C++17 added new overloads to <fstream> class templates to support
opening files from wide character strings "on systems where
filesystem::path::value_type is not char". This patch adds those
overloads conditional on _wfopen being available, and enables them for
pre-C++17 modes as well.

	Add support for opening file streams from wide character strings.
	* config/io/basic_file_stdio.cc [_GLIBCXX_HAVE__WFOPEN]
	(__basic_file<char>::open(const wchar_t*, ios_base::openmode)):
	Define new overload.
	* config/io/basic_file_stdio.h [_GLIBCXX_HAVE__WFOPEN]
	(__basic_file<char>::open(const wchar_t*, ios_base::openmode)):
	Declare new overload.
	* configure.ac: Check for _wfopen.
	* crossconfig.m4: Likewise.
	* configure: Regenerate.
	* config.h.in: Regenerate.
	* include/bits/fstream.tcc [_GLIBCXX_HAVE__WFOPEN]
	(basic_filebuf<C,T>::open(const wchar_t*, ios_base::openmode)):
	Define new overload.
	* include/std/fstream [_GLIBCXX_HAVE__WFOPEN]
	(basic_filebuf<C,T>::open(const wchar_t*, ios_base::openmode)):
	Declare new overload.
	[_GLIBCXX_HAVE__WFOPEN]
	(basic_ifstream<C,T>::basic_ifstream(const wchar_t*, openmode))
	(basic_ifstream<C,T>::basic_open(const wchar_t*, openmode))
	(basic_ofstream<C,T>::basic_ifstream(const wchar_t*, openmode))
	(basic_ofstream<C,T>::basic_open(const wchar_t*, openmode))
	(basic_fstream<C,T>::basic_ifstream(const wchar_t*, openmode))
	(basic_fstream<C,T>::basic_open(const wchar_t*, openmode)): Define
	new overloads.
	* testsuite/27_io/basic_filebuf/open/wchar_t/1.cc: New.
	* testsuite/27_io/basic_ifstream/cons/wchar_t/1.cc: New.
	* testsuite/27_io/basic_ifstream/open/wchar_t/1.cc: New.
	* testsuite/27_io/basic_ofstream/cons/wchar_t/1.cc: New.
	* testsuite/27_io/basic_ofstream/open/wchar_t/1.cc: New.
	* testsuite/27_io/basic_fstream/cons/wchar_t/1.cc: New.
	* testsuite/27_io/basic_fstream/open/wchar_t/1.cc: New.

From-SVN: r260479

2018-05-21  François Dumont  <fdumont@gcc.gnu.org>

	PR libstdc++/85845
	* include/bits/stl_tree.h
	(_Rb_tree_impl(_Rb_tree_impl&&, _Node_allocator&&)): Fix noexcept
	qualification.

From-SVN: r260478

This recently-committed test fails the INS scan for tiny and large memory models.
That is because instead of the:
make_vector:
         adrp    x1, a
         adrp    x0, b
         movi    v0.4s, 0
         ldr     s2, [x1, #:lo12:a]
         ldr     s1, [x0, #:lo12:b]
         ins     v0.s[2], v2.s[0]
         ins     v0.s[3], v1.s[0]
         ret

That we generate for the default small model, we end up with a simple register
addressing mode with no addend/offset for the lane load:
make_vector:
         movi    v0.4s, 0
         adr     x1, a
         adr     x0, b
         ld1     {v0.s}[2], [x1]
         ld1     {v0.s}[3], [x0]
         ret

and

make_vector:
         movi    v0.4s, 0
         adrp    x0, .LC0
         ldr     x1, [x0, #:lo12:.LC0]
         adrp    x0, .LC1
         ldr     x0, [x0, #:lo12:.LC1]
         ld1     {v0.s}[2], [x1]
         ld1     {v0.s}[3], [x0]
         ret

So we end up merging the load and the lane insert.
This patch adjusts the testcase to scan for the right thing accordingly.
Checked that the testcase passes with -mcmodel=tiny, -mcmodel=small, -mcmodel=large.

	* gcc.target/aarch64/vec_init_1.c: Scan for LD1 instead of INS for
	tiny and large memory models.

From-SVN: r260474

From-SVN: r260472

The compiler warns on an object declaration with default initialization
and an address clause, to indicate that the overlay implied by the address
clause might affect a value elsewhere. The warning is suppressed if the type
carries the Suppress_Initialization aspect. With this patch the compiler
also inhibits the warning if the aspect is specified for the object itself.

2018-05-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* freeze.adb (Warn_Overlay): Do not emit a wawrning on an object
	declaration with an explicit address clause and a type with default
	initialization, if the declaration carries an aspect
	Suppress_Initialization.

gcc/testsuite/

	* gnat.dg/suppress_initialization.adb,
	gnat.dg/suppress_initialization_pkg.ads: New testcase.

From-SVN: r260471

2018-05-21  Daniel Mercier  <mercier@adacore.com>

gcc/ada/

	* pprint.adb: Use mixed case for attribute names.

From-SVN: r260470

This patch ensures that aspect specifications which appear on package,
protected, and task body stubs are properly analyzed.

------------
-- Source --
------------

--  pack.ads

package Pack
  with SPARK_Mode,
       Abstract_State => State
is
   -------------------------------------
   -- Refined_Depends, Refined_Global --
   -------------------------------------

   procedure Proc_1;
   procedure Proc_2
     with Global  => (In_Out => State),
          Depends => (State  => State);

   task Task_Obj_1;
   task Task_Obj_2
     with Global  => (In_Out => State),
          Depends => (State  => State);

   ------------------
   -- Refined_Post --
   ------------------

   function Func_1 (Formal : Integer) return Integer;
   function Func_2 (Formal : Integer) return Integer
     with Post => Func_2'Result > Formal;

   -------------------
   -- Refined_State --
   -------------------

   package Pack_1 is end Pack_1;
   package Pack_2 with Abstract_State => State_2 is end Pack_2;

   ----------------
   -- SPARK_Mode --
   ----------------

   package Pack_3 with SPARK_Mode => Off is end Pack_3;
   package Pack_4 with SPARK_Mode => Off is end Pack_4;
   package Pack_5 is end Pack_5;

   protected type Prot_Typ_1 with SPARK_Mode => Off is end Prot_Typ_1;
   protected type Prot_Typ_2 with SPARK_Mode => Off is end Prot_Typ_2;
   protected type Prot_Typ_3 is end Prot_Typ_3;

   procedure Proc_3 with SPARK_Mode => Off;
   procedure Proc_4 with SPARK_Mode => Off;
   procedure Proc_5;

   task type Task_Typ_1 with SPARK_Mode => Off;
   task type Task_Typ_2 with SPARK_Mode => Off;
   task type Task_Typ_3;
end Pack;

--  pack.adb

package body Pack
  with SPARK_Mode,
       Refined_State => (State => Constit)
is
   Constit : Integer := 0;

   -------------------------------------
   -- Refined_Depends, Refined_Global --
   -------------------------------------

   procedure Proc_1 is separate
     with Refined_Global  => (In_Out  => Constit),                   --  Error
          Refined_Depends => (Constit => Constit);                   --  Error

   procedure Proc_2 is separate
     with Refined_Global  => (In_Out  => Constit),                   --  OK
          Refined_Depends => (Constit => Constit);                   --  OK

   task body Task_Obj_1 is separate
     with Refined_Global  => (In_Out  => Constit),                   --  Error
          Refined_Depends => (Constit => Constit);                   --  Error

   task body Task_Obj_2 is separate
     with Refined_Global  => (In_Out  => Constit),                   --  OK
          Refined_Depends => (Constit => Constit);                   --  OK

   ------------------
   -- Refined_Post --
   ------------------

   function Func_1 (Formal : Integer) return Integer is separate
     with Refined_Post => Func_1'Result > Formal;                    --  OK

   function Func_2 (Formal : Integer) return Integer is separate
     with Refined_Post => Func_2'Result > Formal;                    --  OK

   -------------------
   -- Refined_State --
   -------------------

   package body Pack_1 is separate
     with Refined_State => (State_1 => Constit_1);                   --  Error

   package body Pack_2 is separate
     with Refined_State => (State_2 => Constit_2);                   --  Error

   ----------------
   -- SPARK_Mode --
   ----------------

   package body Pack_3 is separate with SPARK_Mode => On;            --  Error
   package body Pack_4 is separate;
   package body Pack_5 is separate with SPARK_Mode => Off;           --  Error

   protected body Prot_Typ_1 is separate with SPARK_Mode => On;      --  Error
   protected body Prot_Typ_2 is separate;
   protected body Prot_Typ_3 is separate with SPARK_Mode => Off;     --  Error

   procedure Proc_3 is separate with SPARK_Mode => On;               --  Error
   procedure Proc_4 is separate;
   procedure Proc_5 is separate with SPARK_Mode => Off;              --  Error

   task body Task_Typ_1 is separate with SPARK_Mode => On;           --  Error
   task body Task_Typ_2 is separate;
   task body Task_Typ_3 is separate with SPARK_Mode => Off;          --  Error
end Pack;

--  pack-func_1.adb

separate (Pack)

function Func_1 (Formal : Integer) return Integer
  with Refined_Post => Func_1'Result > Formal                        --  Error
is
begin
   return Formal * 10;
end Func_1;

--  pack-func_2.adb

separate (Pack)

function Func_2 (Formal : Integer) return Integer
  with Refined_Post => Func_2'Result > Formal                        --  Error
is
begin
   return Formal * 10;
end Func_2;

--  pack-pack_1.adb

separate (Pack)

package body Pack_1
  with SPARK_Mode,
       Refined_State => (State_1 => Constit_1)                       --  Error
is
   Constit_1 : Integer := 1;
end Pack_1;

--  pack-pack_2.adb

separate (Pack)

package body Pack_2
  with SPARK_Mode,
       Refined_State => (State_2 => Constit_2)                       --  OK
is
   Constit_2 : Integer := 2;
end Pack_2;

--  pack-pack_3.adb

separate (Pack)

package body Pack_3 is end Pack_3;

--  pack-pack_4.adb

separate (Pack)

package body Pack_4 with SPARK_Mode => On is end Pack_4;             --  OK

--  pack-pack_5.adb

separate (Pack)

package body Pack_5 with SPARK_Mode => On is end Pack_5;             --  OK

--  pack-proc_1.adb

separate (Pack)

procedure Proc_1
  with Refined_Global  => (In_Out  => Constit),                      --  Error
       Refined_Depends => (Constit => Constit)                       --  Error
is begin null; end Proc_1;

--  pack-proc_2.adb

separate (Pack)

procedure Proc_2
  with Refined_Global  => (In_Out  => Constit),                      --  Error
       Refined_Depends => (Constit => Constit)                       --  Error
is begin null; end Proc_2;

--  pack-proc_3.adb

separate (Pack)

procedure Proc_3 is begin null; end Proc_3;

--  pack-proc_4.adb

separate (Pack)

procedure Proc_4 with SPARK_Mode => On is begin null; end Proc_4;    --  OK

--  pack-proc_5.adb

separate (Pack)

procedure Proc_5 with SPARK_Mode => On is begin null; end Proc_5;    --  OK

--  pack-prot_typ_1.adb

separate (Pack)

protected body Prot_Typ_1 is end Prot_Typ_1;

--  pack-prot_typ_2.adb

separate (Pack)

protected body Prot_Typ_2 with SPARK_Mode => On is end Prot_Typ_2;   --  OK

--  pack-prot_typ_3.adb

separate (Pack)

protected body Prot_Typ_3 with SPARK_Mode => On is end Prot_Typ_3;   --  OK

--  pack-task_obj_1.adb

separate (Pack)

task body Task_Obj_1
  with Refined_Global  => (In_Out  => Constit),                      --  Error
       Refined_Depends => (Constit => Constit)                       --  Error
is begin null; end Task_Obj_1;

--  pack-task_obj_2.adb

separate (Pack)

task body Task_Obj_2
  with Refined_Global  => (In_Out  => Constit),                      --  Error
       Refined_Depends => (Constit => Constit)                       --  Error
is begin null; end Task_Obj_2;

--  pack-task_typ_1.adb

separate (Pack)

task body Task_Typ_1 is begin null; end Task_Typ_1;

--  pack-task_typ_2.adb

separate (Pack)

task body Task_Typ_2 with SPARK_Mode => On is                        --  OK
begin null; end Task_Typ_2;

--  pack-task_typ_3.adb

separate (Pack)

task body Task_Typ_3 with SPARK_Mode => On is                        --  OK
begin null; end Task_Typ_3;

----------------------------
-- Compilation and output --
----------------------------

$ gcc -c pack.adb
pack.adb:12:11: useless refinement, declaration of subprogram "Proc_1" lacks
  aspect or pragma Global
pack.adb:13:11: useless refinement, declaration of subprogram "Proc_1" lacks
  aspect or pragma Depends
pack.adb:20:11: useless refinement, declaration of task type "Task_Obj_1" lacks
  aspect or pragma Global
pack.adb:21:11: useless refinement, declaration of task type "Task_Obj_1" lacks
  aspect or pragma Depends
pack.adb:42:11: aspect "Refined_State" must apply to a package body
pack.adb:45:11: aspect "Refined_State" must apply to a package body
pack.adb:51:41: incorrect placement of aspect "Spark_Mode"
pack.adb:53:41: incorrect placement of aspect "Spark_Mode"
pack.adb:55:47: incorrect placement of aspect "Spark_Mode"
pack.adb:57:47: incorrect placement of aspect "Spark_Mode"
pack.adb:59:38: incorrect placement of aspect "Spark_Mode"
pack.adb:61:38: incorrect placement of aspect "Spark_Mode"
pack.adb:63:42: incorrect placement of aspect "Spark_Mode"
pack.adb:65:42: incorrect placement of aspect "Spark_Mode"
pack-proc_1.adb:4:08: aspect "Refined_Global" cannot apply to a subunit
pack-proc_1.adb:5:08: aspect "Refined_Depends" cannot apply to a subunit
pack-proc_2.adb:4:08: aspect "Refined_Global" cannot apply to a subunit
pack-proc_2.adb:5:08: aspect "Refined_Depends" cannot apply to a subunit
pack-task_obj_1.adb:4:08: aspect "Refined_Global" cannot apply to a subunit
pack-task_obj_1.adb:5:08: aspect "Refined_Depends" cannot apply to a subunit
pack-task_obj_2.adb:4:08: aspect "Refined_Global" cannot apply to a subunit
pack-task_obj_2.adb:5:08: aspect "Refined_Depends" cannot apply to a subunit
pack-func_1.adb:4:08: aspect "Refined_Post" cannot apply to a subunit
pack-func_2.adb:4:08: aspect "Refined_Post" cannot apply to a subunit
pack-pack_1.adb:3:14: body of package "Pack_1" has unused hidden states
pack-pack_1.adb:3:14: variable "Constit_1" defined at line 7
pack-pack_1.adb:5:08: useless refinement, package "Pack_1" does not define
  abstract states
pack-pack_1.adb:5:26: "State_1" is undefined
pack-pack_3.adb:3:01: incorrect use of SPARK_Mode at pack.adb:2
pack-pack_3.adb:3:01: value Off was set for SPARK_Mode on "Pack_3" at
  pack.ads:38
pack-pack_4.adb:3:01: incorrect use of SPARK_Mode at pack.adb:2
pack-pack_4.adb:3:01: value Off was set for SPARK_Mode on "Pack_4" at
  pack.ads:39
pack-pack_4.adb:3:26: incorrect use of SPARK_Mode
pack-pack_4.adb:3:26: value Off was set for SPARK_Mode on "Pack_4" at
  pack.ads:39
pack-prot_typ_2.adb:3:32: incorrect use of SPARK_Mode
pack-prot_typ_2.adb:3:32: value Off was set for SPARK_Mode on "Prot_Typ_2" at
  pack.ads:43
pack-proc_3.adb:3:01: incorrect use of SPARK_Mode at pack.adb:2
pack-proc_3.adb:3:01: value Off was set for SPARK_Mode on "Proc_3" at
  pack.ads:46
pack-proc_4.adb:3:01: incorrect use of SPARK_Mode at pack.adb:2
pack-proc_4.adb:3:01: value Off was set for SPARK_Mode on "Proc_4" at
  pack.ads:47
pack-proc_4.adb:3:23: incorrect use of SPARK_Mode
pack-proc_4.adb:3:23: value Off was set for SPARK_Mode on "Proc_4" at
  pack.ads:47
pack-task_typ_2.adb:3:27: incorrect use of SPARK_Mode
pack-task_typ_2.adb:3:27: value Off was set for SPARK_Mode on "Task_Typ_2" at
  pack.ads:51

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* sem_ch6.adb (Analyze_Generic_Subprogram_Body): Rename the call to
	Analyze_Aspect_Specifications_On_Body_Or_Stub.
	(Analyze_Subprogram_Body_Helper): Rename the calls to
	Analyze_Aspect_Specifications_On_Body_Or_Stub.
	* sem_ch9.adb (Analyze_Entry_Body): Rename the call to
	Analyze_Aspect_Specifications_On_Body_Or_Stub.
	* sem_ch10.adb: Add with and use clause for Sem_Ch13.
	(Analyze_Package_Body_Stub): Add constant Id. Decorate the package stub
	prior to analyzing its aspects.
	(Analyze_Protected_Body_Stub): Add constant Id. Decorate the package
	stub prior to analyzing its aspects. Save and restore the configuration
	switches.
	(Analyze_Task_Body_Stub): Add constant Id. Decorate the package stub
	prior to analyzing its aspects.
	* sem_ch13.adb (Analyze_Aspect_Specifications_On_Body_Or_Stub): Renamed
	to Analyze_Aspects_On_Subprogram_Body_Or_Stub.
	* sem_ch13.ads (Analyze_Aspect_Specifications_On_Body_Or_Stub): Renamed
	to Analyze_Aspects_On_Subprogram_Body_Or_Stub.
	* sem_prag.adb: Code reformatting.
	(Analyze_Refined_Depends_Global_Post): Consider task body stubs.

From-SVN: r260469

This properly links with libsocket when needed by the user code.

2018-05-21  Jerome Lambourg  <lambourg@adacore.com>

gcc/ada/

	* gcc-interface/Makefile.in: Add g-soliop__qnx.ads to the runtime build
	for QNX.

From-SVN: r260468

This patch corrects the part of the access-before-elaboration mechanism which
ensures that the freeze node of a tagged type is within the early call region
of all its overriding bodies to ignore predefined primitives.

------------
-- Source --
------------

--  pack.ads

package Pack with SPARK_Mode is
   type Parent_Typ is tagged null record;
   procedure Prim (Obj : Parent_Typ);

   type Deriv_Typ is new Parent_Typ with private;
   overriding procedure Prim (Obj : Deriv_Typ);

private
   type Deriv_Typ is new Parent_Typ with null record;
end Pack;

-----------------
-- Compilation --
-----------------

$ gcc -c pack.ads

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* exp_cg.adb: Remove with and use clause for Exp_Disp.
	* exp_ch9.adb: Remove with and use clause for Exp_Disp.
	* exp_disp.adb (Is_Predefined_Dispatching_Operation): Moved to Sem_Util.
	(Is_Predefined_Interface_Primitive): Moved to Sem_Util.
	(Is_Predefined_Internal_Operation): Moved to Sem_Util.
	* exp_disp.ads (Is_Predefined_Dispatching_Operation): Moved to Sem_Util.
	(Is_Predefined_Interface_Primitive): Moved to Sem_Util.
	(Is_Predefined_Internal_Operation): Moved to Sem_Util.
	* exp_dist.adb: Remove with and use clause for Exp_Disp.
	* freeze.adb: Remove with and use clause for Exp_Disp.
	* sem_cat.adb: Remove with and use clause for Exp_Disp.
	* sem_ch6.adb: Remove with and use clause for Exp_Disp.
	* sem_ch12.adb: Remove with and use clause for Exp_Disp.
	* sem_elab.adb (Check_Overriding_Primitive): Do not process predefined
	primitives.
	* sem_util.adb: Remove with and use clause for Exp_Disp.
	(Is_Predefined_Dispatching_Operation): Moved from Exp_Disp.
	(Is_Predefined_Interface_Primitive): Moved from Exp_Disp.
	(Is_Predefined_Internal_Operation): Moved from Exp_Disp.
	* sem_util.ads (Is_Predefined_Dispatching_Operation): Moved from
	Exp_Disp.
	(Is_Predefined_Interface_Primitive): Moved from Exp_Disp.
	(Is_Predefined_Internal_Operation): Moved from Exp_Disp.

From-SVN: r260467

A type conversion may be illegal if the expression in the conversion has a
limited view of a type. This patch expands the error report to indicate the
presence of a limited view, and when the context is a package body it suggests
the addition of a regular with-clause to make the full view available.

Compiling client.adb must yield:

   client.adb:6:16: invalid conversion, not compatible with limited view
      of type "Map_Type" defined at maps.ads:2
   client.adb:6:16: add with_clause for "Maps" to current unit

----
package Maps is
  type Map_Type is null record;
end;
----
limited with Maps;
package Payloads is
  function Get_Map return access Maps.Map_Type;
end;
----
with Maps;
package Maps2 is
  type New_Map_Type is new Maps.Map_Type;
end;
----
with Maps2;
package Client is
  procedure Foo (Map : Maps2.New_Map_Type) is null;
  procedure Bar;
end;
----
with Payloads;
package body Client is
  procedure Bar is
  begin
     Foo (Maps2.New_Map_Type (Payloads.Get_Map.all));
  end;
end;

2018-05-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* sem_res.adb (Valid_Conversion): Improve error message on an illegal
	type conversion whose expression has a limited view of a type.

From-SVN: r260466

This patch fixes an omission in the expansion of loops over GNAT-specific
iterable objects. If the source includes an explicit name for the loop,
that name has to be preserved in the expanded code to allow exit statements
to mention it.

2018-05-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* exp_ch5.adb (Build_Formal_Container_Iteration): If source has
	explicit name for iterator loop, preserve that name in expanded
	construct, for possible use in exit statements.

gcc/testsuite/

	* gnat.dg/exit1.adb: New testcase.

From-SVN: r260465

2018-05-21  Javier Miranda  <miranda@adacore.com>

gcc/ada/

	* sem_ch4.adb (Analyze_Membership_Op): Avoid compiler crash when the
	spec of a unit has Ada 2012 membership tests with multiple choices and
	the unit body is not compiled under Ada 2012 mode.

From-SVN: r260464

2018-05-21  Doug Rupp  <rupp@adacore.com>

gcc/ada/

	* sigtramp-vxworks-target.inc: Set cfa_reg properly from sigcontext
	pregs.
	(CFI_COMMON_REGS): Restore LR jic probed from prologue.
	(REGNO_PC_OFFSET): Change to correct value for Aarch64.

From-SVN: r260463

2018-05-21  Jose Ruiz  <ruiz@adacore.com>

gcc/ada/

	* doc/gnat_ugn/gnat_utility_programs.rst, exp_attr.adb,
	libgnarl/s-tassta.adb: Minor typo fixes

From-SVN: r260462

This patch refines the handling of the well-known syntactic ambiguity created
by a function with defaulted parameters that returns an array, so that F (X)
may designate a call to the function, or an indexing of a parameterless call.
This patch handles the case where such a call is itself the prefix of another
call, and the function is a primitive operation invoked in prefix form.

2018-05-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* sem_ch4.adb (Analyze_One_Call): Recognize complex cases where an
	indexed call originally in prefix forn is itself the prefix of a
	further call.

gcc/testsuite/

	* gnat.dg/array30.adb: New testcase.

From-SVN: r260461

2018-05-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

	* sem_eval.adb (Is_Null_Range): Clarify access to the full view of a
	type.
	(Not_Null_Range): Same as above.

From-SVN: r260460

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* exp_ch3.adb: Minor reformatting.
	* exp_ch6.adb: Likewise.
	* freeze.adb: Likewise.
	* inline.adb: Likewise.
	* sem_util.adb: Likewise.

From-SVN: r260459

In the frontend inlining used in GNATprove, inlining of a return statement
was using an unchecked type conversion, which could cause a necessary
run-time check on the conversion to be skipped. Now fixed.

There is no impact on compilation.

2018-05-21  Yannick Moy  <moy@adacore.com>

gcc/ada/

	* inline.adb (Expand_Inlined_Call.Process_Formals): Use a type
	conversion instead of an unchecked type conversion when inlining a
	return statement, unless type qualification is required (for character
	and string literal) or no check can result from the conversion (for
	access types).
	* opt.ads: Update comment.

From-SVN: r260458

This patch modifies the semantics of pragma Elaboration_Checks. The pragma
was intended to be a configuration pragma, however its placement was never
verified until now.

The pragma may appear in the following contexts:

   * Configuration pragmas file

   * Prior to the context clauses of a compilation unit's initial declaration

Any other placement of the pragma will result in a warning and the effects of
the offending pragma will be ignored.

------------
-- Source --
------------

--  elab_checks_1.adc

pragma Elaboration_Checks (Dynamic);

--  elab_checks_2.adc

pragma Elaboration_Checks (Dynamic);
pragma Elaboration_Checks (Static);                                  --  Error

--  pack_1.ads

pragma Elaboration_Checks (Static);                                  --  OK

package Pack_1 is
end Pack_1;

--  pack_2.ads

pragma Elaboration_Checks (Static);                                  --  OK
pragma Elaboration_Checks (Static);                                  --  Error

package Pack_2 is
end Pack_2;

--  pack_3.ads

package Pack_3 is
   procedure Proc;
end Pack_3;

--  pack_3.adb

pragma Elaboration_Checks (Static);                                  --  Error

package body Pack_3 is
   procedure Proc is begin null; end Proc;
end Pack_3;

--  pack_4.ads

package Pack_4 is
   procedure Proc;
end Pack_4;

--  pack_4.adb

package body Pack_4 is
   procedure Proc is separate;
end Pack_4;

--  pack_4-proc.adb

pragma Elaboration_Checks (Static);                                  --  Error

separate (Pack_4)
procedure Proc is begin null; end Proc;

--  gen.ads

generic
   with function Called_At_Elaboration return Boolean;

package Gen is
   procedure Proc;
end Gen;

--  gen.adb

package body Gen is
   procedure Proc is
      Obj : constant Boolean := Called_At_Elaboration;
   begin null; end Proc;
begin
   Proc;
end Gen;

--  abe_static.ads

pragma Elaboration_Checks (Static);

with Gen;

package ABE_Static is
   function ABE return Boolean;

   package Inst_1 is new Gen (ABE);
end ABE_Static;

--  abe_static.adb

package body ABE_Static is
   package Inst_2 is new Gen (ABE);

   package Subunit is
   end Subunit;

   package body Subunit is separate;

   function ABE return Boolean is
   begin
      return True;
   end ABE;
end ABE_Static;

--  abe_static-subunit.adb

separate (ABE_Static)

package body Subunit is
   package Inst_3 is new Gen (ABE);

   package Nested_Subunit is
   end Nested_Subunit;

   package body Nested_Subunit is separate;
end Subunit;

--  abe_static-subunit-nested_subunit.adb

separate (ABE_Static.Subunit)

package body Nested_Subunit is
   package Inst_4 is new Gen (ABE);
end Nested_Subunit;

--  abe_static_main.adb

with ABE_Static;

procedure ABE_Static_Main is begin null; end ABE_Static_Main;

----------------------------
-- Compilation and output --
----------------------------

$ gcc -c pack_1.ads -gnatec=elab_checks_1.adc
$ gcc -c pack_1.ads -gnatec=elab_checks_2.adc
$ gcc -c pack_1.ads
$ gcc -c pack_2.ads
$ gcc -c pack_3.adb
$ gcc -c pack_4.adb
$ gnatmake -q -gnatE abe_static_main.adb
elab_checks_2.adc:2:01: pragma "Elaboration_Checks" duplicates pragma declared
  at line 1
pack_2.ads:2:01: pragma "Elaboration_Checks" duplicates pragma declared at line
  1
pack_3.adb:1:01: warning: effects of pragma "Elaboration_Checks" are ignored
pack_3.adb:1:01: warning: place pragma on initial declaration of library unit
pack_4-proc.adb:1:01: warning: effects of pragma "Elaboration_Checks" are
  ignored
pack_4-proc.adb:1:01: warning: place pragma on initial declaration of library
  unit
abe_static.adb:2:04: warning: in instantiation at gen.adb:3
abe_static.adb:2:04: warning: cannot call "ABE" before body seen
abe_static.adb:2:04: warning: Program_Error may be raised at run time
abe_static.adb:2:04: warning:   body of unit "ABE_Static" elaborated
abe_static.adb:2:04: warning:   procedure "Proc" called at gen.adb:6, instance
  at line 2
abe_static.adb:2:04: warning:   function "ABE" called at gen.adb:3, instance at
  line 2
abe_static.ads:8:04: warning: in instantiation at gen.adb:3
abe_static.ads:8:04: warning: cannot call "ABE" before body seen
abe_static.ads:8:04: warning: Program_Error may be raised at run time
abe_static.ads:8:04: warning:   spec of unit "ABE_Static" elaborated
abe_static.ads:8:04: warning:   procedure "Proc" called at gen.adb:6, instance
  at line 8
abe_static.ads:8:04: warning:   function "ABE" called at gen.adb:3, instance at
  line 8
abe_static-subunit.adb:4:04: warning: in instantiation at gen.adb:3
abe_static-subunit.adb:4:04: warning: cannot call "ABE" before body seen
abe_static-subunit.adb:4:04: warning: Program_Error may be raised at run time
abe_static-subunit.adb:4:04: warning:   body of unit "ABE_Static" elaborated
abe_static-subunit.adb:4:04: warning:   procedure "Proc" called at gen.adb:6,
  instance at line 4
abe_static-subunit.adb:4:04: warning:   function "ABE" called at gen.adb:3,
  instance at line 4
abe_static-subunit-nested_subunit.adb:4:04: warning: in instantiation at
  gen.adb:3
abe_static-subunit-nested_subunit.adb:4:04: warning: cannot call "ABE" before
  body seen
abe_static-subunit-nested_subunit.adb:4:04: warning: Program_Error may be
  raised at run time
abe_static-subunit-nested_subunit.adb:4:04: warning:   body of unit
  "ABE_Static" elaborated
abe_static-subunit-nested_subunit.adb:4:04: warning:   procedure "Proc" called
  at gen.adb:6, instance at line 4
abe_static-subunit-nested_subunit.adb:4:04: warning:   function "ABE" called at
  gen.adb:3, instance at line 4
warning: "abe_static_main.adb" has dynamic elaboration checks and with's
warning:   "abe_static.ads" which has static elaboration checks

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Install the elaboration
	model of the compilation unit spec, if any.
	* sem_ch7.adb (Analyze_Package_Body_Helper): Install the elaboration
	model of the compilation unit spec, if any.
	* sem_ch10.adb (Analyze_Subunit): Install the elaboration model of the
	parent compilation unit spec, if any.
	* sem_elab.adb (Check_Elaboration_Scenarios): Restore the elaboration
	model of the main unit.
	(Is_Same_Unit): The routine now uses Unit_Entity.
	(Is_Subunit): Removed.
	(Normalize_Unit): Removed.
	(Unit_Entity): New routine.
	* sem_prag.adb (Analyze_Pragma): Reimplement the handling of pragma
	Elaboration_Checks. The analysis now ensures that the pragma appears at
	the configuration level, and on the initial declaration of a unit.
	Other placements are either flagged as illegal, or ignored.
	(Check_Duplicate_Elaboration_Checks_Pragma): New routine.
	(Ignore_Elaboration_Checks_Pragma): New routine.
	* sem_util.adb (Install_Elaboration_Model): New routine.
	* sem_util.ads (Install_Elaboration_Model): New routine.
	* doc/gnat_rm/implementation_defined_pragmas.rst: Update the
	documentation of pragma Elaboration_Checks.
	* gnat_rm.texi: Regenerate.

From-SVN: r260457

Any program calling Gnat.Traceback.Symbolic.Enable_Cache for
dwarf based symbolization fails with a segmentation violation
when spawned with an inaccurate argv[0] such that it couldn't
be found on PATH.

argv[0] is most often found on PATH. One plausible case where
it isn't is when argv[0] is a mere file name and . isn't on PATH,
which might happen out of imprecise exec calls.

This change robustifies the Traceback.Symbolic implementation
to work in this case as well, by just trying to work with argv[0]
untouched as the executable file to fetch dwarf info from.

2018-05-21  Olivier Hainque  <hainque@adacore.com>

gcc/ada/

	* libgnat/s-trasym__dwarf.adb (Executable_Name): Return argv[0] instead
	of empty string when argv[0] couldn't be found on PATH.
	(Enable_Cache): Raise Program_Error instead of attempting a null
	pointer dereference when the Exec_Module initialization failed.

From-SVN: r260456

Explicit External aspect was an equivalant to an implicit default.  It was only
needed as a workaround for a frontend bug. (If it meant to serve as
documentation, there should be explicit Effective_Reads and Effective_Writes
set to False too.)

No test, because these changes are semantically neutral.

2018-05-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

	* libgnarl/a-reatim.ads (Clock_Time): Remove External aspect.
	* libgnarl/a-taside.ads (Tasking_State): Likewise.
	* libgnat/a-calend.ads (Clock_Time): Likewise.

From-SVN: r260455

During the special inlining done in GNATprove mode, a call in prefix
notation leads to a spurious error. Now fixed.

2018-05-21  Yannick Moy  <moy@adacore.com>

gcc/ada/

	* sem_ch6.adb (Analyze_Procedure_Call): Refine test to recognize prefix
	call notation in inlined call in GNATprove mode.

From-SVN: r260454

This patch ensures that an abstract state declared with simple option
"synchronous" is automatically considered "external".

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* einfo.adb (Is_External_State): An abstract state is also external
	when it is declared with option "synchronous".
	* einfo.ads: Update the documentation of synthesized attribute
	Is_External_State.
	* sem_util.adb (Find_Simple_Properties): New routine.
	(Is_Enabled_External_Property): New routine.
	(State_Has_Enabled_Property): Reimplemented. The two flavors of option
	External have precedence over option Synchronous when determining
	whether a property is in effect.

gcc/testsuite/

	* gnat.dg/sync2.adb, gnat.dg/sync2.ads: New testcase.

From-SVN: r260453

In some cases, the inlining performed in GNATprove mode leads to a crash,
when inlining a call where a return statement of the inlined function
returns a string literal. Now fixed.

2018-05-21  Yannick Moy  <moy@adacore.com>

gcc/ada/

	* sem_eval.adb (Static_Length): Take into account case of variable of
	subtype string literal.

From-SVN: r260452

Symbolization of traceback entries from dwarf info was
failing in some cases with shared libraries on ELF targets,
from unexpected overlapping of what we believed were code
regions for distinct modules.

This is caused by the inclusion of all SHF_ALLOC sections in
the set of sections of possible relevance to determine the span
of possible code addresses for a module.

This change renames the Get_memory_Bound subprogram to better
convey that we really care about sections hosting executable code
in particular, matching what the spec comments already claims.

It also renames the boolean flag conveying the info of relevance
in the Object_Section record, and adjusts the ELF implementation
of Get_Section to feed this flag from SHF_EXECINSTR instead of
SHF_ALLOC.

2018-05-21  Olivier Hainque  <hainque@adacore.com>

gcc/ada/

	* libgnat/s-objrea.ads (Get_Memory_Bounds): Rename as Get_Xcode_Bounds.
	(Object_Section): Rename Flag_Alloc component as Flag_Xcode.
	* libgnat/s-objrea.adb (Get_Xcode_Bounds): Adjust to new subprogram and
	component name.
	(Get_Section, ELF case): Set Flag_Xcode from SHF_EXECINSTR.
	* libgnat/s-dwalin.adb (Open): Adjust to the Get_Memory_Bounds name
	change.

From-SVN: r260451

Symbolization of traceback entries from dwarf info is
failing in multiple cases for addresses originating from
shared libraries.

Part of the problem is a confusion across different functions
regarding the kind of "address" at hand, sometimes full process
runtime addresses (e.g. in traceback entries), sometimes module
relative (e.g. in dwarf info segments).

This change fixes this by introducing the use of distinct types
for the two kinds of addresses, resorting to System.Address
for runtime addresses and to Storage_Elements.Storage_Offset
for module relative values. The accompanying code changes
tidy a few places where we can now use standard operators
to combine offets & addresses, and include a few corrections
of consistency problems at spots where comparisons were done
between runtime addresses and relative offsets.

2018-05-21  Olivier Hainque  <hainque@adacore.com>

gcc/ada/

	* libgnat/s-dwalin.ads (Dwarf_Context): Change type of Load_Address to
	Address, and type of Low, High to Storage_Offset.
	(Low): Rename as Low_Address and convey that the return value is a
	runtime reference accounting for a load address.
	* libgnat/s-dwalin.adb (Read_Aranges_Entry): Adjust to the
	address/offset type changes.
	(Aranges_Lookup): Likewise.
	(Symbolic_Address): Likewise.
	(Symbolic_Traceback): Likewise.
	(Dump_Cache): Likewise.
	(Is_Inside): Likewise.
	(Open): Likewise.
	(Set_Load_Address): Likewise.
	(Low_Address): Likewise, and account for C.Load_Address.
	* libgnat/s-trasym__dwarf.adb (Lt): Use Low_Address instead of Low.
	(Multi_Module_Symbolic_Traceback): Compare address in traceback
	with module Low_Address instead of Low.

From-SVN: r260450

Load_Address is just more consistent with the overall use of the component.

2018-05-21  Olivier Hainque  <hainque@adacore.com>

gcc/ada/

	* libgnat/s-dwalin.ads (Dwarf_Context): Rename Load_Slide as
	Load_Address.
	* libgnat/s-dwalin.adb (Is_Inside): Adjust accordingly.
	(Set_Load_Address): Likewise.
	(Symbolic_Traceback): Likewise.

From-SVN: r260449

Symbolization of traceback addresses through shared libraries
requires information on the shared libraries load addresses, which
was at hand on Linuxbut not propagated through the runtime when
caching is enabled.  This change fixes this.

2018-05-21  Olivier Hainque  <hainque@adacore.com>

gcc/ada/

	* libgnat/s-trasym__dwarf.adb (Add_Module_To_Cache): Expect a
	Load_Address argument and pass it down to Init_Module.
	* libgnat/s-tsmona__linux.adb (Build_Cache_For_All_Modules): Pass
	lm.l_addr as the Load_Address to Add_Module_To_Cache.

From-SVN: r260448

This patch enforces what the comment for Has_Discriminant says:

--    Has_Discriminants (Flag5)
--       Defined in all types and subtypes.

to avoid semantically undefined calls on non-type entities. It also adapts
other routines to respect this comment.

No user-visible impact.

2018-05-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

	* einfo.adb (Has_Discriminants): Stronger assertion.
	(Set_Has_Discriminants): Stronger assertion.
	* sem_ch13.adb (Push_Scope_And_Install_Discriminants): Adapt to respect
	the stronger assertion on Has_Discriminant.
	(Uninstall_Discriminants_And_Pop_Scope): Same as above.
	* sem_util.adb (New_Copy_Tree): Same as above.
	* sem_ch7.adb (Generate_Parent_References): Prevent calls to
	Has_Discriminant on non-type entities that might happen when the
	compiled code has errors.
	* sem_ch3.adb (Derived_Type_Declaration): Only call
	Set_Has_Discriminant on type entities.

From-SVN: r260447

2018-05-21  Arnaud Charlet  <charlet@adacore.com>

gcc/ada/

	* exp_unst.adb (Unnest_Subprogram): Unnest all subprograms relevant for
	code generation.

From-SVN: r260446

2018-05-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

	* lib.ads: Fix typo in enumeration in comment.

From-SVN: r260445

This patch extends the legality of the GNAT attribute Scalar_Storage_Order,
to apply to formal private types. Previously this extension applied only
in GNAT_Mode, to support instantiations of Ada.Sequential_IO, but it is more
generally useful.

The following must compile quietly:

----
with Memory_View_Generic;
procedure Main is
   type T is array (1..10) of integer;
   package OK is new Memory_View_Generic (T);

   type T2 is new Long_Float;
   package Wrong is new Memory_View_Generic (T2);
begin
   null;
end;
----
with System;
generic
   type Source_Type is private;
package Memory_View_Generic is
   -- various declarations ...
   SSO : System.Bit_Order := Source_Type'Scalar_Storage_Order;
end Memory_View_Generic;

2018-05-21  Ed Schonberg  <schonberg@adacore.com>

gcc/ada/

	* sem_attr.adb (Analyze_Attribute, case Scalar_Storage_Order): The
	attribute reference is legal within a generic unit when the prefix is a
	formal private type.

From-SVN: r260444

This patch modifies the creation of transient scopes to eliminate potential
premature secondary stack reclamations when there is no suitable transient
context and the scope was intended to manage the secondary stack. Instead,
the logic was changed to accommodate a special case where an assignment with
suppressed controlled actions that appears within a type initialization
procedure requires secondary stack reclamation.

The patch also corrects the handling of function calls which utilize the
secondary stack in loop parameter specifications. Previously the predicate
which determined whether the function will utilize the secondary stack was
not accurate enough, and in certain cases could lead to leaks.

------------
-- Source --
------------

--  iterators.ads

package Iterators is
   type Iterator is limited interface;
   type Iterator_Access is access all Iterator'Class;

   function Next
     (I       : in out Iterator;
      Element : out Character) return Boolean is abstract;

   procedure Iterate
     (I    : in out Iterator'Class;
      Proc : access procedure (Element : Character));
end Iterators;

--  iterators.adb

package body Iterators is
   procedure Iterate
     (I    : in out Iterator'Class;
      Proc : access procedure (Element : Character))
   is
      Element : Character;
   begin
      while I.Next (Element) loop
         Proc (Element);
      end loop;
   end Iterate;
end Iterators;

--  base.ads

with Iterators; use Iterators;

package Base is
   type String_Access is access all String;
   type Node is tagged record
      S : String_Access;
   end record;

   type Node_Access is access all Node'Class;
   type Node_Array is array (Positive range <>) of Node_Access;

   function As_Array (N : Node_Access) return Node_Array;
   function Get_String (C : Character) return String;

   type Node_Iterator is limited new Iterator with record
      Node : Node_Access;
      I    : Positive;
   end record;

   overriding function Next
     (It      : in out Node_Iterator;
      Element : out Character) return Boolean;

   function Constructor_1 (N : Node_Access) return Node_Iterator;
   function Constructor_2 (N : Node_Access) return Node_Iterator;
end Base;

--  base.adb

package body Base is
   function As_Array (N : Node_Access) return Node_Array is
   begin
      return (1 => N);
   end As_Array;

   function Get_String (C : Character) return String is
   begin
      return (1 .. 40 => C);
   end Get_String;

   function Next
     (It      : in out Node_Iterator;
      Element : out Character) return Boolean
   is
   begin
      if It.I > It.Node.S'Last then
         return False;
      else
         It.I := It.I + 1;
         Element := It.Node.S (It.I - 1);
         return True;
      end if;
   end Next;

   function Constructor_1 (N : Node_Access) return Node_Iterator is
   begin
      return Node_Iterator'(N, 1);
   end Constructor_1;

   function Constructor_2 (N : Node_Access) return Node_Iterator is
   begin
      return Constructor_1 (As_Array (N) (1));
   end Constructor_2;
end Base;

--  main.adb

with Ada.Text_IO; use Ada.Text_IO;
with Base;        use Base;
with Iterators;   use Iterators;

procedure Main is
   N : constant Node_Access := new Node'(S => new String'("hello world"));

   procedure Process (C : Character) is
   begin
      Put_Line (Get_String (C));
   end Process;

   C : Iterator'Class := Constructor_2 (N);

begin
   C.Iterate (Process'Access);
end Main;

----------------------------
-- Compilation and output --
----------------------------

$ gnatmake -q main.adb
$ ./main
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
llllllllllllllllllllllllllllllllllllllll
llllllllllllllllllllllllllllllllllllllll
oooooooooooooooooooooooooooooooooooooooo

wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww
oooooooooooooooooooooooooooooooooooooooo
rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr
llllllllllllllllllllllllllllllllllllllll
dddddddddddddddddddddddddddddddddddddddd

2018-05-21  Hristian Kirtchev  <kirtchev@adacore.com>

gcc/ada/

	* exp_ch7.adb (Establish_Transient_Scope): Code cleanup. Do not
	delegate the secondary stack management when there is no suitable
	transient context, and the transient scope was intended to manage the
	secondary stack because this causes premature reclamation. Change the
	transient scope creation logic by special casing assignment statements
	of controlled components for type initialization procedures.
	(Find_Node_To_Be_Wrapped): Renamed to Find_Transient_Context. Update
	the comment on usage.
	(Find_Transient_Context): Change the initinte loop into a while loop.
	Iterations schemes and iterator specifications are not valid transient
	contexts because they rely on special processing. Assignment statements
	are now treated as a normal transient context, special cases are
	handled by the caller. Add special processing for pragma Check.
	(Is_OK_Construct): Removed. Its functionality has been merged in
	routine Find_Transient_Context.
	* sem_ch5.adb (Check_Call): Reimplemented. Add code to properly
	retrieve the subprogram being invoked. Use a more accurate predicate
	(Requires_Transient_Scope) to determine that the function will emply
	the secondary stack.

From-SVN: r260443

These are GNAT style violations detected with a trivial Libadalang checker.

2018-05-21  Piotr Trojanek  <trojanek@adacore.com>

gcc/ada/

	* ada_get_targ.adb: Fix subprogram body headers.
	* adabkend.adb: Likewise.
	* checks.adb: Likewise.
	* exp_ch3.adb: Likewise.
	* exp_ch5.adb: Likewise.
	* exp_ch9.adb: Likewise.
	* exp_dist.adb: Likewise.
	* exp_tss.adb: Likewise.
	* inline.adb: Likewise.
	* lib-writ.adb: Likewise.
	* lib-xref-spark_specific.adb: Likewise.
	* libgnarl/s-osinte__darwin.adb: Likewise.
	* libgnarl/s-stusta.adb: Likewise.
	* libgnarl/s-taprop__solaris.adb: Likewise.
	* libgnarl/s-tposen.adb: Likewise.
	* libgnarl/s-vxwext__kernel-smp.adb: Likewise.
	* libgnarl/s-vxwext__kernel.adb: Likewise.
	* libgnat/a-btgbso.adb: Likewise.
	* libgnat/a-cfdlli.adb: Likewise.
	* libgnat/a-cfhama.adb: Likewise.
	* libgnat/a-cfinve.adb: Likewise.
	* libgnat/a-cimutr.adb: Likewise.
	* libgnat/a-coboho.adb: Likewise.
	* libgnat/a-cofove.adb: Likewise.
	* libgnat/a-cofuve.adb: Likewise.
	* libgnat/a-comutr.adb: Likewise.
	* libgnat/a-exexda.adb: Likewise.
	* libgnat/a-tags.adb: Likewise.
	* libgnat/a-tideau.adb: Likewise.
	* libgnat/a-wtdeau.adb: Likewise.
	* libgnat/a-ztdeau.adb: Likewise.
	* libgnat/g-alleve.adb: Likewise.
	* libgnat/s-excdeb.adb: Likewise.
	* libgnat/s-parint.adb: Likewise.
	* libgnat/s-shasto.adb: Likewise.
	* libgnat/s-traceb__hpux.adb: Likewise.
	* prepcomp.adb: Likewise.
	* sem_ch4.adb: Likewise.
	* sem_ch6.adb: Likewise.
	* sem_dist.adb: Likewise.
	* sem_prag.adb: Likewise.
	* sem_util.adb: Likewise.
	* sinfo.adb: Likewise.
	* switch.adb: Likewise.

From-SVN: r260442

	* src/filesystem/std-ops.cc (absolute): Report an error for empty
	paths.
	(weakly_canonical(const path&)): Do not call canonical on empty path.
	(weakly_canonical(const path&, error_code&)): Likewise.
	* testsuite/27_io/filesystem/operations/absolute.cc: Check for errors.

From-SVN: r260441

	PR libstdc++/85818
	* testsuite/experimental/filesystem/path/preferred_separator.cc: Add
	dg-require-filesystem-ts.

From-SVN: r260439

From-SVN: r260438

This patch implements the usadv16qi and ssadv16qi standard names.
See the thread at on gcc@gcc.gnu.org [1] for background.

The V16QImode variant is important to get right as it is the most commonly used pattern:
reducing vectors of bytes into an int.
The midend expects the optab to compute the absolute differences of operands 1 and 2 and
reduce them while widening along the way up to SImode. So the inputs are V16QImode and
the output is V4SImode.

I've tried out a few different strategies for that, the one I settled with is to emit:
UABDL2    tmp.8h, op1.16b, op2.16b
UABAL    tmp.8h, op1.16b, op2.16b
UADALP    op3.4s, tmp.8h

To work through the semantics let's say operands 1 and 2 are:
op1 { a0, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11, a12, a13, a14, a15 }
op2 { b0, b1, b2, b3, b4, b5, b6, b7, b8, b9, b10, b11, b12, b13, b14, b15 }
op3 { c0, c1, c2, c3 }

The UABDL2 takes the upper V8QI elements, computes their absolute differences, widens them and stores them into the V8HImode tmp:

tmp { ABS(a[8]-b[8]), ABS(a[9]-b[9]), ABS(a[10]-b[10]), ABS(a[11]-b[11]), ABS(a[12]-b[12]), ABS(a[13]-b[13]), ABS(a[14]-b[14]), ABS(a[15]-b[15]) }

The UABAL after that takes the lower V8QI elements, computes their absolute differences, widens them and accumulates them into the V8HImode tmp from the previous step:

tmp { ABS(a[8]-b[8])+ABS (a[0]-b[0]), ABS(a[9]-b[9])+ABS(a[1]-b[1]), ABS(a[10]-b[10])+ABS(a[2]-b[2]), ABS(a[11]-b[11])+ABS(a[3]-b[3]), ABS(a[12]-b[12])+ABS(a[4]-b[4]), ABS(a[13]-b[13])+ABS(a[5]-b[5]), ABS(a[14]-b[14])+ABS(a[6]-b[6]), ABS(a[15]-b[15])+ABS(a[7]-b[7]) }

Finally the UADALP does a pairwise widening reduction and accumulation into the V4SImode op3:
op3 { c0+ABS(a[8]-b[8])+ABS(a[0]-b[0])+ABS(a[9]-b[9])+ABS(a[1]-b[1]), c1+ABS(a[10]-b[10])+ABS(a[2]-b[2])+ABS(a[11]-b[11])+ABS(a[3]-b[3]), c2+ABS(a[12]-b[12])+ABS(a[4]-b[4])+ABS(a[13]-b[13])+ABS(a[5]-b[5]), c3+ABS(a[14]-b[14])+ABS(a[6]-b[6])+ABS(a[15]-b[15])+ABS(a[7]-b[7]) }

(sorry for the text dump)

Remember, according to [1] the exact reduction sequence doesn't matter (for integer arithmetic at least).
I've considered other sequences as well (thanks Wilco), for example
* UABD + UADDLP + UADALP
* UABLD2 + UABDL + UADALP + UADALP

I ended up settling in the sequence in this patch as it's short (3 instructions) and in the future we can potentially
look to optimise multiple occurrences of these into something even faster (for example accumulating into H registers for longer
before doing a single UADALP in the end to accumulate into the final S register).

If your microarchitecture has some some strong preferences for a particular sequence, please let me know or, even better, propose a patch
to parametrise the generation sequence by code (or the appropriate RTX cost).


This expansion allows the vectoriser to avoid unpacking the bytes in two steps and performing V4SI arithmetic on them.
So, for the code:

unsigned char pix1[N], pix2[N];

int foo (void)
{
  int i_sum = 0;
  int i;

  for (i = 0; i < 16; i++)
    i_sum += __builtin_abs (pix1[i] - pix2[i]);

  return i_sum;
}

we now generate on aarch64:
foo:
        adrp    x1, pix1
        add     x1, x1, :lo12:pix1
        movi    v0.4s, 0
        adrp    x0, pix2
        add     x0, x0, :lo12:pix2
        ldr     q2, [x1]
        ldr     q3, [x0]
        uabdl2  v1.8h, v2.16b, v3.16b
        uabal   v1.8h, v2.8b, v3.8b
        uadalp  v0.4s, v1.8h
        addv    s0, v0.4s
        umov    w0, v0.s[0]
        ret


instead of:
foo:
        adrp    x1, pix1
        adrp    x0, pix2
        add     x1, x1, :lo12:pix1
        add     x0, x0, :lo12:pix2
        ldr     q0, [x1]
        ldr     q4, [x0]
        ushll   v1.8h, v0.8b, 0
        ushll2  v0.8h, v0.16b, 0
        ushll   v2.8h, v4.8b, 0
        ushll2  v4.8h, v4.16b, 0
        usubl   v3.4s, v1.4h, v2.4h
        usubl2  v1.4s, v1.8h, v2.8h
        usubl   v2.4s, v0.4h, v4.4h
        usubl2  v0.4s, v0.8h, v4.8h
        abs     v3.4s, v3.4s
        abs     v1.4s, v1.4s
        abs     v2.4s, v2.4s
        abs     v0.4s, v0.4s
        add     v1.4s, v3.4s, v1.4s
        add     v1.4s, v2.4s, v1.4s
        add     v0.4s, v0.4s, v1.4s
        addv    s0, v0.4s
        umov    w0, v0.s[0]
        ret

So I expect this new expansion to be better than the status quo in any case.
Bootstrapped and tested on aarch64-none-linux-gnu.
This gives about 8% on 525.x264_r from SPEC2017 on a Cortex-A72.

	* config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
	UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
	UNSPEC_UADALP values.
	* config/aarch64/iterators.md (ABAL): New int iterator.
	(ABDL2): Likewise.
	(ADALP): Likewise.
	(sur): Add mappings for the above.
	* config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
	New define_insn.
	(aarch64_<sur>abal<mode>_4): Likewise.
	(aarch64_<sur>adalp<mode>_3): Likewise.
	(<sur>sadv16qi): New define_expand.

	* gcc.c-torture/execute/ssad-run.c: New test.
	* gcc.c-torture/execute/usad-run.c: Likewise.
	* gcc.target/aarch64/ssadv16qi.c: Likewise.
	* gcc.target/aarch64/usadv16qi.c: Likewise.

From-SVN: r260437

2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>

gcc/
        * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
        (*movdf_internal): Ditto.
        (*rcpsf2_sse): Ditto.
        (*rsqrtsf2_sse): Ditto.
        (*sqrt<mode>2_sse): Ditto.

From-SVN: r260436

This patch adds the missing neon intrinsics for all 128 bit vector Integer modes for the
three-way XOR and negate and xor instructions for Arm8.2-a to Armv8.4-a.

gcc/
2018-05-21  Tamar Christina  <tamar.christina@arm.com>

	* config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
	eor3q<mode>4.
	(aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
	* config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
	veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
	vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
	vbcaxq_s64): New.
	* config/aarch64/arm_neon.h: Likewise.
	* config/aarch64/iterators.md (VQ_I): New.

gcc/testsuite/
2018-05-21  Tamar Christina  <tamar.christina@arm.com>

	* gcc.target/gcc.target/aarch64/sha3.h (veor3q_u8, veor3q_u32,
	veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
	vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
	vbcaxq_s64): New.
	* gcc.target/gcc.target/aarch64/sha3_1.c: Likewise.
	* gcc.target/gcc.target/aarch64/sha3_1.c: Likewise.
	* gcc.target/gcc.target/aarch64/sha3_1.c: Likewise.

From-SVN: r260435