[Ada] Unnesting: do not set size of access subprograms

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* layout.adb: Do not set size of access subprogram if unnesting.

From-SVN: r263721

[Ada] Unnesting: do not emit warnings for access to subprograms

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* freeze.adb: Remove warnings for access to subprograms when
unnesting is active.

From-SVN: r263720

[Ada] Improper copying of limited arrays with default initialization

This patch fixes an improper expansion of aggregates for limited array
types in an object declaration. Prior to this patch, The presence of the
aggregate (which can only consist of box initializations) would create a
temporary that was then assigned to the object in the declaration.
Apart from a violation of the semantics of build-in-place limited
objects, this can also lead to out-of-scope access in LLVM.

Executing the following;

   gcc -c -gnatDG nocopy.adb
   grep quintet nocopy.adb.dg | wc -l

must yield:

   5

----
procedure NoCopy is

  --  Task used in this example to test that the limited component
  --  is properly initialized.

  task type T_Task (Disc : Natural);

  task body T_Task is
  begin
     null;
  end T_Task;

  type My_Rec (D : Natural := 9999) is record

     --  Components initialized by means of the current value
     --  of the record discriminant

     T : T_Task (D);
  end record;

  type TR is array (1 .. 5) of My_Rec;
  Quintet : TR := (others => (others => <>));
begin
   null;
end NoCopy;

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* exp_aggr.adb (Expand_Array_Aggregate): If the component type
is limited, the array must be constructed in place, so set flag
In_Place_Assign_OK_For_Declaration accordingly. This prevents
improper copying of an array of tasks during initialization.

From-SVN: r263719

[Ada] Fix spurious error on unchecked conversion on VFA record type

2018-08-21  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* gcc-interface/trans.c (Call_to_gnu): Always suppress an
unchecked conversion around the actual for an In parameter
passed by copy.

From-SVN: r263718

[Ada] Fix internal error on extension of record with representation clause

This fixes a long-standing issue present for extensions of tagged record
types with a representation clause: the clause is correctly inherited
for components inherited in the extension but the position and size are
not, which fools the logic of Is_Possibly_Unaligned_Object.

This can result in an attempt to take the address of a component not
aligned on a byte boundary, which is then flagged as an internal error.

2018-08-21  Eric Botcazou  <ebotcazou@adacore.com>

gcc/ada/

* exp_util.adb (Is_Possibly_Unaligned_Object): For the case of a
selected component inherited in a record extension and subject
to a representation clause, retrieve the position and size from
the original record component.

gcc/testsuite/

* gnat.dg/rep_clause7.adb: New testcase.

From-SVN: r263717

[Ada] Spurious "Duplicated symbol" error with discriminated tasks

This patch fixes a spurious error in a program that contains a
discriminated task type and several of its subtype in the same
declarative part, when the corresponding discriminant constraints are
expressions.

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_util.ads, sem_util.adb (New_External_Entity): Type of
Suffix_Index must be Int, not Nat, so that a negative value can
be used to generate a unique name for an external object, as
specified in Tbuild.New_External_Name.
(Scope_Within): Handle private type whose completion is a
synchronized type (For unnesting).
* itypes.ads, itypes.adb (Create_Itype): Ditto
* sem_ch3.adb (Constrain_Corresponding_Record): Generate a
unique name for the created subtype, because there may be
several discriminated tasks present in the same scope, and each
needs its distinct corresponding record subtype.

gcc/testsuite/

* gnat.dg/task1.adb, gnat.dg/task1.ads, gnat.dg/task1_pkg.adb,
gnat.dg/task1_pkg.ads: New testcase.

From-SVN: r263716

[Ada] Define versions of dimension system for Float and Long_Float

The dimension system in System.Dim.Mks is based on Long_Long_Float,
which may not be convenient to people who want to use Float or
Long_Float as basis. These new files provide units that define the
dimension system based on Float in System.Dim.Float_Mks and on
Long_Float in System.Dim.Long_Mks.

Child packages Other_Prefixes and Mks_IO are also defined appropriately,
with new instantiations for the various sizes of floating-point types.

2018-08-21  Yannick Moy  <moy@adacore.com>

gcc/ada/

* doc/gnat_ugn/gnat_and_program_execution.rst: Update
documentation of dimensionality analysis.
* gnat_ugn.texi: Regenerate.
* Makefile.rtl, impunit.adb: Consider the new units.
* libgnat/s-dfmkio.ads, libgnat/s-dfmopr.ads,
libgnat/s-diflmk.ads: New units based on Float.
* libgnat/s-dilomk.ads, libgnat/s-dlmkio.ads,
libgnat/s-dlmopr.ads: New units based on Long_Float.
* libgnat/s-dmotpr.ads: Rename to libgnat/s-dgmgop.ads and turn
into an instance of
System.Dim.Generic_Mks.Generic_Other_Prefixes.
* libgnat/s-dimmks.ads: Rename to libgnat/s-digemk.ads and turn
into an instance of System.Dim.Generic_Mks.

From-SVN: r263715

[Ada] General purpose doubly linked list for compiler and tool use

This patch adds unit GNAT.Lists which currently contains the
implementation of a general purpose doubly linked list intended for use
by the compiler and the tools around it.

2018-08-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* impunit.adb: Add g-lists to the set of non-implementation
units.
* libgnat/g-lists.adb, libgnat/g-lists.ads: New unit.
* Makefile.rtl: Add g-lists to the set of non-tasking units.
* gcc-interface/Make-lang.in: Add g-lists to the set of files
used by gnat1.

gcc/testsuite/

* gnat.dg/linkedlist.adb: New testcase.

From-SVN: r263714

[Ada] Fix scope computation for entry bodies and accept alternatives

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* exp_ch9.adb (Reset_Scopes): Do not recurse into type
declarations when resetting the scope of entities declared the
procedures generated for entry bodies and accept alternatives.
Use the entity of the procedure declaration, not its body, as
the new scope.

From-SVN: r263713

[Ada] Crash on entry in generic with dynamic elaboration checks

This patch modifies the set of attributes that applies to entries and
entry families to include elaboration entities used by the
access-before-elaboration mechanism.

2018-08-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* einfo.adb (Elaboration_Entity): Include entries and entry
families in the set of legal entities.
(Elaboration_Entity_Required): Include entries and entry
families in the set of legal entities.
(Set_Elaboration_Entity): Include entries and entry families in
the set of legal entities.
(Set_Elaboration_Entity_Required): Include entries and entry
families in the set of legal entities.
(Write_Field13_Name): Update the output of attribute
Elaboration_Entity.
* einfo.ads: Attributes Elaboration_Entity and
Elaboration_Entity_Required now apply to entries and entry
families.

gcc/testsuite/

* gnat.dg/elab6.adb, gnat.dg/elab6.ads, gnat.dg/elab6_pkg.adb,
gnat.dg/elab6_pkg.ads: New testcase.

From-SVN: r263712

[Ada] Set_Targ: add a CodePeer annotation

2018-08-21  Arnaud Charlet  <charlet@adacore.com>

gcc/ada/

* set_targ.adb: Mark some CodePeer message as Intentional.

From-SVN: r263711

[Ada] Compiler abort on call to expr. function for default discriminant

If a discriminant specification has a default that is a call to an
expression function, that function has to be frozen at the point of a
call to the initialization procedure for an object of the record type,
even though the call does not appear to come from source.

2018-08-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

* sem_res.adb (Resolve_Call): Force the freezing of an
expression function that is called to provide a default value
for a defaulted discriminant in an object initialization.

gcc/testsuite/

* gnat.dg/expr_func5.adb: New testcase.

From-SVN: r263710

[Ada] Dynamically resizable, load factor-based hash table

This patch introduces a dynamically resizable, load factor-based hash
table in unit GNAT.Dynamic_HTables.

2018-08-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

* libgnat/g-dynhta.adb, libgnat/g-dynhta.ads: New package
Dynamic_HTable.

gcc/testsuite/

* gnat.dg/dynhash.adb: New testcase.

From-SVN: r263709

[Ada] Enumeration types with non-standard representation

The compiler may report errors on enumeration types with non-standard
representation (i.e. at least one literal has a representation value
different from its 'Pos value) processing attribute 'Enum_Rep.

It may also generate wrong code for the evaluation of 'Enum_Rep raising
Constraint_Error at runtime.

2018-08-21  Javier Miranda  <miranda@adacore.com>

gcc/ada/

* checks.ads (Determine_Range): Adding documentation.
* checks.adb (Determine_Range): Don't deal with enumerated types
with non-standard representation.
(Convert_And_Check_Range): For conversion of enumeration types
with non standard representation to an integer type perform a
direct conversion to the target integer type.

gcc/testsuite/

* gnat.dg/enum4.adb: New testcase.

From-SVN: r263708

[Ada] Handle pragmas that come from aspects for GNATprove

In GNATprove we appear to abuse a routine related to cross-references and
expect it to deliver exact results, which is not what it was designed for.

This patch is a temporary solution to avoid crashes in GNATprove; it doesn't
affect the frontend or other backends, because this code is used exclusively
by GNATprove (it is located in the frontend for technical reasons). No tests
provided.

2018-08-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

* lib-xref.ads, lib-xref-spark_specific.adb
(Enclosing_Subprogram_Or_Library_Package): Now roughtly works
for pragmas that come from aspect specifications.

From-SVN: r263707

[Ada] Add sa_messages.ad[sb] for SPARK 2014

These new source files will make it possible to build SPARK 2014 from
a snapshot of GCC FSF sources.

2018-08-21  Pierre-Marie de Rodat  <derodat@adacore.com>

gcc/ada/

* sa_messages.ads, sa_messages.adb: New source files.

From-SVN: r263706

* testsuite/18_support/new_nothrow.cc: XFAIL on AIX.

From-SVN: r263703

Update global reviewers list

2018-08-21  Richard Sandiford  <richard.sandiford@arm.com>

* MAINTAINERS: Add self to global reviewers list.

From-SVN: r263702

tree.c (find_decls_types_r): Do not check for redundant typedefs.

* tree.c (find_decls_types_r): Do not check for redundant typedefs.
* tree.h (is_redundant_typedef): Remove.
* dwarf2out.c (is_redundant_typedef): Turn into static function.

From-SVN: r263699

[Ada] Bump copyright notices to 2018

2018-08-21  Pierre-Marie de Rodat  <derodat@adacore.com>

gcc/ada/

* libgnarl/a-intnam__dragonfly.ads,
libgnarl/s-osinte__dragonfly.adb,
libgnarl/s-osinte__dragonfly.ads, libgnarl/s-osinte__gnu.adb,
libgnarl/s-osinte__gnu.ads, libgnarl/s-osinte__hpux-dce.adb,
libgnarl/s-osinte__hpux-dce.ads,
libgnarl/s-taprop__hpux-dce.adb,
libgnarl/s-taspri__hpux-dce.ads: Bump copyright notices to 2018.

From-SVN: r263698

tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT when possible.

* tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
when possible.

From-SVN: r263697

Fix DSE big-endian subreg crash on AArch64.

This patch fixes an ICE that would happen when extract_low_bits
is called with modes for which you can't extract a valid subreg.
e.g. taking a 32 bytes subreg from a 48 byte mode.

The ICE happens because convert_modes which eventually calls
simplify_gen_subreg does not expect the convertion to fail.

The assert in gen_lowpart_general would then be hit.  The patch
changes it to validate the subreg before trying to convert the
modes.  If the subreg is not possible we return NULL_RTX and bail
out early.

gcc/ChangeLog:

2018-08-21  Tamar Christina  <tamar.christina@arm.com>

* expmed.c (extract_low_bits): Reject invalid subregs early.

gcc/testsuite/ChangeLog:

2018-08-21  Tamar Christina  <tamar.christina@arm.com>

* gcc.target/aarch64/large_struct_copy.c: New test.

From-SVN: r263696

Remove redundant { dg-do run } directives in tests

These tests accidentally had two dg-do directives. Only the second one
is needed.

* testsuite/26_numerics/bit/bitops.count/countl_one.cc: Remove
redundant dg-do directive.
* testsuite/26_numerics/bit/bitops.count/countl_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_one.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/popcount.cc: Likewise.

From-SVN: r263695

Restrict the pr86763 test to *-*-linux* targets

clock_gettime is not available on some baremetal targets
and may require -lrt on some non-linux targets.

gcc/testsuite/ChangeLog:

* g++.dg/torture/pr86763.C: Restrict to *-*-linux*.

From-SVN: r263694

re PR middle-end/86121 (missing -Wstringop-overflow on strcpy followed by strcat)

2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/86121
        * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
        behaviour.

testsuite:
2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/86121
        * gcc.dg/Wstringop-overflow-6.c: Remove xfail.

From-SVN: r263693

vxworks: don't define vxworks_asm_out_constructor when using .init_array

When the compiler is configured with --enable-initfini-array,
config/initfini-array.h gets included after config/vxworks.h by tm.h, so
the definitions of TARGET_ASM_CONSTRUTOR/TARGET_ASM_DESTRUCTOR in
vxworks.h get undone in initfini-array.h. Hence, we might as well not
define the vxworks_asm_out_* functions.

From-SVN: r263692

vxworks: enable use of .init_array/.fini_array for cdtors

The target OS actually runs all function pointers found in the _ctors
array when the module is loaded. So it is not that hard to make use of
the "modern" .init_array/.fini_array mechanism - it mostly just requires
a linker script adding the _ctors and _dtors symbols and terminating
LONG(0) entries.

Assume that if the user passed --enable-initfini-array when building
gcc, the rest of the toolchain (including the link spec and linker
script) is set up appropriately.

Note that configuring with --enable-initfini-array may prevent the -mrtp
mode from working, due to the (unconditional) use of .init_array.*
sections instead of .ctors.* - however, that is the case regardless of this
patch.

From-SVN: r263691

libgcc: add crt{begin,end} for powerpc-wrs-vxworks target

In order to allow ZCX on VxWorks, we need the frame_dummy function to do
the register_frame_info(). So make sure crtbegin.o and crtend.o are
available for use with a custom spec file.

From-SVN: r263690

vxworks: add target/h/wrn/coreip to the set of system include paths

In order to build crtstuff.c, I need to ensure the headers in
target/h/wrn/coreip are also searched. Of course, that can be done
similar to how wrn/coreip gets manually added for libgcc, e.g. by adding

  CRTSTUFF_T_CFLAGS += -I$(WIND_BASE)/target/h -I$(WIND_BASE)/target/h/wrn/coreip

But without target/h/wrn/coreip in the default search path, all
user-code that include <unistd.h> (crtstuff.c just being one such
example) will have to manually add an -isystem
flag adding the wrn/coreip directory: unistd.h include ioLib.h, which
has

#include "net/uio.h"

and that header is found in target/h/wrn/coreip. In other words, the
VxWorks system headers (at least for VxWorks 5.5) are written in a way
that assumes wrn/coreip is in the search path, so I think it makes sense
to have that by default.

It will change the search order for existing setups that pass
-I.../target/h/wrn/coreip (without -nostdinc), since that flag will now
be ignored. I can't know whether that will break anything, but I do
believe it makes sense to have the defaults actually usable without
expecting all invocations to add -I/-isystem flags.

From-SVN: r263689

gimple-ssa-sprintf.c (decl_constant_value): Remove.

2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * gimple-ssa-sprintf.c (decl_constant_value): Remove.
        (get_format_string): Refer to c_getstr.

From-SVN: r263688

[debug] Add debug and earlydebug dumps

With the introduction of early debug, we've added a phase in the compiler which
produces information which is not visible, unless we run the compiler in the
debugger and call debug_dwarf from dwarf2out_early_finish or some such.

This patch adds dumping of "early" and "final" debug info, into .earlydebug
and .debug dump files, enabled by -fdump-earlydebug and -fdumpdebug, such that
we can follow f.i. the upper bound of a vla type from early debug:
...
          DW_AT_upper_bound: location descriptor:
            (0x7f0d645b7550) DW_OP_GNU_variable_value , 0
...
to final debug:
...
          DW_AT_upper_bound: location descriptor:
            (0x7f0d645b7550) DW_OP_fbreg 18446744073709551592, 0
            (0x7f0d645b7a00) DW_OP_deref 8, 0
...
to -dA annotated assembly file:
...
        .uleb128 0x3    # DW_AT_upper_bound
        .byte   0x91    # DW_OP_fbreg
        .sleb128 -24
        .byte   0x6     # DW_OP_deref
...

The .debug file shows the same information as the annotated assembly, but in
the same format as the "early" debug info.

Bootstrapped and reg-tested on x86_64.

2018-08-21  Tom de Vries  <tdevries@suse.de>

* cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
(debuginfo_start, debuginfo_stop, debuginfo_early_start)
(debuginfo_early_stop): Declare.
* cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
(debuginfo_start, debuginfo_stop, debuginfo_early_start)
(debuginfo_early_stop): New function.
(symbol_table::finalize_compilation_unit): Call debuginfo_early_start
and debuginfo_early_stop.
* dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
* toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
(general_init): Call debuginfo_early_init.
(finalize): Call debuginfo_fini.
(do_compile): Call debuginfo_init.
* doc/invoke.texi (@gccoptlist): Add -fdump-debug and
-fdump-early-debug.
(@item -fdump-debug, @item -fdump-earlydebug): Add.

* lto.c (lto_main):  Call debuginfo_early_start and
debuginfo_early_stop.

* gcc.c-torture/unsorted/dump-noaddr.x: Use -gno-record-gcc-switches
to avoid mismatch in .debug and .earlydebug dump files.

From-SVN: r263687

[debug] Respect fdump-noaddr and fdump-unnumbered in print_die

2018-08-21  Tom de Vries  <tdevries@suse.de>

* dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
flag_dump_noaddr and flag_dump_unnumbered.

From-SVN: r263686

wide-int-range.cc (wide_int_range_abs): New.

* wide-int-range.cc (wide_int_range_abs): New.
(wide_int_range_order_set): Rename from wide_int_range_min_max.
* wide-int-range.h (wide_int_range_abs): New.
(wide_int_range_min_max): New.
* tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
case to call wide_int_range_abs.
Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
(extract_range_from_abs_expr): Delete.

From-SVN: r263685

Daily bump.

From-SVN: r263683

[libiberty patch] PEX-unix forking

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01170.html
* pex-unix.c (pex_child_error): Delete.
(pex_unix_exec_child): Commonize error paths to single message &
exit.

From-SVN: r263679

re PR target/87033 (The compiler does not generate the LWAX instruction)

[gcc]
2018-08-20  Michael Meissner  <meissner@linux.ibm.com>

PR target/87033
* config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
from 'Y' to 'YZ' to enable the LWAX instruction to be generated
for indexed loads.

[gcc/testsuite]
2018-08-20  Michael Meissner  <meissner@linux.ibm.com>

PR target/87033
* gcc.target/powerpc/pr87033.c: New test.

From-SVN: r263678

[PATCH] fix some build breakage

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01214.html
* config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
* config/spu/spu-c.c (spu_macro_to_expand): Likewise.

Co-Authored-By: Jeff Law <law@redhat.com>
From-SVN: r263677

PR tree-optimization/87034 - missing -Wformat-overflow on a sprintf %s with a wide string

gcc/testsuite/ChangeLog:
* gcc.dg/builtin-sprintf-warn-20.c: New test.

From-SVN: r263676

Add support for grouping of related diagnostics (PR other/84889)

We often emit logically-related groups of diagnostics.

A relatively simple case is this:

  if (warning_at (body_loc, OPT_Wmultistatement_macros,
                  "macro expands to multiple statements"))
    inform (guard_loc, "some parts of macro expansion are not guarded by "
            "this %qs clause", guard_tinfo_to_string (keyword));

where the "note" diagnostic issued by the "inform" call
is guarded by the -Wmultistatement_macros warning.

More complicated examples can be seen in the C++ frontend,
where e.g. print_z_candidates can lead to numerous "note"
diagnostics being emitted.

I'm looking at various ways to improve how we handle such related
diagnostics, but, prior to this patch, there was no explicit
relationship between these diagnostics: the diagnostics subsystem
had no way of "knowing" that these were related.

This patch introduces a simple way to group the diagnostics:
an auto_diagnostic_group class: all diagnostics emitted within
the lifetime of an auto_diagnostic_group instance are logically
grouped.

Hence in the above example, the two diagnostics can be grouped
by simply adding an auto_diagnostic_group instance:

  auto_diagnostic_group d;
  if (warning_at (body_loc, OPT_Wmultistatement_macros,
                  "macro expands to multiple statements"))
    inform (guard_loc, "some parts of macro expansion are not guarded by "
            "this %qs clause", guard_tinfo_to_string (keyword));

Some more awkard cases are of the form:

  if (some_condition
      && warning_at (...)
      && more_conditions)
    inform (...);

which thus need restructuring to:

  if (some_condition)
    {
      auto_diagnostic_group d;
      warning_at (...);
      if (more_conditions)
        inform (...);
    }

so that the lifetime of the auto_diagnostic_group groups the
warning_at and the inform call.

Nesting is handled by simply tracking a nesting depth within the
diagnostic_context.: all diagnostics are treated as grouped until the
final auto_diagnostic_group is popped.

diagnostic.c uses this internally, so that all diagnostics are part of
a group - those that are "by themselves" are treated as being part of
a group with one element.

The diagnostic_context gains optional callbacks for displaying the
start of a group (when the first diagnostic is emitted within it), and
the end of a group (for when the group was non-empty); these callbacks
are unused by default, but a test plugin demonstrates them (and verifies
that the machinery is working).

As noted above, I'm looking at various ways to use the grouping to
improve how we output the diagnostics.

FWIW, I experimented with a more involved implementation, of the form:

  diagnostic_group d;
  if (d.warning_at (body_loc, OPT_Wmultistatement_macros,
                    "macro expands to multiple statements"))
    d.inform (guard_loc, "some parts of macro expansion are not guarded by "
              "this %qs clause", guard_tinfo_to_string (keyword));

which had the advantage of allowing auto-detection of the places where
groups were needed (by converting ::warning_at's return type to bool),
but it was a much more invasive patch, especially when dealing with
the places in the C++ frontend that can emit numerous notes after
an error or warning (and thus having to pass the group around)
Hence I went with this simpler approach.

gcc/c-family/ChangeLog:
PR other/84889
* c-attribs.c (common_handle_aligned_attribute): Add
auto_diagnostic_group instance.
* c-indentation.c (warn_for_misleading_indentation): Likewise.
* c-opts.c (c_common_post_options): Likewise.
* c-warn.c (warn_logical_not_parentheses): Likewise.
(warn_duplicated_cond_add_or_warn): Likewise.
(warn_for_multistatement_macros): Likewise.

gcc/c/ChangeLog:
PR other/84889
* c-decl.c (pushtag): Add auto_diagnostic_group instance.
(diagnose_mismatched_decls): Likewise, in various places.
(warn_if_shadowing): Likewise.
(implicit_decl_warning): Likewise.
(implicitly_declare): Likewise.
(undeclared_variable): Likewise.
(declare_label): Likewise.
(grokdeclarator): Likewise.
(start_function): Likewise.
* c-parser.c (c_parser_declaration_or_fndef): Likewise.
(c_parser_parameter_declaration): Likewise.
(c_parser_binary_expression): Likewise.
* c-typeck.c (c_expr_sizeof_expr): Likewise.
(parser_build_binary_op): Likewise.
(build_unary_op): Likewise.
(error_init): Likewise.
(pedwarn_init): Likewise.
(warning_init): Likewise.
(convert_for_assignment): Likewise.

gcc/cp/ChangeLog:
PR other/84889
* call.c (build_user_type_conversion_1): Add auto_diagnostic_group
instance(s).
(print_error_for_call_failure): Likewise.
(build_op_call_1): Likewise.
(build_conditional_expr_1): Likewise.
(build_new_op_1): Likewise.
(build_op_delete_call): Likewise.
(convert_like_real): Likewise.
(build_over_call): Likewise.
(build_new_method_call_1): Likewise.
(joust): Likewise.
* class.c (check_tag): Likewise.
(finish_struct_anon_r): Likewise.
(one_inherited_ctor): Likewise.
(finalize_literal_type_property): Likewise.
(explain_non_literal_class): Likewise.
(find_flexarrays): Likewise.
(resolve_address_of_overloaded_function): Likewise.
* constexpr.c (ensure_literal_type_for_constexpr_object): Likewise.
(is_valid_constexpr_fn): Likewise.
(cx_check_missing_mem_inits): Likewise.
* cp-gimplify.c (cp_genericize_r): Likewise.
* cvt.c (maybe_warn_nodiscard): Likewise.
* decl.c (warn_extern_redeclared_static): Likewise.
(check_redeclaration_exception_specification): Likewise.
(check_no_redeclaration_friend_default_args): Likewise.
(duplicate_decls): Likewise.
(redeclaration_error_message): Likewise.
(warn_misplaced_attr_for_class_type): Likewise.
* decl2.c (finish_static_data_member_decl): Likewise.
(no_linkage_error): Likewise.
(cp_warn_deprecated_use): Likewise.
* error.c (qualified_name_lookup_error): Likewise.
* friend.c (make_friend_class): Likewise.
(do_friend): Likewise.
* init.c (perform_member_init): Likewise.
(build_new_1): Likewise.
(build_vec_delete_1): Likewise.
(build_delete): Likewise.
* lex.c (unqualified_name_lookup_error): Likewise.
* name-lookup.c (check_extern_c_conflict): Likewise.
(inform_shadowed): New function.
(check_local_shadow): Add auto_diagnostic_group instances,
replacing goto "inform_shadowed" label with call to subroutine.
(set_local_extern_decl_linkage): Add auto_diagnostic_group
instance(s).
* parser.c (cp_parser_diagnose_invalid_type_name): Likewise.
(cp_parser_namespace_name): Likewise.
* pt.c (check_specialization_namespace): Likewise.
(check_template_variable): Likewise.
(warn_spec_missing_attributes): Likewise.
(check_explicit_specialization): Likewise.
(process_partial_specialization): Likewise.
(lookup_template_class_1): Likewise.
(finish_template_variable): Likewise.
(do_auto_deduction): Likewise.
* search.c (check_final_overrider): Likewise.
(look_for_overrides_r): Likewise.
* tree.c (maybe_warn_parm_abi): Likewise.
* typeck.c (cxx_sizeof_expr): Likewise.
(cp_build_function_call_vec): Likewise.
(cp_build_binary_op): Likewise.
(convert_for_assignment): Likewise.
(maybe_warn_about_returning_address_of_local): Likewise.
* typeck2.c (abstract_virtuals_error_sfinae): Likewise.
(check_narrowing): Likewise.

gcc/ChangeLog:
PR other/84889
* attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
(decl_attributes): Likewise.
* calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
instance.
* cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
* diagnostic-core.h (class auto_diagnostic_group): New class.
* diagnostic.c (diagnostic_initialize): Initialize the new fields.
(diagnostic_report_diagnostic): Handle the first diagnostics within
a group.
(emit_diagnostic): Add auto_diagnostic_group instance.
(inform): Likewise.
(inform_n): Likewise.
(warning): Likewise.
(warning_at): Likewise.
(warning_n): Likewise.
(pedwarn): Likewise.
(permerror): Likewise.
(error): Likewise.
(error_n): Likewise.
(error_at): Likewise.
(sorry): Likewise.
(fatal_error): Likewise.
(internal_error): Likewise.
(internal_error_no_backtrace): Likewise.
(auto_diagnostic_group::auto_diagnostic_group): New ctor.
(auto_diagnostic_group::~auto_diagnostic_group): New dtor.
* diagnostic.h (struct diagnostic_context): Add fields
"diagnostic_group_nesting_depth",
"diagnostic_group_emission_count", "begin_group_cb",
"end_group_cb".
* gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
Add auto_diagnostic_group instance(s).
(find_explicit_erroneous_behavior): Likewise.
* gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
* gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
* gimplify.c (warn_implicit_fallthrough_r): Likewise.
(gimplify_va_arg_expr): Likewise.
* hsa-gen.c (HSA_SORRY_ATV): Likewise.
(HSA_SORRY_AT): Likewise.
* ipa-devirt.c (compare_virtual_tables): Likewise.
(warn_odr): Likewise.
* multiple_target.c (expand_target_clones): Likewise.
* opts-common.c (cmdline_handle_error): Likewise.
* reginfo.c (globalize_reg): Likewise.
* substring-locations.c (format_warning_n_va): Likewise.
* tree-inline.c (expand_call_inline): Likewise.
* tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
* tree-ssa-loop-niter.c
(do_warn_aggressive_loop_optimizations): Likewise.
* tree-ssa-uninit.c (warn_uninit): Likewise.
* tree.c (warn_deprecated_use): Likewise.

gcc/testsuite/ChangeLog:
PR other/84889
* gcc.dg/plugin/diagnostic-group-test-1.c: New test.
* gcc.dg/plugin/diagnostic_group_plugin.c: New test.
* gcc.dg/plugin/plugin.exp (plugin_test_list): Add the new tests.

From-SVN: r263675

x86: Always update EH return address in word_mode

On x86, return address is always popped in word_mode.  eh_return needs
to put EH return address in word_mode on stack.

gcc/

PR target/87014
* config/i386/i386.md (eh_return): Always update EH return
address in word_mode.

gcc/testsuite/

PR target/87014
* g++.dg/torture/pr87014.C: New file.

From-SVN: r263672

targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when TARGET_SPLIT_COMPLEX_ARG is defined.

* targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
TARGET_SPLIT_COMPLEX_ARG is defined.

From-SVN: r263671

[CPP PATCH] node field bits

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01171.html
* include/cpplib.h: Fixup some whitespace.
(cpp_hashnode): Reduce type to 2 bit & flags to 8.

From-SVN: r263669

[CPP PATCH] node type

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01164.html
* include/cpplib.h (NODE_BUILTIN, NODE_MACRO_ARG): Delete.
Renumber others.
(enum node_type): Replace NT_MACRO with NT_USER_MACRO,
NT_BUILTIN_MACRO, NT_MACRO_ARG.  Delete NT_ASSERTION.
(NTV_MACRO, NTV_ANSWER, NTV_BUILTIN, NTV_ARGUMENT, NTV_NONE):
Delete.
(CPP_HASHNODE_VALUE_IDX): Delete.
(union _cpp_hashnode_value): GTY tag from enum node_type directly.
(struct cpp_hashnode): Adjust GTY desc for value field.
(cpp_user_macro_p, cpp_builtin_macro_p, cpp_macro_p): Adjust.
* directives.c (undefine_macros): Clear value.anwers, adjust flag
clearing.
(_cpp_test_assertion): No need to check NT_ASSERTION.
(do_assert, do_unassert): Likewise.
* init.c (cpp_init_special_builtins): Set type not flags.
* macro.c (struct macro_arg_saved_data): Add type field.
(cpp_get_token_1): Check type not NT_VOID.
(_cpp_free_definition): Adjust flag clearing.  Nullify
value.answers.
(_cpp_save_parameter, _cpp_unsave_parameters): Save and restore
type.
(lex_expansion_token): Check type not flags.
(_cpp_create_definition): Set type to NT_USER_MACRO.
(_cpp_notify_macro_use): Adjust type checking.
* pch.c (write_macdef, count_defs, write_defs, cpp_valid_state)
(save_macros): Adjust node type/flag handling.
* traditional.c (_cpp_scan_out_logical_line): Check type not flags.

From-SVN: r263667

[CPP PATCH] Fix warning & other cleanups.

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01162.html
* directives.c (do_undef): Use cpp_macro_p & cpp_builtin_macro_p.
* include/cpplib.h (enum cpp_macro_kind): Remove trailing comma.
(cpp_fun_like_macro_p): Make inline, define.
* macro.c (cpp_define_lazily): Use UCHAR_MAX.
(cpp_fun_like_macro_p): Delete.

From-SVN: r263666

expr.c (store_field): Change gcc_assert to gcc_checking_assert.

2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>

* expr.c (store_field): Change gcc_assert to gcc_checking_assert.

From-SVN: r263665

re PR target/86984 (invalid relocation accessing a const char array)

PR target/86984
* expr.c (expand_assignment): Assert that bitpos is positive.
(store_field): Likewise
(expand_expr_real_1): Make sure that bitpos is positive.
* config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
integer overflow.

PR target/86984
* gcc.target/alpha/pr86984.c: New test.

From-SVN: r263664

[PATCH] Kill cpp-id-data.h

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01149.html
libcpp/
* Makefile.in (TAGS_SOURCES): Remove cpp-id-data.h.
* include/cpp-id-data.h: Delete.
* internal.h: Include cpplib.h not cpp-id-data.h.
gcc/
* Makefile.in (CPP_ID_DATA_H): Delete.
(CPP_INTERNAL_H): Don't add it.
(GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
* gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h

From-SVN: r263663

re PR c++/78655 (gcc doesn't exploit the fact that the result of pointer addition can not be nullptr)

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

PR libstdc++/86963 Remove use of __tuple_base in std::tuple

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

[debug] Fix handling of vlas in lto

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

[PATCH] #assert becomes macro-like

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01140.html
libcpp/
* include/cpp-id-data.h (struct answer): Delete.
* include/cpplib.h (struct answer): Don't forward-declare.
(enum cpp_macro_kind): Add cmk_assert.
(struct cpp_macro): Union parms and next assert chain.
(union _cpp_hashnode_value): 'answer' field is cpp_macro.
* directives.c (parse_answer): Convert to use cpp_macro. Return
true on success.
(parse_assertion, find_answer, _cpp_test_assertion, cpp_do_assert)
(cpp_do_unassert): Convert to use cpp_macro.
* macro.c (warn_of_redefinition, _cpp_new_macro)
(check_trad_stringification, cpp_macro_definition): Adjust macro
parm access.
* traditional.c (_cpp_replacement_text_len)
(_cpp_copy_replacement_text, _cpp_create_trad_definition): Likewise.
gcc/c-family/
* c-ada-spec.c (macro_length, dump_ada_macros): Adjust macro parm
access.

From-SVN: r263658

Refactor std::optional SFINAE constraints

* 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

Daily bump.

From-SVN: r263656

re PR target/86994 (64-bit gcc.target/i386/20040112-1.c FAILs)

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

Daily bump.

From-SVN: r263651

86658.cc: Use dg-options to define _GLIBCXX_DEBUG.

2018-08-18  François Dumont  <fdumont@gcc.gnu.org>

* testsuite/25_algorithms/copy/86658.cc: Use dg-options to define
_GLIBCXX_DEBUG.

From-SVN: r263647

Darwin emits pubnames/types by default which mask the intent of the test.

gcc/testsuite

2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.dg/debug/dwarf2/pr80263.c: Suppress pubtypes output
for Darwin.

From-SVN: r263646

Darwin does not support gsplit-dwarf.

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

Daily bump.

From-SVN: r263644

[PATCH] Adjust lazy macro definition

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01072.html
libcpp/
* include/cpplib.h (struct cpp_callbacks): Replace
user_builtin_macro with user_lazy_macro.
(struct cpp_macro): add lazy field.
(enum cpp_builtin_type): Remove BT_FIRST_USER, BT_LAST_USER.
(cpp_define_lazily): Declare.
* macro.c (enter_macro_context) Use _cpp_maybe_notify_macro_use.
(warn_of_redefinition): Use cpp_builtin_macro_p, directly call
user_lazy_macro hook.
(_cpp_new_macro): Clear lazy field.
(cpp_define_lazily): Define.
(_cpp_notify_macro_use): Adjust lazy definition code.
(cpp_macro_definition): No need to do lazy definition here.
* pch.c (write_macdef, save_macros): Likewise.
gcc/c-family/
* c-cppbuiltin.c (struct lazy_hex_fp_value_struct): Remove macro
field.
(laxy_hex_fp_value_count): Make unsigned.
(lazy_hex_fp_value): Provided with macro & lazy number.  Directly
manipulate the macro.
(builtin_defin_with_hex_fp_value): Adjust callback name, use
cpp_define_lazily.

From-SVN: r263640

re PR testsuite/86996 (gcc.dg/tree-ssa/builtin-sprintf-warn-1.c FAILs)

PR testsuite/86996

gcc/testsuite/CHangeLog:
* gcc.dg/tree-ssa/builtin-sprintf-warn-1.c: Adjust.

From-SVN: r263636

C++: -Wwrite-strings: use location of string constant

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

doc: Use @ref, not @xref, in the middle of a sentence

makeinfo warns about this.

* doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
a sentence.

From-SVN: r263633

2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>

MAINTAINERS: Add c-sky port maintainers.

From-SVN: r263632

C-SKY port: libgcc

2018-08-17  Jojo  <jijie_rong@c-sky.com>
    Huibin Wang  <huibin_wang@c-sky.com>
    Sandra Loosemore  <sandra@codesourcery.com>
    Chung-Lin Tang  <cltang@codesourcery.com>

C-SKY port: libgcc

libgcc/
* config.host: Add C-SKY support.
* config/csky/*: New.

Co-Authored-By: Chung-Lin Tang <cltang@codesourcery.com>
Co-Authored-By: Huibin Wang <huibin_wang@c-sky.com>
Co-Authored-By: Sandra Loosemore <sandra@codesourcery.com>
From-SVN: r263631

C-SKY port: Testsuite

2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
    Chung-Lin Tang  <cltang@codesourcery.com>
    Xianmiao Qu  <xianmiao_qu@c-sky.com>

C-SKY port: Testsuite

gcc/testsuite/
* g++.dg/Wno-frame-address.C: Adjust for C-SKY.
* g++.dg/torture/type-generic-1.C: Likewise.
* gcc.c-torture/compile/20000804-1.c: Likewise.
* gcc.c-torture/execute/20101011-1.c: Likewise.
* gcc.c-torture/execute/ieee/mul-subnormal-single-1.x: Likewise.
* gcc.dg/20020312-2.c: Likewise.
* gcc.dg/Wno-frame-address.c: Likewise.
* gcc.dg/c11-true_min-1.c: Likewise.
* gcc.dg/sibcall-10.c: Likewise.
* gcc.dg/sibcall-9.c: Likewise.
* gcc.dg/stack-usage-1.c: Likewise.
* gcc.dg/torture/float32-tg-3.c: Likewise.
* gcc.dg/torture/float32x-tg-3.c: Likewise.
* gcc.dg/torture/float64-tg-3.c: Likewise.
* gcc.dg/torture/float64x-tg-3.c: Likewise.
* gcc.dg/torture/type-generic-1.c: Likewise.
* gcc.target/csky/*: New.
* lib/target-supports.exp (check_profiling_available): Add
csky-*-elf.
(check_effective_target_hard_float): Handle C-SKY targets with
single-precision hard float only.
(check_effective_target_logical_op_short_circuit): Handle C-SKY.

Co-Authored-By: Chung-Lin Tang <cltang@codesourcery.com>
Co-Authored-By: Xianmiao Qu <xianmiao_qu@c-sky.com>
From-SVN: r263630

C-SKY port: Documentation

2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>

C-SKY port: Documentation

gcc/
* doc/extend.texi (C-SKY Function Attributes): New section.
* doc/invoke.texi (Option Summary): Add C-SKY options.
(C-SKY Options): New section.
* doc/md.texi (Machine Constraints): Document C-SKY constraints.

From-SVN: r263629

C-SKY port: Backend implementation

2018-08-17  Jojo  <jijie_rong@c-sky.com>
    Huibin Wang  <huibin_wang@c-sky.com>
    Sandra Loosemore  <sandra@codesourcery.com>
    Chung-Lin Tang  <cltang@codesourcery.com>

C-SKY port: Backend implementation

gcc/
* config/csky/*: New.
* common/config/csky/*: New.

Co-Authored-By: Chung-Lin Tang <cltang@codesourcery.com>
Co-Authored-By: Huibin Wang <huibin_wang@c-sky.com>
Co-Authored-By: Sandra Loosemore <sandra@codesourcery.com>
From-SVN: r263628

C-SKY port: Configury

2018-08-17  Jojo  <jijie_rong@c-sky.com>
    Huibin Wang  <huibin_wang@c-sky.com>
    Sandra Loosemore  <sandra@codesourcery.com>
    Chung-Lin Tang  <cltang@codesourcery.com>
    Andrew Jenner  <andrew@codesourcery.com>

C-SKY port: Configury

gcc/
* config.gcc (csky-*-*): New.
* configure.ac: Add csky to targets for dwarf2 debug_line support.
* configure: Regenerated.

contrib/
* config-list.mk (LIST): Add csky-elf and csky-linux-gnu.

Co-Authored-By: Andrew Jenner <andrew@codesourcery.com>
Co-Authored-By: Chung-Lin Tang <cltang@codesourcery.com>
Co-Authored-By: Huibin Wang <huibin_wang@c-sky.com>
Co-Authored-By: Sandra Loosemore <sandra@codesourcery.com>
From-SVN: r263627

Formatted printing for dump_* in the middle-end

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 Implement LWG 2729 constraints on tuple assignment

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

Fix poly types after PR tree-optimization/71625 strlen optimization

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

rs6000: Give names to cbranch/creturn patterns

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

[PATCH] Macro body is trailing array

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01037.html
* include/cpplib.h (enum cpp_macro_kind): New.
(struct cpp_macro): Make body trailing array.  Add kind field,
delete traditional flag.
* internal.h (_cpp_new_macro): Declare.
(_cpp_reserve_room): New inline.
(_cpp_commit_buf): Declare.
(_cpp_create_trad_definition): Return new macro.
* lex.c (_cpp_commit_buff): New.
* macro.c (macro_real_token_count): Count backwards.
(replace_args): Pointer equality not orderedness.
(_cpp_save_parameter): Use _cpp_reserve_room.
(alloc_expansion_token): Delete.
(lex_expansion_token): Return macro pointer.  Use _cpp_reserve_room.
(create_iso_definition): Allocate macro itself.  Adjust for
different allocation ordering.
(_cpp_new_macro): New.
(_cpp_create_definition): Adjust for API changes.
* traditional.c (push_replacement_text): Don't set traditional
flag.
(save_replacement_text): Likewise.
(_cpp_create_trad_definition): Allocate macro itself, Adjust for
different allocation ordering.

From-SVN: r263622

[PATCH] Move cpp_macro to cpplib.h

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg01016.html
libcpp/
* cpp-id-data.h (uchar, UC): Move to internal.h
(struct cpp_macro): Move to cpplib.h.
* internal.h (uchar, UC): From cpp-id-data.h.
* include/cpplib.h (struct cpp_macro): From cpp-id-data.h.
gcc/c-family/
* c-ada-spec.c: Don't #include "cpp-id-data.h"
* c-cppbuiltin.c: Likewise.
gcc/
* cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".

From-SVN: r263618

re PR tree-optimization/86841 (ICE in gcc/gcc/tree-vrp.c:1325 with graphite)

2018-08-17  Richard Biener  <rguenther@suse.de>

PR tree-optimization/86841
* wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.

From-SVN: r263615

Merge Ignore and Deprecated in .opt files.

2018-08-17  Martin Liska  <mliska@suse.cz>

* common.opt: Remove Warn, Init and Report for options with
        Ignore/Deprecated flag. Warning is done automatically for
        Deprecated flags.
* config/i386/i386.opt: Likewise.
* config/ia64/ia64.opt: Likewise.
* config/rs6000/rs6000.opt: Likewise.
* cppbuiltin.c (define_builtin_macros_for_compilation_flags):
        Remove usage of flag_check_pointer_bounds.
* lto-wrapper.c (merge_and_complain): Do not handle
        OPT_fcheck_pointer_bounds.
(append_compiler_options): Likewise.
* opt-functions.awk: Do not handle Deprecated.
* optc-gen.awk: Check that Var, Report and Init are not
        used for an option with Ignore/Deprecated flag.
* opts-common.c (decode_cmdline_option): Do not report
        CL_ERR_DEPRECATED.
(read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
        options.
* opts.h (struct cl_option): Remove cl_deprecated flag.
(CL_ERR_DEPRECATED): Remove error enum value.
2018-08-17  Martin Liska  <mliska@suse.cz>

* g++.dg/opt/mpx.C: Fix scanned pattern.
* gcc.target/i386/mpx.c: Likewise.
* g++.dg/warn/Wunreachable-code-1.C: Remove.
* g++.dg/warn/Wunreachable-code-2.C: Likewise.
* gcc.dg/torture/pr52969.c: Likewise.
* g++.dg/warn/pr31246-2.C: Likewise.
* g++.dg/warn/pr31246.C: Likewise.
* gcc.dg/pr33092.c: Likewise.
* g++.dg/opt/eh1.C: Remove a deprecated option.
* g++.dg/template/inline1.C: Likewise.
* g++.dg/tree-ssa/pr81408.C: Likewise.
* gcc.dg/pr41837.c: Likewise.
* gcc.dg/pr41841.c: Likewise.
* gcc.dg/pr42250.c: Likewise.
* gcc.dg/pr43084.c: Likewise.
* gcc.dg/pr43317.c: Likewise.
* gcc.dg/pr51879-18.c: Likewise.
* gcc.dg/torture/pr36066.c: Likewise.
* gcc.dg/tree-ssa/ifc-8.c: Likewise.
* gcc.dg/tree-ssa/ifc-cd.c: Likewise.
* gcc.dg/tree-ssa/pr19210-1.c: Likewise.
* gcc.dg/tree-ssa/pr45122.c: Likewise.
* gcc.target/i386/pr45352-2.c: Likewise.
* gcc.target/i386/zee.c: Likewise.
* gfortran.dg/auto_char_len_2.f90: Likewise.
* gfortran.dg/auto_char_len_4.f90: Likewise.
* gfortran.dg/c_ptr_tests_15.f90: Likewise.
* gfortran.dg/char_array_structure_constructor.f90: Likewise.
* gfortran.dg/gomp/pr47331.f90: Likewise.
* gfortran.dg/pr40999.f: Likewise.
* gfortran.dg/pr41011.f: Likewise.
* gfortran.dg/pr42051.f03: Likewise.
* gfortran.dg/pr46804.f90: Likewise.
* gfortran.dg/pr83149_1.f90: Likewise.
* gfortran.dg/pr83149_b.f90: Likewise.
* gfortran.dg/whole_file_1.f90: Likewise.
* gfortran.dg/whole_file_10.f90: Likewise.
* gfortran.dg/whole_file_11.f90: Likewise.
* gfortran.dg/whole_file_12.f90: Likewise.
* gfortran.dg/whole_file_13.f90: Likewise.
* gfortran.dg/whole_file_14.f90: Likewise.
* gfortran.dg/whole_file_15.f90: Likewise.
* gfortran.dg/whole_file_16.f90: Likewise.
* gfortran.dg/whole_file_17.f90: Likewise.
* gfortran.dg/whole_file_18.f90: Likewise.
* gfortran.dg/whole_file_19.f90: Likewise.
* gfortran.dg/whole_file_2.f90: Likewise.
* gfortran.dg/whole_file_20.f03: Likewise.
* gfortran.dg/whole_file_3.f90: Likewise.
* gfortran.dg/whole_file_4.f90: Likewise.
* gfortran.dg/whole_file_5.f90: Likewise.
* gfortran.dg/whole_file_6.f90: Likewise.
* gfortran.dg/whole_file_7.f90: Likewise.
* gfortran.dg/whole_file_8.f90: Likewise.
* gfortran.dg/whole_file_9.f90: Likewise.
* gcc.dg/vect/vect.exp: Likewise.
2018-08-17  Martin Liska  <mliska@suse.cz>

* c.opt: Remove Warn, Init and Report for options with
        Ignore/Deprecated flag. Warning is done automatically for
        Deprecated flags.

From-SVN: r263614

re PR middle-end/86505 (__builtin_va_arg_pack_len() computes the number of arguments wrongly)

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

re PR tree-optimization/86853 (sprintf optimization for wide strings doesn't account for conversion failure)

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

Daily bump.

From-SVN: r263611

builtins.c (c_strlen): Add new parameter eltsize.

* builtins.c (c_strlen): Add new parameter eltsize.  Use it
for determining how to count the elements.
* builtins.h (c_strlen): Adjust prototype.
* expr.c (string_constant): Add new parameter mem_size.
Set *mem_size appropriately.
* expr.h (string_constant): Adjust protoype.
* gimple-fold.c (get_range_strlen): Add new parameter eltsize.
* gimple-fold.h (get_range_strlen): Adjust prototype.
* gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
(format_string): Call get_string_length with eltsize.

From-SVN: r263607

diagnostics: fix bad interaction between line spans and line numbers

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

diagnostics: tweak to line-insertion fix-it hints with line-numbering

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

rs6000: Delete old add+cmp patterns

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

rs6000: Remove "length 4" from other insns

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

rs6000: Remove "length 4" from branch insns

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

rs6000: Change the length attribute default

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

[PATCH] Macro definition parameter parsing

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg00977.html
libcpp/
* internal.h (_cpp_save_parameter): Take parmno, not macro.
(_cpp_unsave_parameters): Declare.
* macro.c (_cpp_save_parameter): Take parm number, not macro.
Return true on success.
(_cpp_unsave_parameters): New.
(parse_params): Take parm_no and variadic pointers, not macro.
Reimplement parsing logic.
(create_iso_definition): Adjust parse_params changes.  Call
_cpp_unsave_parameters here.
(_cpp_create_definition): Don't unsave params here.
* traditional.c (scan_parameters): Take n_param pointer, adjust.
(_cpp_create_trad_definition): Ajust scan_parameters change.  Call
_cpp_unsave_parameters.
gcc/testsuite/
* gcc.dg/cpp/macsyntx.c: Adjust expected errors.
* gcc.dg/cpp/macsyntx2.c: likewise.

From-SVN: r263600

re PR testsuite/86745 (gcc.target/i386/avx-cvt-2.c etc. FAIL on 64-bit x86)

PR testsuite/86745
* gcc.target/i386/avx-cvt-2.c: Loosen scan-assembler strings.
* gcc.target/i386/avx-cvt-2.c: Ditto.

From-SVN: r263598

Fix warning with -Wsign-compare -Wsystem-headers

* include/tr1/legendre_function.tcc (__sph_legendre): Avoid warning
about signed/unsigned comparison.

From-SVN: r263596

-Wmisleading-indentation: fix ICE in get_visual_column (PR c++/70693)

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

Allow larger copies when not slow_unaligned_access and no padding.

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

Avoid deprecation warning with -Wsystem-headers

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

Remove rtl.texi references to old RTX code class names

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.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

Fix bootstrap with --enable-fully-dynamic-string

PR libstdc++/86447
* src/c++11/cow-stdexcept.cc [_GLIBCXX_FULLY_DYNAMIC_STRING]
(logic_error::logic_error(logic_error&&))
(logic_error::operator=(logic_error&&))
(runtime_error::runtime_error(runtime_error&&))
(runtime_error::operator=(runtime_error&&)): Copy strings instead of
moving, to avoid allocating empty reps for moved-from strings.

From-SVN: r263590

[PATCH] CPP Macro predicates

https://gcc.gnu.org/ml/gcc-patches/2018-08/msg00897.html
libcpp/
* include/cpplib.h (cpp_user_macro_p, cpp_builtin_macro_p)
(cpp_macro_p): New inlines.
* directives.c (do_pragma_poison): Use cpp_macro_p.
(do_ifdef, do_ifndef): Likewise.  Use _cpp_maybe_notify_macro_use.
(cpp_pop_definition): Use cpp_macro_p.  Move _cpp_free_definition
earlier.  Don't zap node directly.
* expr.c (parse_defined): Use _cpp_maybe_notify_macro_use &
cpp_macro_p.
* files.c (should_stack_file): Use cpp_macro_p.
* identifiers.c (cpp_defined): Likewise.
* internal.h (_cpp_mark_macro): Use cpp_user_macro_p.
(_cpp_notify_macro_use): Declare.
(_cpp_maybe_notify_macro_use): New inline.
* lex.c (is_macro): Use cpp_macro_p.
* macro.c (_cpp_warn_if_unused_macro): Use cpp_user_macro_p.
(enter_macro_context): Likewise.
(_cpp_create_definition): Use cpp_builtin_macro_p,
cpp_user_macro_p.  Move _cpp_free_definition earlier.
(_cpp_notify_macro_use): New, broken out of multiple call sites.
* traditional.c (fun_like_macro_p): Use cpp_builtin_macro_p.
(maybe_start_funlike, _cpp_scan_out_logical_line)
(push_replacement_text): Likewise.
gcc/c-family/
* c-ada-spec.c (count_ada_macro): Use cpp_user_macro_p.
(store_ada_macro): Likewise.
* c-ppoutput.c (cb_used_define, dump_macro): Likewise.
* c-spellcheck.cc (should-suggest_as_macro_p): Likewise,
gcc/
* config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
* config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
gcc/cp/
* name-lookup.c (lookup_name_fuzzy): Likewise.
gcc/fortran/
* cpp.c (dump_macro): Use cpp_user_macro_p.

From-SVN: r263587

Add __USER_LABEL_PREFIX__ to the asm names for targets that need it.

2018-08-16  Iain Sandoe <iain@sandoe.co.uk>

gcc/testsuite

        * gcc.dg/memcmp-1.c (lib_memcmp): Apply __USER_LABEL_PREFIX__.
        (lib_strncmp): Likewise.

From-SVN: r263586

Darwin doesn't have pthread_barrier, at least up to Darwin18
so don't try to do tests with it.

2018-08-16  Iain Sandoe <iain@sandoe.co.uk>

gcc/testsuite/
* c-c++-common/asan/pointer-subtract-3.c: Skip for Darwin.
* c-c++-common/asan/pointer-subtract-4.c: Likewise.
* g++.dg/torture/pr44295.C : Likewise.

From-SVN: r263585

Update fall through pattern for FP16 patterns in ARM.

The original issue comes from the fact that the code does

... foo (... bar)
{
  return bar;
}

The expansion of the return statement causes GCC to try to return the value in
a register.  GCC will try to emit the move then, from MEM to REG (due to the SSA
temporary.).  It checks for a mov optab for this which isn't available and
then tries to do the move in bits using emit_move_multi_word.

emit_move_multi_word will split the move into sub parts, but then needs to get
the sub parts and does this using subregs, but it's told it can't do subregs!

The compiler is now stuck in an infinite loop.

The way this is worked around in the back-end is that we have move patterns in
neon.md that usually just force the register instead of checking with the
back-end. This prevents emit_move_multi_word from being needed.  However the
pattern for V4HF and V8HF were guarded by TARGET_NEON && TARGET_FP16.

I don't believe the TARGET_FP16 guard to be needed, because the pattern doesn't
actually generate code and requires another pattern for that, and a reg to reg move
should always be possible anyway. So allowing the force to register here is safe
and it allows the compiler to generate a correct error instead of ICEing in an
infinite loop.

gcc/
2018-08-16  Tamar Christina  <tamar.christina@arm.com>

PR target/84711
* config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
* config/arm/neon.md (movv4hf, movv8hf): Refactored to..
(mov<mov>): ..this and enable unconditionally.

From-SVN: r263584