, uqdech, uqdecp, uqdecw, uqincd, uqinch, uqincp, uqincw, uqsub,
uxtb, uxth, uxtw, bic, eon, orn, mov, fmov): Change _SVE_INSN into _SVE_INSNC and add
C_SCAN_MOVPRFX and C_MAX_ELEM constraints.
I tried a build on macOS today and it failed due to a mismatch between
the printf format and the type in aarch64-linux-tdep.c. This patch
fixes the problem by using pulongest and %s rather than %ld.
gdb/ChangeLog
2018-10-02 Tom Tromey <tom@tromey.com>
* aarch64-linux-tdep.c (aarch64_linux_sigframe_init): Use pulongest.
The RISC-V memory model has been ratified, and it includes an additional
fence: "fence.tso". This pseudo instruction extends one of the
previously reserved full fence patterns to be less restrictive, and
therefor will execute correctly on all existing microarchitectures.
Thus there is no reason to allow this instruction to be disabled (or
unconverted to a full fence), so it's just unconditionally allowed.
I've added a test case for GAS to check that "fence.tso" correctly
assembles on rv32i-based targets. I checked to see that "fence.tso"
appears in "gas.log", but that's the only testing I've done.
gas/ChangeLog
2018-10-02 Palmer Dabbelt <palmer@sifive.com>
* testsuite/gas/riscv/fence-tso.d: New file.
* testsuite/gas/riscv/fence-tso.s: Likewise.
include/ChangeLog
2018-10-02 Palmer Dabbelt <palmer@sifive.com>
* opcode/riscv-opc.h (MATCH_FENCE_TSO): New define.
(MASK_FENCE_TSO): Likewise.
opcodes/ChangeLog
2018-10-02 Palmer Dabbelt <palmer@sifive.com>
* riscv-opc.c (riscv_opcodes) <fence.tso>: New opcode.
