Jonathan Wakely [Mon, 21 May 2018 17:18:35 +0000 (18:18 +0100)]
Add support for opening file streams from wide character strings
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.
Kyrylo Tkachov [Mon, 21 May 2018 15:58:32 +0000 (15:58 +0000)]
[AArch64][committed] Fix gcc.target/aarch64/vec_init_1.c for tiny and large mcmodels
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
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.
Ed Schonberg [Mon, 21 May 2018 14:52:36 +0000 (14:52 +0000)]
[Ada] Spurious warning on object declaration with address clause
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.
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);
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;
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
* 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.
[Ada] Spurious error on early call region of tagged type
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;
* 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.
Ed Schonberg [Mon, 21 May 2018 14:52:05 +0000 (14:52 +0000)]
[Ada] Error message on invalid conversion involving limited views
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.
Ed Schonberg [Mon, 21 May 2018 14:52:00 +0000 (14:52 +0000)]
[Ada] Exit statement in loops over iterable objects
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.
Javier Miranda [Mon, 21 May 2018 14:51:55 +0000 (14:51 +0000)]
[Ada] Avoid compiler crash for body in non Ada 2012 mode
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.
Doug Rupp [Mon, 21 May 2018 14:51:48 +0000 (14:51 +0000)]
[Ada] Enhance stack unwinding on VxWorks for AArch64
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.
Ed Schonberg [Mon, 21 May 2018 14:51:35 +0000 (14:51 +0000)]
[Ada] Spurious error on indexed call as prefix of a call
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.
Yannick Moy [Mon, 21 May 2018 14:51:19 +0000 (14:51 +0000)]
[Ada] Use type conversion when inlining may trigger a run-time check
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.
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.
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
* 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.
Olivier Hainque [Mon, 21 May 2018 14:51:09 +0000 (14:51 +0000)]
[Ada] Robustify traceback caching for executable in current dir
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.
Piotr Trojanek [Mon, 21 May 2018 14:51:04 +0000 (14:51 +0000)]
[Ada] Remove External aspect from predefined abstract states
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.
* 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.
Yannick Moy [Mon, 21 May 2018 14:50:49 +0000 (14:50 +0000)]
[Ada] Crash on incorrect frontend inlining in GNATprove mode
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.
Olivier Hainque [Mon, 21 May 2018 14:50:44 +0000 (14:50 +0000)]
[Ada] Tighten Object_Reader.Get_Memory_Bounds
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.
Olivier Hainque [Mon, 21 May 2018 14:50:38 +0000 (14:50 +0000)]
[Ada] Robustify traceback symbolization from dwarf info
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.
Olivier Hainque [Mon, 21 May 2018 14:50:28 +0000 (14:50 +0000)]
[Ada] Propagate load addresses for traceback cache on Linux
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.
Piotr Trojanek [Mon, 21 May 2018 14:50:23 +0000 (14:50 +0000)]
[Ada] Only allow Has_Discriminants on type entities
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.
Ed Schonberg [Mon, 21 May 2018 14:50:06 +0000 (14:50 +0000)]
[Ada] Extend legality of Scalar_Storage_Order to formal types
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.
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 --
----------------------------
* 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.
Jonathan Wakely [Mon, 21 May 2018 12:52:44 +0000 (13:52 +0100)]
Fix std::filesystem::absolute for empty paths
* 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.
Kyrylo Tkachov [Mon, 21 May 2018 11:21:07 +0000 (11:21 +0000)]
[AArch64] Implement usadv16qi and ssadv16qi standard names
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:
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:
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]);
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.
Tamar Christina [Mon, 21 May 2018 10:33:30 +0000 (10:33 +0000)]
Add missing AArch64 NEON instrinctics for Armv8.2-a to Armv8.4-a
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>
Alexey Brodkin [Mon, 21 May 2018 09:56:57 +0000 (09:56 +0000)]
[ARC] Add multilib support for linux targets
We used to build baremetal (AKA Elf32) multilibbed toolchains for years
now but never made that for Linux targets since there were problems with
uClibc n multilib setup. Now with help of Crosstool-NG it is finally
possible to create uClibc-based multilibbed toolchains and so we add
relevant CPUs for multilib in case of configuration for "arc*-*-linux*".
This will be essentially useful for glibc-based multilibbbed toolchains
in the future.
Janus Weil [Mon, 21 May 2018 06:45:55 +0000 (08:45 +0200)]
re PR fortran/85841 ([F2018] reject deleted features)
2018-05-21 Janus Weil <janus@gcc.gnu.org>
PR fortran/85841
* libgfortran.h: New macros GFC_STD_OPT_*.
* error.c (notify_std_msg): New function.
(gfc_notify_std): Adjust such that it can handle combinations of
GFC_STD_* flags in the 'std' argument, not just a single one.
* match.c (match_arithmetic_if, gfc_match_if): Reject arithmetic if
in Fortran 2018.
(gfc_match_stopcode): Use GFC_STD_OPT_* macros.
* options.c (set_default_std_flags): Warn for F2018 deleted features
by default.
(gfc_handle_option): F2018 deleted features are allowed in earlier
standards.
* symbol.c (gfc_define_st_label, gfc_reference_st_label): Reject
nonblock do constructs in Fortran 2018.
Paul Thomas [Sun, 20 May 2018 10:54:24 +0000 (10:54 +0000)]
re PR fortran/82275 (gfortran rejects valid & accepts invalid reference to dimension-remapped type SELECT TYPE selector)
2018-05-20 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82275
Correcting ChangeLogs
* match.c (gfc_match_type_spec): Go through the array ref and
decrement 'rank' for every dimension that is an element.
2018-05-20 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82275
Correcting ChangeLogs
* gfortran.dg/select_type_42.f90: New test.
Paul Thomas [Sun, 20 May 2018 10:08:24 +0000 (10:08 +0000)]
re PR fortran/80657 (Loop in character function declaration)
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/80657
* resolve.c (flag_fn_result_spec): Use the 'sym' argument to
test for self refs to the function result in the character len
expression. If a self reference is found, emit an error and
return true.
(resolve_fntype): Use the function symbol in the calls to the
above.
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/80657
* gfortran.dg/char_result_18.f90: New test.
Paul Thomas [Sun, 20 May 2018 10:04:46 +0000 (10:04 +0000)]
re PR fortran/49636 ([F03] ASSOCIATE construct confused with slightly complicated case)
2018-05-20 Paul Thomas <pault@gcc.gnu.org>
PR fortran/49636
* trans-array.c (gfc_get_array_span): Renamed from
'get_array_span'.
(gfc_conv_expr_descriptor): Change references to above.
* trans-array.h : Add prototype for 'gfc_get_array_span'.
* trans-stmt.c (trans_associate_var): If the associate name is
a subref array pointer, use gfc_get_array_span for the span.
2018-05-20 Paul Thomas <pault@gcc.gnu.org>
PR fortran/49636
* gfortran.dg/associate_38.f90: New test.
Paul Thomas [Sun, 20 May 2018 09:59:54 +0000 (09:59 +0000)]
re PR fortran/82923 (Automatic allocation of deferred length character using function result)
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82923
PR fortran/66694
PR fortran/82617
* trans-array.c (gfc_alloc_allocatable_for_assignment): Set the
charlen backend_decl of the rhs expr to ss->info->string_length
so that the value in the current scope is used.
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82923
* gfortran.dg/allocate_assumed_charlen_4.f90: New test. Note
that the patch fixes PR66694 & PR82617, although the testcases
are not explicitly included.
Uros Bizjak [Sat, 19 May 2018 13:43:06 +0000 (15:43 +0200)]
i386.md (rex64namesuffix): New mode attribute.
* config/i386/i386.md (rex64namesuffix): New mode attribute.
* config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
Merge insn pattern from sse_cvtsi2ss<round_name> and
sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
(sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
using SWI48 mode iterator.
(sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
(sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
pattern from sse_cvttss2si<round_saeonly_name>
and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
(avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
using SWI48 mode iterator.
(avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
(avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
using SWI48 mode iterator.
(avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
(sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
SWI48 mode iterator.
(sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
(sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
pattern from sse_cvttsd2si<round_saeonly_name>
and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
Paul Thomas [Sat, 19 May 2018 10:49:50 +0000 (10:49 +0000)]
re PR fortran/82923 (Automatic allocation of deferred length character using function result)
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82923
PR fortran/66694
PR fortran/82617
* trans-array.c (gfc_alloc_allocatable_for_assignment): Set the
charlen backend_decl of the rhs expr to ss->info->string_length
so that the value in the current scope is used.
2018-05-19 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82923
* gfortran.dg/allocate_assumed_charlen_4.f90: New test. Note
that the patch fixes PR66694 & PR82617, although the testcases
are not explicitly included.
Jonathan Wakely [Sat, 19 May 2018 02:03:42 +0000 (03:03 +0100)]
Fix std::codecvt_utf8<wchar_t> for Mingw
* src/c++11/codecvt.cc (__codecvt_utf8_base<wchar_t>::do_in)
[__SIZEOF_WCHAR_T__==2 && __BYTE_ORDER__!=__ORDER_BIG_ENDIAN__]: Set
little_endian element in bitmask.
* testsuite/22_locale/codecvt/codecvt_utf8/69703.cc: Run all tests.
* testsuite/22_locale/codecvt/codecvt_utf8/wchar_t/1.cc: New.
Jonathan Wakely [Fri, 18 May 2018 16:14:04 +0000 (17:14 +0100)]
PR libstdc++/85098 add missing definitions for static constants
In C++11 and C++14 any odr-use of these constants requires a definition
at namespace-scope. In C++17 they are implicitly inline and so the
namespace-scope redeclarations are redundant (and allowing them is
deprecated).
PR libstdc++/85098
* include/bits/regex.h [__cplusplus < 201703L] (basic_regex::icase)
(basic_regex::nosubs, basic_regex::optimize, basic_regex::collate)
(basic_regex::ECMAScript, basic_regex::basic, basic_regex::extended)
(basic_regex::awk, basic_regex::grep, basic_regex::egrep): Add
definitions.
* include/bits/regex_automaton.h (_NFA::_M_insert_state): Adjust
whitespace.
* include/bits/regex_compiler.tcc (__INSERT_REGEX_MATCHER): Add
braces around body of do-while.
* testsuite/28_regex/basic_regex/85098.cc: New
Kyrylo Tkachov [Fri, 18 May 2018 13:10:36 +0000 (13:10 +0000)]
[arm][2/2] Remove support for -march=armv3 and older
We deprecated architecture versions earlier than Armv4T in GCC 6 [1].
This patch removes support for architectures lower than Armv4.
That is the -march values armv2, armv2a, armv3, armv3m are removed
with this patch. I did not remove armv4 because it's a bit more
involved code-wise and there has been some pushback on the implications
for -mcpu=strongarm support.
Removing armv3m and earlier though is pretty straightforward.
This allows us to get rid of the armv3m and mode32 feature bits
in arm-cpus.in as they can be assumed to be universally available.
Consequently the mcpu values arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dm are now also removed.
Bootstrapped and tested on arm-none-linux-gnueabihf and on arm-none-eabi
with an aprofile multilib configuration (which builds quite a lot of library
configurations).
Kyrylo Tkachov [Fri, 18 May 2018 13:08:16 +0000 (13:08 +0000)]
[arm][1/2] Remove support for deprecated -march=armv5 and armv5e
The -march=armv5 and armv5e options have been deprecated in GCC 7 [1].
This patch removes support for them.
It's mostly mechanical stuff. The functionality that was previously
gated on arm_arch5 is now gated on arm_arch5t and the functionality
that was gated on arm_arch5e is now gated on arm_arch5te.
A path in TARGET_OS_CPP_BUILTINS for VxWorks is now unreachable and
therefore is deleted.
References to armv5 and armv5e are deleted/updated throughout the
source tree and testsuite.
Bootstrapped and tested on arm-none-linux-gnueabihf.
Also built a cc1 for arm-wrs-vxworks as a sanity check.
* config/arm/arm-cpus.in (armv5, armv5e): Delete features.
(armv5t, armv5te): New features.
(ARMv5, ARMv5e): Delete fgroups.
(ARMv5t, ARMv5te): Adjust for above changes.
(ARMv6m): Likewise.
(armv5, armv5e): Delete arches.
* config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
arm_arch5.
(*call_reg_arm): Likewise.
(*call_value_reg_armv5): Likewise.
(*call_value_reg_arm): Likewise.
(*call_symbol): Likewise.
(*call_value_symbol): Likewise.
(*sibcall_insn): Likewise.
(*sibcall_value_insn): Likewise.
(clzsi2): Likewise.
(prefetch): Likewise.
(define_split and define_peephole2 dependent on arm_arch5):
Likewise.
* config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
arm_arch5e.
(TARGET_ARM_QBIT): Likewise.
(TARGET_DSP_MULTIPLY): Likewise.
(enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
(arm_arch5, arm_arch5e): Delete.
(arm_arch5t, arm_arch5te): Declare.
* config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
(arm_arch5t): Declare.
(arm_option_reconfigure_globals): Update for the above.
(arm_options_perform_arch_sanity_checks): Update comment, replace
use of arm_arch5 with arm_arch5t.
(use_return_insn): Likewise.
(arm_emit_call_insn): Likewise.
(output_return_instruction): Likewise.
(arm_final_prescan_insn): Likewise.
(arm_coproc_builtin_available): Likewise.
* config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
arm_arch5e with arm_arch5t and arm_arch5te.
* config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
(arm_arch5t, arm_arch5te): Declare.
* config/arm/arm-tables.opt: Regenerate.
* config/arm/t-arm-elf: Remove references to armv5, armv5e.
* config/arm/t-multilib: Likewise.
* config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
instead of arm_arch5.
(*call_reg_thumb1): Likewise.
(*call_value_reg_thumb1_v5): Likewise.
(*call_value_reg_thumb1): Likewise.
* config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
unreachable path.
* doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
* gcc.target/arm/pr40887.c: Update comment.
* lib/target-supports.exp: Don't generate effective target checks
and related helpers for armv5. Update comment.
* gcc.target/arm/armv5_thumb_isa.c: Delete.
* gcc.target/arm/di-longlong64-sync-withhelpers.c: Update effective
target check and options.
* config/arm/libunwind.S: Update comment relating to armv5.
Martin Liska [Fri, 18 May 2018 13:06:31 +0000 (15:06 +0200)]
gcov: add new option -t that prints output to stdout (PR gcov-profile/84846).
2018-05-18 Martin Liska <mliska@suse.cz>
PR gcov-profile/84846
* gcov.c (print_usage): Add new -t option.
(process_args): Handle the option.
(generate_results): Use stdout as output when requested by
the option.
2018-05-18 Martin Liska <mliska@suse.cz>
PR gcov-profile/84846
* doc/gcov.texi: Document -t option of gcov tool.
Toon Moene [Fri, 18 May 2018 09:07:39 +0000 (09:07 +0000)]
invoke.texi: Move -floop-unroll-and-jam documentation directly after that of -floop-interchange.
2018-05-18 Toon Moene <toon@moene.org>
* doc/invoke.texi: Move -floop-unroll-and-jam documentation
directly after that of -floop-interchange. Indicate that both
options are enabled by default when specifying -O3.
Kyrylo Tkachov [Fri, 18 May 2018 08:52:30 +0000 (08:52 +0000)]
[AArch64] Unify vec_set patterns, support floating-point vector modes properly
We've a deficiency in our vec_set family of patterns.
We don't support directly loading a vector lane using LD1 for V2DImode and all the vector floating-point modes.
We do do it correctly for the other integer vector modes (V4SI, V8HI etc) though.
The alternatives on the relative floating-point patterns only allow a register-to-register INS instruction.
That means if we want to load a value into a vector lane we must first load it into a scalar register and then
perform an INS, which is wasteful.
There is also an explicit V2DI vec_set expander dangling around for no reason that I can see. It seems to do the
exact same things as the other vec_set expanders. This patch removes that.
It now unifies all vec_set expansions into a single "vec_set<mode>" define_expand using the catch-all VALL_F16 iterator.
With this patch we avoid loading values into scalar registers and then doing an explicit INS on them to move them into
the desired vector lanes. For example for:
typedef float v4sf __attribute__ ((vector_size (16)));
typedef long long v2di __attribute__ ((vector_size (16)));
v2di
foo_v2di (long long *a, long long *b)
{
v2di res = { *a, *b };
return res;
}
* config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
iterator. Delete separate integer-mode vec_set<mode> expander.
(aarch64_simd_vec_setv2di): Delete.
(vec_setv2di): Delete.
(aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
the "w, r" alternative.
Martin Liska [Fri, 18 May 2018 08:43:19 +0000 (10:43 +0200)]
Radically simplify emission of balanced tree for switch statements.
2018-05-18 Martin Liska <mliska@suse.cz>
* passes.def: Add pass_lower_switch and pass_lower_switch_O0.
* tree-pass.h (make_pass_lower_switch_O0): New function.
* tree-switch-conversion.c (node_has_low_bound): Remove.
(node_has_high_bound): Likewise.
(node_is_bounded): Likewise.
(class pass_lower_switch): Make it a template type and create
two instances.
(pass_lower_switch::execute): Add template argument.
(make_pass_lower_switch): New function.
(make_pass_lower_switch_O0): New function.
(do_jump_if_equal): Remove.
(emit_case_nodes): Simplify to just handle all 3 cases and leave
all the hard work to tree optimization passes.
2018-05-18 Martin Liska <mliska@suse.cz>
* gcc.dg/tree-ssa/vrp104.c: Adjust dump file that is scanned.
* gcc.dg/tree-prof/update-loopch.c: Likewise.
Martin Liska [Fri, 18 May 2018 08:42:15 +0000 (10:42 +0200)]
Support lower and upper limit for -fdbg-cnt flag.
2018-05-18 Martin Liska <mliska@suse.cz>
* dbgcnt.c (limit_low): Renamed from limit.
(limit_high): New variable.
(dbg_cnt_is_enabled): Check for upper limit.
(dbg_cnt): Adjust dumping.
(dbg_cnt_set_limit_by_index): Add new argument for high
value.
(dbg_cnt_set_limit_by_name): Likewise.
(dbg_cnt_process_single_pair): Parse new format.
(dbg_cnt_process_opt): Use strtok.
(dbg_cnt_list_all_counters): Remove 'value' and add
'limit_high'.
* doc/invoke.texi: Document changes.
2018-05-18 Martin Liska <mliska@suse.cz>
* gcc.dg/ipa/ipa-icf-39.c: New test.
* gcc.dg/pr68766.c: Adjust pruned output.
There are four optabs for various forms of fused multiply-add:
fma, fms, fnma and fnms. Of these, only fma had a direct gimple
representation. For the other three we relied on special pattern-
matching during expand, although tree-ssa-math-opts.c did have
some code to try to second-guess what expand would do.
This patch removes the old FMA_EXPR representation of fma and
introduces four new internal functions, one for each optab.
IFN_FMA is tied to BUILT_IN_FMA* while the other three are
independent directly-mapped internal functions. It's then
possible to do the pattern-matching in match.pd and
tree-ssa-math-opts.c (via folding) can select the exact
FMA-based operation.
The BRIG & HSA parts are a best guess, but seem relatively simple.
2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
* doc/sourcebuild.texi (scalar_all_fma): Document.
* tree.def (FMA_EXPR): Delete.
* internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
* internal-fn.c (ternary_direct): New macro.
(expand_ternary_optab_fn): Likewise.
(direct_ternary_optab_supported_p): Likewise.
* Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
* builtins.c (fold_builtin_fma): Delete.
(fold_builtin_3): Don't call it.
* cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
* expr.c (expand_expr_real_2): Likewise.
* fold-const.c (operand_equal_p): Likewise.
(fold_ternary_loc): Likewise.
* gimple-pretty-print.c (dump_ternary_rhs): Likewise.
* gimple.c (DEFTREECODE): Likewise.
* gimplify.c (gimplify_expr): Likewise.
* optabs-tree.c (optab_for_tree_code): Likewise.
* tree-cfg.c (verify_gimple_assign_ternary): Likewise.
* tree-eh.c (operation_could_trap_p): Likewise.
(stmt_could_throw_1_p): Likewise.
* tree-inline.c (estimate_operator_cost): Likewise.
* tree-pretty-print.c (dump_generic_node): Likewise.
(op_code_prio): Likewise.
* tree-ssa-loop-im.c (stmt_cost): Likewise.
* tree-ssa-operands.c (get_expr_operands): Likewise.
* tree.c (commutative_ternary_tree_code, add_expr): Likewise.
* fold-const-call.h (fold_fma): Delete.
* fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
CFN_FNMA and CFN_FNMS.
(fold_fma): Delete.
* genmatch.c (combined_fn): New enum.
(commutative_ternary_tree_code): Remove FMA_EXPR handling.
(commutative_op): New function.
(commutate): Use it. Handle more than 2 operands.
(dt_operand::gen_gimple_expr): Use commutative_op.
(parser::parse_expr): Allow :c to be used with non-binary
operators if the commutative operand is known.
* gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
CFN_FMS, CFN_FNMA and CFN_FNMS.
(backprop::process_assign_use): Remove FMA_EXPR handling.
* hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
(gen_hsa_fma): New function.
(gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
IFN_FNMA and IFN_FNMS.
* match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
* gimple-fold.h (follow_all_ssa_edges): Declare.
* gimple-fold.c (follow_all_ssa_edges): New function.
* tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
gimple_build interface and use follow_all_ssa_edges to fold the result.
(convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
instead of checking for optabs directly.
* config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
rather than FMA_EXPRs.
* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
call to IFN_FMA instead of an FMA_EXPR.
gcc/brig/
* brigfrontend/brig-function.cc
(brig_function::get_builtin_for_hsa_opcode): Use BUILT_IN_FMA
for BRIG_OPCODE_FMA.
(brig_function::get_tree_code_for_hsa_opcode): Treat BUILT_IN_FMA
as a call.
Jim Wilson [Thu, 17 May 2018 22:37:38 +0000 (22:37 +0000)]
RISC-V: Optimize switch with sign-extended index.
gcc/
* expr.c (do_tablejump): When converting index to Pmode, if we have a
sign extended promoted subreg, and the range does not have the sign bit
set, then do a sign extend.
* config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
test, check for sign extended subreg and/or constant operands, and
do a sign extend in that case.
Jonathan Wakely [Thu, 17 May 2018 15:36:25 +0000 (16:36 +0100)]
PR libstdc++/85818 ensure path::preferred_separator is defined
Because path.cc is compiled with -std=gnu++17 the static constexpr
data member is implicitly 'inline' and so no definition gets emitted
unless it gets used in that translation unit. Other translation units
built as C++11 or C++14 still require a namespace-scope definition of
the variable, so mark the definition as used.
Jonathan Wakely [Thu, 17 May 2018 15:03:29 +0000 (16:03 +0100)]
PR libstdc++/85812 fix memory leak in std::make_exception_ptr
PR libstdc++/85812
* libsupc++/cxxabi_init_exception.h (__cxa_free_exception): Declare.
* libsupc++/exception_ptr.h (make_exception_ptr) [__cpp_exceptions]:
Refactor to separate non-throwing and throwing implementations.
[__cpp_rtti && !_GLIBCXX_HAVE_CDTOR_CALLABI]: Deallocate the memory
if constructing the object throws.