2018-08-20 Richard Biener <rguenther@suse.de>
PR tree-optimization/78655
* tree-vrp.c (extract_range_from_binary_expr_1): Make
pointer + offset nonnull if either operand is nonnull work.
* gcc.dg/tree-ssa/evrp11.c: New testcase.
From-SVN: r263662
The _Tuple_impl base class can be used to disable copy/move assignment,
without requiring an extra base class.
Exception specifications on std::tuple assignment and swap functions can
be defined directly using is_nothrow_swappable, instead of querying the
base classes.
PR libstdc++/86963
* include/std/tuple (_Tuple_impl::operator=): Define as deleted.
(_Tuple_impl::_M_assign): New functions to perform assignment instead
of assignment operators.
(_Tuple_impl::_M_swap): Remove exception specification.
(_Tuple_impl<_Idx, _Head>): Likewise.
(_TC::_NonNestedTuple, _TC::_NotSameTuple): Use __remove_cvref_t.
(__tuple_base): Remove.
(tuple, tuple<_T1, _T2>): Remove inheritance from __tuple_base.
(tuple::operator=, tuple<_T1, _T2>::operator=): Call _M_assign.
(tuple::swap, tuple<_T1, _T2>::swap): Define exception specification
using __is_nothrow_swappable.
(tuple<_T1, _T2>::tuple(_U1&&, _U2&&)): Use __remove_cvref_t.
From-SVN: r263661
Atm, when running vla-1.c with -O0 -flto, we have:
...
FAIL: gcc.dg/guality/vla-1.c -O0 -flto -fuse-linker-plugin \
-fno-fat-lto-objects line 17 sizeof (a) == 6
...
The vla a[i + 1] in f1 is gimplified into:
...
f1 (int i)
{
char a[0:D.1922] [value-expr: *a.0];
char[0:D.1922] * a.0;
D.1921 = i + 1;
D.1926 = (sizetype) D.1921;
a.0 = __builtin_alloca_with_align (D.1926, 8);
...
The early debug info for the upper bound of the type of vla a that we stream
out is:
...
DIE 0: DW_TAG_subrange_type (0x7f85029a90f0)
DW_AT_upper_bound: location descriptor:
(0x7f85029a9230) DW_OP_GNU_variable_value die -> 0 (0x7f85029a94b0), 0
DIE 0: DW_TAG_variable (0x7f85029a94b0)
DW_AT_name: "D.1922"
DW_AT_type: die -> 0 (0x7f85029a3d70)
DW_AT_artificial: 1
...
and in ltrans we have for that same upper bound:
...
DIE 0: DW_TAG_subrange_type (0x7f5183b57d70)
DW_AT_upper_bound: die -> 0 (0x7f5183b576e0)
DIE 0: DW_TAG_variable (0x7f5183b576e0)
DW_AT_name: "D.4278"
DW_AT_abstract_origin: die -> label: vla_1.c.6719312a + 193 (0x7f5183b57730)
...
where D.4278 has abstract origin D.1922.
The D.4278 die has no DW_AT_location, so when evaluting "sizeof (a)" in the
debugger, we can't find the information to get the value of D.4278, and the
debugger prints "<optimized out>".
This patch fixes that by either:
- adding DW_AT_location to the referenced variable die, or
- instead of using a ref for the upper bound, using an exprloc.
When changing gcc.dg/guality/guality.exp to run the usual flto flavours
"-fno-use-linker-plugin -flto-partition=none" and "-fuse-linker-plugin
-fno-fat-lto-objects" in combination with O0, Og, O1, O2, O3 and Os, this patch
fixes all (20) failures in vla-1.c, leaving only:
...
No symbol "i" in current context.
UNSUPPORTED: gcc.dg/guality/vla-1.c -O3 -flto -fno-use-linker-plugin \
-flto-partition=none line 17 i == 5
'a' has unknown type; cast it to its declared type
UNSUPPORTED: gcc.dg/guality/vla-1.c -O3 -flto -fno-use-linker-plugin \
-flto-partition=none line 17 sizeof (a) == 6
...
Bootstrapped and reg-tested on x86_64.
2018-08-20 Tom de Vries <tdevries@suse.de>
* dwarf2out.c (add_scalar_info): Don't add reference to existing die
unless the referenced die describes the added property using
DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
Otherwise, add a DW_AT_location to the referenced die.
From-SVN: r263660
* include/std/optional (_Optional_payload): Use variable templates
for conditions in default template arguments and exception
specifications.
(optional): Likewise. Adjust indentation.
(optional::__not_self, optional::__not_tag, optional::_Requires): New
SFINAE helpers.
(optional::optional): Use new helpers in constructor constraints.
* include/std/type_traits (__or_v, __and_v): New variable templates.
* testsuite/20_util/optional/cons/value_neg.cc: Change dg-error to
dg-prune-output. Remove unused header.
From-SVN: r263657
PR target/86994
* config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
register_operand when calling ix86_set_reg_reg_cost.
[case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
Set *total to 0 for operands that satisfy x86_64_immediate_operand
predicate and to 1 otherwise.
From-SVN: r263652
2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
gcc/
* config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
emit a diagnostic that it is not supported and reset the option.
* config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
supported and consume the option. (ASM_FINAL_SPEC): New.
gcc/testsuite
* g++.dg/debug/dwarf2/pr85302.C: Skip unsupported split DWARF
test on Darwin.
* g++.dg/debug/dwarf2/pr85302.C: Likewise.
* gcc.dg/lto/pr83719_0.c: Likewise.
From-SVN: r263645
Consider:
extern int callee (const char *one, char *two, const char *three);
int test ()
{
return callee ("first", "second", "third");
}
for which -Wwrite-strings was emitting:
Wwrite-strings.C: In function 'int test()':
Wwrite-strings.C:10:44: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
10 | return callee ("first", "second", "third");
| ^
This patch fixes the warning so that it underlines the pertinent argument
at the callsite:
Wwrite-strings.C: In function 'int test()':
Wwrite-strings.C:10:27: warning: ISO C++ forbids converting a string constant to 'char*' [-Wwrite-strings]
10 | return callee ("first", "second", "third");
| ^~~~~~~~
Ideally we ought to also issue a "note" highlighting the pertinent
parameter within the decl, but that's not readily available, so I'm
saving it for another patch.
gcc/cp/ChangeLog:
* typeck.c (string_conv_p): Extract location from EXP and use it
in preference to input_location when issuing warnings.
gcc/testsuite/ChangeLog:
* g++.dg/conversion/Wwrite-strings.C: New test.
From-SVN: r263635
This patch converts dump_print and dump_printf_loc from using
printf (and thus ATTRIBUTE_PRINTF) to using a new pretty-printer
based on pp_format, which supports formatting middle-end types.
In particular, the following codes are implemented (in addition
to the standard pretty_printer ones):
%E: gimple *:
Equivalent to: dump_gimple_expr (MSG_*, TDF_SLIM, stmt, 0)
%G: gimple *:
Equivalent to: dump_gimple_stmt (MSG_*, TDF_SLIM, stmt, 0)
%T: tree:
Equivalent to: dump_generic_expr (MSG_*, arg, TDF_SLIM).
Hence it becomes possible to convert e.g.:
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: different sized vector "
"types in statement, ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM, vectype);
dump_printf (MSG_MISSED_OPTIMIZATION, " and ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM, nunits_vectype);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
into a one-liner:
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: different sized vector "
"types in statement, %T and %T\n",
vectype, nunits_vectype);
Unlike regular pretty-printers, this one captures optinfo_item
instances for the formatted chunks as appropriate, so that when
written out to a JSON optimization record, the relevant parts of
the message are labelled by type, and by source location (so that
e.g. %G is entirely equivalent to using dump_gimple_stmt).
dump_printf and dump_printf_loc become marked with
ATTRIBUTE_GCC_DUMP_PRINTF, which the patch also implements.
gcc/c-family/ChangeLog:
* c-format.c (enum format_type): Add gcc_dump_printf_format_type.
(gcc_dump_printf_length_specs): New.
(gcc_dump_printf_flag_pairs): New.
(gcc_dump_printf_flag_specs): New.
(gcc_dump_printf_char_table): New.
(format_types_orig): Add entry for "gcc_dump_printf".
(init_dynamic_diag_info): Set up length_char_specs and
conversion_specs for gcc_dump_printf_format_type.
(handle_format_attribute): Handle gcc_dump_printf_format_type.
gcc/ChangeLog:
* dump-context.h: Include "dumpfile.h".
(dump_context::dump_printf_va): Convert final param from va_list
to va_list *. Convert from ATTRIBUTE_PRINTF to
ATTRIBUTE_GCC_DUMP_PRINTF.
(dump_context::dump_printf_loc_va): Likewise.
* dumpfile.c: Include "stringpool.h".
(make_item_for_dump_printf_va): Delete.
(make_item_for_dump_printf): Delete.
(class dump_pretty_printer): New class.
(dump_pretty_printer::dump_pretty_printer): New ctor.
(dump_pretty_printer::emit_items): New member function.
(dump_pretty_printer::emit_any_pending_textual_chunks): New member
function.
(dump_pretty_printer::emit_item): New member function.
(dump_pretty_printer::stash_item): New member function.
(dump_pretty_printer::format_decoder_cb): New member function.
(dump_pretty_printer::decode_format): New member function.
(dump_context::dump_printf_va): Reimplement in terms of
dump_pretty_printer.
(dump_context::dump_printf_loc_va): Convert final param from va_list
to va_list *.
(dump_context::begin_scope): Reimplement call to
make_item_for_dump_printf.
(dump_printf): Update for change to dump_printf_va.
(dump_printf_loc): Likewise.
(selftest::test_capture_of_dump_calls): Convert "stmt" from
greturn * to gimple *. Add a test_decl. Add tests of dump_printf
with %T, %E, and %G.
* dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
(dump_printf): Replace ATTRIBUTE_PRINTF_2 with
ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
(dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
* tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
within a dump_printf_loc call to "%wu".
(vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
missing space after "=".
* tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
* tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
* tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
duplicate "vectorized" from message.
gcc/testsuite/ChangeLog:
* gcc.dg/format/gcc_diag-1.c: Fix typo. Add test coverage for
gcc_dump_printf.
* gcc.dg/format/gcc_diag-10.c: Add gimple typedef. Add test
coverage for gcc_dump_printf.
From-SVN: r263626
PR libstdc++/86963
* include/std/tuple (__tuple_base): New class template with deleted
copy assignment operator.
(tuple, tuple<_T1, _T2>): Derive from __tuple_base<tuple> so that
implicit copy/move assignment operator will be deleted/suppressed.
(tuple::__assignable, tuple<_T1, _T2>::__assignable): New helper
functions for SFINAE constraints on assignment operators.
(tuple::__nothrow_assignable, tuple<_T1, _T2>::__nothrow_assignable):
New helper functions for exception specifications.
(tuple::operator=(const tuple&), tuple::operator=(tuple&&))
(tuple<_T1, _T2>::operator=(const tuple&))
(tuple<_T1, _T2>::operator=(tuple&&)): Change parameter types to
__nonesuch_no_braces when the operator should be defined implicitly.
Use __nothrow_assignable for exception specifications.
(tuple::operator=(const tuple<_UElements...>&))
(tuple::operator=(tuple<_UElements...>&&))
(tuple<_T1, _T2>::operator=(const tuple<_U1, _U2>&))
(tuple<_T1, _T2>::operator=(tuple<_U1, _U2>&&))
(tuple<_T1, _T2>::operator=(const pair<_U1, _U2>&))
(tuple<_T1, _T2>::operator=(pair<_U1, _U2>&&)): Constrain using
__assignable and use __nothrow_assignable for exception
specifications.
* python/libstdcxx/v6/printers.py (is_specialization_of): Accept
gdb.Type as first argument, instead of a string.
(StdTuplePrinter._iterator._is_nonempty_tuple): New method to check
tuple for expected structure.
(StdTuplePrinter._iterator.__init__): Use _is_nonempty_tuple.
* testsuite/20_util/tuple/dr2729.cc: New test.
* testsuite/20_util/tuple/element_access/get_neg.cc: Change dg-error
to dg-prune-output.
From-SVN: r263625
Same as r263561, but for arm: avoid compilation errors caused by poly
initializers getting treated as string literals.
Tested on arm-none-linux-gnueabihf.
gcc/ChangeLog:
* config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
polyNxK_t element's TYPE_STRING_FLAG.
From-SVN: r263624
This gives a name to the conditional branch and conditional return
patterns, so that it looks neater in dumps and verbose asm. Also, the
comment for conditional branch was out of date; this fixes it.
* config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
(they were unnamed before). Fix comments.
From-SVN: r263623
2018-08-17 Richard Biener <rguenther@suse.de>
PR middle-end/86505
* tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
across a va-arg-pack using call adjust its return value accordingly.
* gcc.dg/torture/pr86505.c: New testcase.
From-SVN: r263613
gcc/ChangeLog:
PR tree-optimization/86853
* gimple-ssa-sprintf.c (struct format_result): Rename member.
(struct fmtresult): Add member and initialize it in ctors.
(format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
(format_string): Handle %S the same as %ls. Set MAYFAIL.
(format_directive): Set POSUNDER4K when MAYFAIL is set.
(parse_directive): Handle %C same as %c.
(sprintf_dom_walker::compute_format_length): Adjust.
(is_call_safe): Adjust.
gcc/testsuite/ChangeLog:
PR tree-optimization/86853
* gcc.dg/tree-ssa/builtin-sprintf-10.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-11.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-warn-18.c: Adjust.
From-SVN: r263612
Without this patch, the "line span" markers and the line numbering
interacted badly, leading to stray copies of the line-span markers
appearing as prefixes on the first source line in a span:
missing-header-fixit-3.c: In function 'test':
missing-header-fixit-3.c:9:3: warning: implicit declaration of function 'printf' [-Wimplicit-function-declaration]
9 | printf ("%i of %i\n", i, j);
| ^~~~~~
missing-header-fixit-3.c:9:3: warning: incompatible implicit declaration of built-in function 'printf'
missing-header-fixit-3.c:9:3: note: include '<stdio.h>' or provide a declaration of 'printf'
missing-header-fixit-3.c:1:1:
|+#include <stdio.h>
missing-header-fixit-3.c:1:1:1 | /* Example of a fix-it hint that adds a #include directive,
missing-header-fixit-3.c:9:3:
missing-header-fixit-3.c:9:3:9 | printf ("%i of %i\n", i, j);
| ^~~~~~
With this patch, we now correctly print:
missing-header-fixit-3.c: In function 'test':
missing-header-fixit-3.c:9:3: warning: implicit declaration of function 'printf' [-Wimplicit-function-declaration]
9 | printf ("%i of %i\n", i, j);
| ^~~~~~
missing-header-fixit-3.c:9:3: warning: incompatible implicit declaration of built-in function 'printf'
missing-header-fixit-3.c:9:3: note: include '<stdio.h>' or provide a declaration of 'printf'
missing-header-fixit-3.c:1:1:
+ |+#include <stdio.h>
1 | /* Example of a fix-it hint that adds a #include directive,
missing-header-fixit-3.c:9:3:
9 | printf ("%i of %i\n", i, j);
| ^~~~~~
gcc/ChangeLog:
* diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
to emit the span, rather than setting it as the prefix.
gcc/testsuite/ChangeLog:
* gcc.dg/missing-header-fixit-3.c: New test.
From-SVN: r263606
This commit slightly tweaks line-insertion fix-it hints, so that
with line-numbering, rather than e.g.:
99 | x = a;
|+ break;
110 | case 'b':
| ^~~~~~~~
we fill the margin with "+":
99 | x = a;
+++ |+ break;
110 | case 'b':
| ^~~~~~~~
to emphasize that this is a suggested new line, rather than the user's
source.
gcc/ChangeLog:
* diagnostic-show-locus.c (layout::start_annotation_line): Add
"margin_char" parameter, defaulting to space. Use it in place
of pp_space for the initial part of the margin.
(layout::print_leading_fixits): Use '+' when filling the margin
of line-insertion fix-it hints.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-show-locus-bw-line-numbers.c
(test_fixit_insert_newline): Update expected output to show '+'
characters in margin of line-insertion fix-it hint.
From-SVN: r263605
There are some patterns that recognise the parallel of an add and a
compare, and split it back to the same two insns. This apparently
helped RIOS machines before RTL scheduling existed? Either way, it
isn't helpful anymore, and even hurts a tiny bit. So, delete it.
* config/rs6000/rs6000.md (two unnamed define_insn and define_split):
Delete.
From-SVN: r263604
There were many insns that set "length 4" explicitly while that does
not make anything clearer to the reader. So, simplify the code.
* config/rs6000/altivec.md: Don't set length attribute to the default
value.
* config/rs6000/darwin.md: Ditto.
* config/rs6000/dfp.md: Ditto.
* config/rs6000/htm.md: Ditto.
* config/rs6000/rs6000.md: Ditto.
* config/rs6000/sync.md: Ditto.
* config/rs6000/vsx.md: Ditto.
From-SVN: r263603
Now that it is the default for branch insns like for all other insns,
we don't need to set it explicitly so often anymore.
* config/rs6000/altivec.md: Don't set length attribute to the default
value, for branch instructions.
* config/rs6000/darwin.md: Ditto.
* config/rs6000/rs6000.md: Ditto.
From-SVN: r263602
This moves what is currently the default of the length attribute to
the only branch instruction patterns where it applies, namely, the
B-form instructions. It was used for the "jump" instruction as well
before, but that is an I-form instruction and always has length 4.
* config/rs6000/rs6000.md (length): Always define as const_int 4.
(unnamed conditional branch define_insn): Set length to 4 or 8
depending on offset.
(<bd>_<mode>): Similar, for alternative 0.
(<bd>tf_<mode>): Ditto.
From-SVN: r263601
PR c++/70693 reports a crash within -Wmisleading-indentation in
get_visual_column, reading past the end of a source line.
The issue occurs due to a stray carriage return aka '\r' aka ^M, occurring
towards the end of line 35 of attachment 38289 - but not at the end itself.
This carriage return confuses our line numbering: from that point in the
file, the lexer (and thus location_t values) use line numbers that are
one larger than those seen by input.c, "cat -n" and emacs.
This discrepancy between the lexer's line numbering and input.c's line
numbering leads to an out-of-range read in get_visual_column (trying to
read column 8, to locate the first non-whitespace on the line containing
"break;", but finding the next line, which is only 4 characters long).
This patch fixes the ICE by adding a range check to get_visual_column
before accessing the input.c line buffer. This is arguably papering
over the root cause, but there are presumably other ways of triggering
such an out-of-range read by writing to the source file after the lexer
but before -Wmisleading-indentation, and we ought to be not ICE in the
face of that.
gcc/c-family/ChangeLog:
PR c++/70693
* c-common.c (selftest::c_family_tests): Call
selftest::c_indentation_c_tests.
* c-common.h (selftest::c_indentation_c_tests): New decl.
* c-indentation.c: Include "selftest.h".
(next_tab_stop): Add "tab_width" param, rather than accessing
cpp_opts.
(get_visual_column): Likewise. Clarify comment. Bulletproof
against reading past the end of the line.
(get_first_nws_vis_column): Add "tab_width" param.
(detect_intervening_unindent): Likewise.
(should_warn_for_misleading_indentation): Read tab width from
cpp_opts and pass around.
(selftest::test_next_tab_stop): New test.
(selftest::assert_get_visual_column_succeeds): New function.
(ASSERT_GET_VISUAL_COLUMN_SUCCEEDS): New macro.
(selftest::assert_get_visual_column_fails): New function.
(ASSERT_GET_VISUAL_COLUMN_FAILS): New macro.
(selftest::test_get_visual_column): New test.
(selftest::c_indentation_c_tests): New function.
gcc/testsuite/ChangeLog:
PR c++/70693
* c-c++-common/Wmisleading-indentation-pr70693.c: New test.
From-SVN: r263595
This allows copy_blkmode_to_reg to perform larger copies when it is safe to do so by calculating
the bitsize per iteration doing the maximum copy allowed that does not read more
than the amount of bits left to copy.
Strictly speaking, this copying is only done if:
1. the target supports fast unaligned access
2. no padding is being used.
This should avoid the issues of the first patch (PR85123) but still work for targets that are safe
to do so.
Original patch https://gcc.gnu.org/ml/gcc-patches/2017-11/msg01088.html
Previous respin https://gcc.gnu.org/ml/gcc-patches/2018-04/msg00239.html
gcc/
2018-08-16 Tamar Christina <tamar.christina@arm.com>
* expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
From-SVN: r263594
C++17 says to use std::uncaught_exceptions() here instead of
std::uncaught_exception() but since we only care whether the result is
non-zero (and we aren't planning to remove the deprecated version) we
can just keep using std::uncaught_exception() and suppress the warning.
* include/std/ostream (basic_ostream::sentry::~sentry): Suppress
deprecation warnings for using uncaught_exception().
From-SVN: r263593
Committed on behalf of Matthew Malcomson.
2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
gcc/
* doc/rtl.texi: Replace old RTX class names with new names.
From-SVN: r263592
* expmed.h (canonicalize_comparison): New declaration.
* expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
* expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
* optabs.c (prepare_cmp_insn): Likewise.
* rtl.h (unsigned_condition_p): New function which checks if a
comparison operator is unsigned.
* gcc.target/aarch64/imm_choice_comparison.c: New test.
From-SVN: r263591
