[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* einfo.adb (Derived_Type_Link): New function
(Set_Derived_Type_Link): New procedure.
(Write_Field31_Name): Output Derived_Type_Link.
* einfo.ads: New field Derived_Type_Link.
* exp_ch6.adb (Expand_Call): Warn if change of representation
needed on call.
* sem_ch13.adb: Minor addition of ??? comment.
(Rep_Item_Too_Late): Warn on case that is legal but could cause an
expensive implicit conversion.
* sem_ch3.adb (Build_Derived_Type): Set Derived_Type_Link if needed.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch3.adb (Build_Init_Procedure): Renamed Local_DF_Id
to DF_Id. Add new local variable DF_Call. Do not perform any
elaboration-related checks on the call to the partial finalization
routine within an init proc to avoid generating bogus elaboration
warnings on expansion-related code.
* sem_elab.adb (Check_A_Call): Move constant Access_Case to
the top level of the routine.  Ensure that Output_Calls takes
into account flags -gnatel and -gnatwl when emitting warnings
or info messages.
(Check_Internal_Call_Continue): Update the call to Output_Calls.
(Elab_Warning): Moved to the top level of routine Check_A_Call.
(Emit): New routines.
(Output_Calls): Add new formal parameter Check_Elab_Flag along with a
comment on usage. Output all warnings or info messages only when the
caller context demands it and the proper elaboration flag is set.

2014-07-29  Yannick Moy  <moy@adacore.com>

* sem_attr.adb (Analyze_Attribute/Attribute_Old):
Check rule about Old appearing in potentially unevaluated
expression everywhere, not only in Post.

2014-07-29  Arnaud Charlet  <charlet@adacore.com>

* sem_prag.adb: Update comment.
* a-except.adb, a-except-2005.adb: Minor editing.

2014-07-29  Pierre-Marie Derodat  <derodat@adacore.com>

* exp_dbug.adb (Debug_Renaming_Declaration):
Do not create renaming entities for renamings of non-packed
objects and for exceptions.

From-SVN: r213175

sem_ch3.adb, [...]: Minor reformatting.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_ch3.adb, sinfo.ads, types.ads, sem_prag.adb, a-except-2005.adb,
sem_ch6.adb, par-ch3.adb: Minor reformatting.

From-SVN: r213174

[multiple changes]

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Check_Return_Subtype_Indication): Reject a return
subtype indication in an extended return statement when the
return value is an ancestor of the return type of the function,
and that return type is a null record extension.

2014-07-29  Thomas Quinot  <quinot@adacore.com>

* sem_ch13.adb (Rep_Item_Too_Late): Specialize/clarify error
message produced for the case of a type-related representation
item that is made illegal by 13.10(1).
* gnat_rm.texi (Scalar_Storage_Order): Minor change in
documentation.

From-SVN: r213173

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* gnat_ugn.texi: Add section on Wide_Wide_Character encodings.
* erroutc.adb (Output_Error_Msgs): Take wide characters into
account in computing position of error flags.
* sinput.adb (Get_Column_Number): Take wide characters into
account.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* par-ch3.adb (P_Access_Type_Definition): The subtype indication
in an access type definition can carry a null_exclusion indicator.
* sem_ch3.adb (Access_Type_Declaration): If the subtype indication
carries a null_exclusion indicator, verify that the subtype
indication denotes an access type, and create a null-excluding
subtype for it.
* sinfo.ads, sinfo.adb: New attribute Null_Excluding_Subtype,
defined on N_Access_To_Object_Definition to indicate that the
subtype indication carries a null_exclusion indicator.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch6.adb (Add_Extra_Actual): Do not construct
the extra actual by name, generate a reference instead.

2014-07-29  Arnaud Charlet  <charlet@adacore.com>

* sem_prag.adb (Analyze_Pragma): Do not crash analyzing
Allow_Integer_Address if already set.
* a-except-2005.adb (Rcheck_PE_Stream_Operation_Not_Allowed):
Fix order, for consistency with Rmsg_xx declarations.

From-SVN: r213172

[multiple changes]

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_ch4.adb (Complete_Object_Operation): If the type of the
candidate subprogram is a limited view, use non-limited view
when available.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_ch13.adb: Minor change in RM reference.
* sem_mech.ads: Minor reformatting.
* einfo.ads: Minor comment fix.
* types.ads: Minor correction to range given for Mechanism_Type.
* exp_ch6.adb (Add_Invariant_And_Predicate_Checks): Do not
check predicate on way out for OUT or IN OUT parameters.
* par-ch3.adb (P_Constraint_Opt): Handle missing RANGE keyword
better (P_Range_Constraint): Corresponding fix.
* checks.ads: Minor comment clarification.

2014-07-29  Gary Dismukes  <dismukes@adacore.com>

* sem_ch8.adb (Analyze_Object_Renaming): Set the Is_Volatile
and Treat_As_Volatile flags based on whether the renamed object
is a volatile object.

From-SVN: r213170

[multiple changes]

2014-07-29  Olivier Hainque  <hainque@adacore.com>

* g-debpoo.adb
(Default_Alignment): Rename as Storage_Alignment. This is not
a "default" that can be overriden. Augment comment to clarify
intent and document why we need to manage alignment padding.
(Header_Offset): Set to Header'Object_Size instead of 'Size
rounded up to Storage_Alignment. Storage_Alignment on the
allocation header is not required by our internals so was
overkill. 'Object_Size is enough to ensure proper alignment
of the header address when substracted from a storage address
aligned on Storage_Alignment.
(Minimum_Allocation): Rename as Extra_Allocation, conveying that
this is always added on top of the incoming allocation requests.
(Align): New function, to perform alignment rounding operations.
(Allocate): Add comments on the Storage_Address computation
scheme and adjust so that the alignment padding applies to that
(Storage_Address) only.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* exp_ch3.adb (Default_Initialize_Object): Remove incorrect
pragma Unreferenced.
* cstand.adb (Create_Standard): Use E_Array_Type for standard
string types. Make sure index of Any_String/Any_Array is in a list.
* errout.adb: Minor reformatting.

From-SVN: r213169

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* gnat_ugn.texi: Clean up and correct documentation of warnings.
* usage.adb: Minor corrections to make sure warnings are properly
documented.
* warnsw.adb (Set_Warning_Switch): Remove redundant return statement.
(WA_Warnings): Add Warn_On_Suspicious_Modulus_Value.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* exp_ch3.adb (Expand_N_Object_Declaration): The dummy block
created to match internal sequence numbers between compilations
with/without abort must have its type properly set.

From-SVN: r213168

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* ali.adb (Initialize_ALI): Initialize SSO_Default_Specified
(Scan_ALI): Set SSO_Default in ALIs_Record (Scan_ALI): Set
SSO_Default_Specified.
* ali.ads (ALIs_Record): Add field SSO_Default
(SSO_Default_Specified): New global switch.
* bcheck.adb (Check_Consistent_SSO_Default): New procedure
(Check_Configuration_Consistency): Call this procedure
* einfo.adb (SSO_Set_High_By_Default): New
function (SSO_Set_Low_By_Default): New function
(Set_SSO_Set_High_By_Default): New procedure
(Set_SSO_Set_Low_By_Default): New procedure (Write_Entity_Flags):
List new flags
* einfo.ads (SSO_Set_Low_By_Default): New flag
(SSO_Set_High_By_Default): New flag
* freeze.adb (Set_SSO_From_Default): New procedure
(Freeze_Array_Type): Call Set_SSO_From_Default
(Freeze_Record_Type): Call Set_SSO_From_Default
* gnat_rm.texi: Document pragma Default_Scalar_Storage_Order
* lib-writ.adb (Write_ALI): Set OL/OH in P line as needed
* lib-writ.ads: Add OL/OH parameters to P line
* opt.adb: Set Default_SSO, Default_SSO_Config as appropriate
* opt.ads (Default_SSO): New global switch (Default_SSO_Config):
New global switch
* repinfo.adb (List_Scalar_Storage_Order): List SSO when it is
set by default using pragma Default_Scalar_Storage_Order.
* sem.ads (Scope_Stack_Entry): Add component Save_Default_SSO
* sem_ch13.adb (Inherit_Delayed_Rep_Aspects):
Clear SSO defaults when explicit SSO is inherited.
(Analyze_Attribute_Definition_Clause): Clear SSO defaults when
explicit SSO is specified.
(Inherit_Aspects_At_Freeze_Point):
Clear SSO default when inheriting SSO.
* sem_ch3.adb (Set_Default_SSO): New procedure
(Analyze_Private_Extension_Declaration): Set defualt SSO
(Array_Type_Declaration): ditto (Build_Derived_Array_Type): ditto
(Build_Derived_Private_Type): ditto (Build_Derived_Record_Type):
ditto (Build_Derived_Type): ditto (Make_Class_Wide_Type): ditto
(Record_Type_Declaration): ditto
* sem_ch8.adb (Pop_Scope): Restore Default_SSO (Push_Scope):
Save Default_SSO
* sem_prag.adb (Analyze_Pragma, case
Default_Scalar_Storage_Order): Set Default_SSO

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Valid_Operator_Definition): Verify that
all parameter have mode IN. This check must be done here for
subprogram instantiations that have operator names, because their
analysis does not follow the same path as that for subprogram
declarations.

From-SVN: r213167

freeze.adb (Freeze_Entity, [...]): Add a guard to make sure that the Etype of a component of the corresponding...

2014-07-29  Robert Dewar  <dewar@adacore.com>

* freeze.adb (Freeze_Entity, Concurrent_Type case): Add a guard
to make sure that the Etype of a component of the corresponding
record type is present before trying to freeze it.
* sem_ch5.adb: Minor reformatting.

From-SVN: r213165

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* exp_attr.adb, types.ads, types.h, exp_ch11.adb, a-except.adb,
a-except-2005.adb: Add new reason code PE_Stream_Operation_Not_Allowed,
and then use it when a stream operation is used from a library generic
when the restriction (No_Streams) is active.

2014-07-29  Thomas Quinot  <quinot@adacore.com>

* projects.texi: Fix minor typo.

2014-07-29  Yannick Moy  <moy@adacore.com>

* sem_attr.adb (Analyze_Attribute): Fix generation of warning.

2014-07-29  Arnaud Charlet  <charlet@adacore.com>

* sem_ch5.adb (Check_Unreachable_Code): Do not remove code in
CodePeer mode.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch7.adb (Find_Last_Init): Add local variable
Deep_Init_Found. Check the statement immediately following the
declaration if [Deep_]Initialization was not found.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_util.adb (Is_Aliased): It appears that
'reference-d and renamed objects still play some role in Boolean
expression with actions and cannot be finalized immediately.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* exp_dbug.adb (Qualify_Needed):  For debugging purposes,
Loop names are not part of the full qualification of entity names.

From-SVN: r213164

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* einfo.adb (Has_Protected): Test base type.
* sem_ch4.adb (Analyze_Allocator): Reorganize code to make sure
that we always properly check No_Protected_Type_Allocators.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_util.ads, sem_util.adb (Defining_Entity): Now applies to
loop declarations as well.
* exp_ch5.adb (Expand_Loop_Statement): Apply Qualify_Entity_Names
to an iterator loop, because it may contain local renaming
declarations that require debugging information.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_util.ads, exp_util.adb, sem_attr.adb: Minor reformatting.

From-SVN: r213163

einfo.ads, einfo.adb (Static_Real_Or_String_Predicate): New function

2014-07-29  Robert Dewar  <dewar@adacore.com>

* einfo.ads, einfo.adb (Static_Real_Or_String_Predicate): New function
(Set_Static_Real_Or_String_Predicate): New procedure
* sem_ch13.adb (Build_Predicate_Functions): Accomodate static
string predicates (Is_Predicate_Static): Handle static string
predicates.
* sem_eval.adb (Real_Or_String_Static_Predicate_Matches):
New procedure (Check_Expression_Against_Static_Predicate):
Deal with static string predicates, now fully implemented
(Eval_Relational_Op): Allow string equality/inequality as static
if not comes from source.

From-SVN: r213162

sem_aggr.adb, [...]: General cleanup of static predicate handling.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_aggr.adb, exp_ch5.adb, sem_ch5.adb, exp_util.adb, einfo.adb,
einfo.ads, sem_util.adb, sem_attr.adb, sem_case.adb, sem_eval.adb,
sem_eval.ads, sem_ch13.adb: General cleanup of static predicate
handling. Change name of Discrete_Predicate to
Discrete_Static_Predicate, and replace testing of the presence of this
field by testing the flag Has_Static_Expression.

From-SVN: r213161

gnat_rm.texi: Document pragma Unevaluated_Use_Of_Old.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* gnat_rm.texi: Document pragma Unevaluated_Use_Of_Old.
* opt.adb: Handle Uneval_Old.
* opt.ads (Uneval_Old, Uneval_Old_Config): New variables.
* par-prag.adb: Add dummy entry for pragma Unevaluated_Use_Of_Old.
* sem.ads (Save_Uneval_Old): New field in Scope_Stack_Entry.
* sem_attr.adb (Uneval_Old_Msg): New procedure.
* sem_ch8.adb (Push_Scope): Save Uneval_Old.
(Pop_Scope): Restore Uneval_Old.
* sem_prag.adb (Analyze_Pragma, case Unevaluated_Use_Of_Old):
Implemented.
* snames.ads-tmpl: Add entries for pragma Unevaluated_Use_Of_Old
Add entries for Name_Warn, Name_Allow.

From-SVN: r213160

sem_aggr.adb (Resolve_Array_Aggregate): Change Is_Static_Range to Is_OK_Static_Range.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_aggr.adb (Resolve_Array_Aggregate): Change Is_Static_Range
to Is_OK_Static_Range.
* sem_attr.adb (Eval_Attribute): Make sure we properly flag
static attributes (Eval_Attribute, case Size): Handle size of
zero properly (Eval_Attribute, case Value_Size): Handle size of
zero properly.
* sem_ch13.adb: Minor reformatting.
* sem_ch3.adb (Process_Range_Expr_In_Decl): Change
Is_Static_Range to Is_OK_Static_Range.
* sem_eval.adb (Eval_Case_Expression): Total rewrite, was
wrong in several ways (Is_Static_Range): Moved here from spec
(Is_Static_Subtype): Moved here from spec Change some incorrect
Is_Static_Subtype calls to Is_OK_Static_Subtype.
* sem_eval.ads: Add comments to section on
Is_Static_Expression/Raises_Constraint_Error (Is_OK_Static_Range):
Add clarifying comments (Is_Static_Range): Moved to body
(Is_Statically_Unevaluated): New function.
* sem_util.ads, sem_util.adb (Is_Preelaborable_Expression): Change
Is_Static_Range to Is_OK_Static_Range.
* sinfo.ads: Additional commments for Is_Static_Expression noting
that clients should almost always use Is_OK_Static_Expression
instead. Many other changes throughout front end units to obey
this rule.
* tbuild.ads, tbuild.adb (New_Occurrence_Of): Set Is_Static_Expression
for enumeration literal.
* exp_ch5.adb, sem_intr.adb, sem_ch5.adb, exp_attr.adb, exp_ch9.adb,
lib-writ.adb, sem_ch9.adb, einfo.ads, checks.adb, checks.ads,
sem_prag.adb, sem_ch12.adb, freeze.adb, sem_res.adb, exp_ch4.adb,
exp_ch6.adb, sem_ch4.adb, sem_ch6.adb, exp_aggr.adb, sem_cat.adb:
Replace all occurrences of Is_Static_Expression by
Is_OK_Static_Expression.

From-SVN: r213159

exp_ch4.adb (Process_Transient_Object): Remove constant In_Cond_Expr, use its initialization expression in place.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch4.adb (Process_Transient_Object): Remove constant
In_Cond_Expr, use its initialization expression in place.
* exp_ch7.adb (Process_Declarations): There is no need to check
that a transient object being hooked is controlled as it would
not have been hooked in the first place.
* exp_util.adb (Is_Aliased): 'Reference-d or renamed transient
objects are not considered aliased when the related context is
a Boolean expression_with_actions.
(Requires_Cleanup_Actions): There is no need to check that a transient
object being hooked is controlled as it would not have been hooked in
the first place.

From-SVN: r213158

[multiple changes]

2014-07-29  Robert Dewar  <dewar@adacore.com>

* errout.adb: Minor reformatting.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch3.adb (Default_Initialize_Object): Add new variables
Abrt_Blk and Dummy. Generate a dummy temporary when aborts are
not allowed to ensure the symmetrical generation of symbols.
* exp_ch7.adb (Build_Object_Declarations): Remove variables A_Expr
and E_Decl. Add new variables Decl and Dummy. Generate a dummy
temporary when aborts are not allowed to ensure symmertrical
generation of symbols.
* exp_intr.adb (Expand_Unc_Deallocation): Add new variable
Dummy. Generate a dummy temporary when aborts are not allowed
to ensure symmertrical generation of symbols.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* exp_dbug.adb (Debug_Renaming_Declaration): For an object
renaming, indicate that the renamed entity itself needs debug
information. This is necessary if that entity is a temporary,
e.g. part of the expansion of an explicit dereference in an
iterator.

From-SVN: r213157

[multiple changes]

2014-07-29  Thomas Quinot  <quinot@adacore.com>

* errout.adb (Set_Error_Posted): When propagating flag to
an enclosing named association, also propagate to the parent
of that node, so that named and positional associations are
treated consistently.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_attr.adb (Resolve_Attribute, case 'Update):  Set
Do_Range_Check properly on array component expressions that
have a scalar type. In GNATprove mode, only checks on scalar
components must be marked by the front-end.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sem_res.adb (Resolve_Type_Conversion): If the type of the
expression is a limited view, use the non-limited view when
available.

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch4.adb (Expand_N_Case_Expression): Mark the generated
case statement as coming from a conditional expression.
(Expand_N_If_Expression): Mark the generated if statement as
coming from a conditional expression.
* exp_ch5.adb (Expand_N_Case_Statement): Do not process controlled
objects found in case statement alternatives when the case
statement is actually a case expression.
(Expand_N_If_Statement):
Do not process controlled objects found in an if statement when
the if statement is actually an if expression.
* sinfo.adb (From_Conditional_Expression): New routine.
(Set_From_Conditional_Expression): New routine.
* sinfo.ads Add new semantic flag From_Conditional_Expression and
update related nodes.
(From_Conditional_Expression): New routine along with pragma Inline.
(Set_From_Conditional_Expression): New routine along with pragma Inline.

From-SVN: r213156

[multiple changes]

2014-07-29  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch7.adb (Build_BIP_Cleanup_Stmts): Remove
formal parameter Obj_Id and update the comment on usage. Renamed
Obj_Typ to Func_Typ and update all occurrences.
(Find_Last_Init): Remove formal parameter Decl and update the comment
on usage.
Remove local constants Obj_Id and Obj_Typ. Remove local variables
Init_Typ and Is_Conc. Remove the extraction of the initialization type.
(Find_Last_Init_In_Block): Remove formal parameter
Init_Typ and update the comment on usage.
(Is_Init_Call): Remove formal parameter Init_Typ and update the comment
on usage. Check whether the procedure call is an initialization
procedure of either the object type or the initialization type.
(Is_Init_Proc_Of): New routine.
(Process_Object_Declaration): Obj_Id and Obj_Typ are now global to this
routine. Add new variable Init_Typ. Add circuitry to extract the object
type as well as the initialization type.

2014-07-29  Robert Dewar  <dewar@adacore.com>

* sem_case.adb: Minor reformatting.
* sem_aux.ads: Minor reformatting.

2014-07-29  Ed Schonberg  <schonberg@adacore.com>

* sinfo.adb (Set_Else_Actions, Set_Then_Actions): Set parent
pointer on these fields, even though they are semantic, because
subsequent analysis and expansion of action nades may require
exploring the tree, for example to locate a node to be wrapped
when a function with controlled result is called.

2014-07-29  Claire Dross  <dross@adacore.com>

* sem_aux.adb (Get_Binary_Nkind): Use case on
Name_Id instead of an intermediate string.
(Get_Unary_Nkind): Use case on Name_Id instead of an intermediate
string.

2014-07-29  Sergey Rybin  <rybin@adacore.com frybin>

* gnat_ugn.texi (gnatelim, gnatstub, gnatmetric): Add note
about processing sources with preprocessor directives.

From-SVN: r213155

tree-sra.c (sra_modify_constructor_assign): Change type of stmt parameter to gimple.

2014-07-29  Martin Jambor  <mjambor@suse.cz>

* tree-sra.c (sra_modify_constructor_assign): Change type of stmt
parameter to gimple.
(sra_modify_assign): Likewise.

From-SVN: r213154

re PR middle-end/52478 (-ftrapv calls the wrong functions in libgcc)

2014-07-29  Richard Biener  <rguenther@suse.de>

PR middle-end/52478
* expr.c (expand_expr_real_2): Revert last change.

From-SVN: r213153

cgraph.c (cgraph_node::create_indirect_edge): Copy speculative data.

* cgraph.c (cgraph_node::create_indirect_edge): Copy speculative data.
* cgraph.h (cgraph_indirect_call_info): Add speculative data.
* gimple-fold.c (fold_gimple_assign): Fix check for virtual
call.
* ipa-devirt.c (ipa_dummy_polymorphic_call_context): Update
(contains_type_p): Forward declare.
(polymorphic_call_target_hasher::hash): Hash speculative info.
(polymorphic_call_target_hasher::equal): Compare speculative info.
(get_class_context): Handle speuclation.
(contains_type_p): Update.
(get_polymorphic_call_info_for_decl): Update.
(walk_ssa_copies): Break out from ...
(get_polymorphic_call_info): ... here; set speculative context
before giving up.
* ipa-prop.c (ipa_write_indirect_edge_info, ipa_read_indirect_edge_info):
Stream speculative context.
* ipa-utils.h (ipa_polymorphic_call_context): Add speculative info
(SPECULATIVE_OFFSET, SPECULATIVE_OUTER_TYPE,
SPECULATIVE_MAYBE_DERIVED_TYPE).
(possible_polymorphic_call_targets overriders): Update.
(dump_possible_polymorphic_call_targets overriders): Update.
(dump_possible_polymorphic_call_target_p overriders): Update.

From-SVN: r213152

gimple-fold.c (fold_gimple_assign): Fix condition guarding ipa-devirt path; fix thinko there.

* gimple-fold.c (fold_gimple_assign): Fix condition guarding
ipa-devirt path; fix thinko there.

From-SVN: r213150

Daily bump.

From-SVN: r213145

i386.c (ix86_return_in_memory): replace one ATTRIBUTE_UNUSED where the attribute can actually sometimes be unused.

2014-07-28 Trevor Saunders <tsaunders@mozilla.com>

* config/i386/i386.c (ix86_return_in_memory): replace one
ATTRIBUTE_UNUSED where the attribute can actually sometimes be unused.

From-SVN: r213142

* doc/invoke.texi (-Wno-odr): Fix @item entry.  Tweak wording.

From-SVN: r213136

Import config.sub from upstream.  No changes to config.guess.

From-SVN: r213134

Fix typo in ChangeLog

From-SVN: r213126

config.gcc (powerpc*-*-linux*): Include gnu-user.h in tm_file.

* config.gcc (powerpc*-*-linux*): Include gnu-user.h in tm_file.
* config/rs6000/sysv4.h (CC!_SPEC): Undefine it before defining it.
* config/rs6000/linux.h (CPLUSPLUS_CPP_SPEC): Delete define.
(LINK_GCC_C_SEQUENCE_SPEC): Likewise.
(USE_LD_AS_NEEDED): Likewise.
(ASM_APP_ON): Likewise.
(ASM_APP_OFF): Likewise.
(TARGET_POSIX_IO): Likewise.
* config/rs6000/linux64.h (CPLUSPLUS_CPP_SPEC): Likewise.
(LINK_GCC_C_SEQUENCE_SPEC): Likewise.
(USE_LD_AS_NEEDED): Likewise.
(ASM_APP_ON): Likewise.
(ASM_APP_OFF): Likewise.
(TARGET_POSIX_IO): Likewise.

From-SVN: r213125

re PR rtl-optimization/61801 (sched2 miscompiles syscall sequence with -g)

2014-07-28  Richard Biener  <rguenther@suse.de>

PR rtl-optimization/61801
* gcc.target/i386/pr61801.c: Fix testcase.

From-SVN: r213121

re PR middle-end/61734 (Regression in ABS_EXPR recognition)

PR middle-end/61734
* fold-const.c (fold_comparison): Disable X - Y CMP 0 to X CMP Y for
operators other than the equality operators.

From-SVN: r213118

re PR middle-end/52478 (-ftrapv calls the wrong functions in libgcc)

2014-07-28  Richard Biener  <rguenther@suse.de>

PR middle-end/52478
* optabs.c (gen_int_libfunc): For -ftrapv libfuncs make
sure to register SImode ones, not only >= word_mode ones.
* expr.c (expand_expr_real_2): When expanding -ftrapv
binops do not use OPTAB_LIB_WIDEN.

* gcc.dg/torture/ftrapv-1.c: New testcase.

From-SVN: r213117

re PR middle-end/61919 (FAIL: gfortran.dg/fmt_g0_6.f08   -O2  execution test)

gcc/
PR middle-end/61919
* tree-outof-ssa.c (insert_partition_copy_on_edge)
(insert_value_copy_on_edge, insert_rtx_to_part_on_edge)
(insert_part_to_rtx_on_edge): Copy partition_to_pseudo rtxes before
inserting them in the insn stream.

From-SVN: r213116

re PR middle-end/61913 (ICE in common_handle_option with -Wodr)

PR middle-end/61913
* common.opt (Wodr): Add Var.

From-SVN: r213115

re PR ipa/61921 (ICE: SIGSEGV in varpool_node::get_constructor(), at varpool.c:275 with -O2 -fipa-pta)

2014-07-28  Richard Biener  <rguenther@suse.de>

PR tree-optimization/61921
* tree-ssa-structalias.c (create_variable_info_for_1): Check
if there is a varpool node before dereferencing it.

* gfortran.dg/pr61921.f90: New testcase.

From-SVN: r213114

re PR rtl-optimization/61801 (sched2 miscompiles syscall sequence with -g)

2014-07-28  Richard Biener  <rguenther@suse.de>

PR rtl-optimization/61801
* gcc.target/i386/pr61801.c: New testcase.

From-SVN: r213111

[gcc/]

* graphite-sese-to-poly.c:
(new_pbb_from_pbb): Set a new id of pbb1->domain (instead of using the
id of the pbb), which contains pointer to the pbb1.

[gcc/testsuite]

* gcc.dg/graphite/isl-ast-gen-if-2.c: New testcase.

From-SVN: r213110

[gcc/]

* graphite-isl-ast-to-gimple.c:
(graphite_create_new_guard): New function.
(translate_isl_ast_node_if): New function.
(translate_isl_ast): Add calling of translate_isl_ast_node_if.

[gcc/testsuite]

* gcc.dg/graphite/isl-ast-gen-if-1.c: New testcase.

From-SVN: r213109

Add missing change from moxiebox patch posted earlier today.

From-SVN: r213108

Daily bump.

From-SVN: r213107

* MAINTAINERS: Remove my MIPS maintainer entry.

From-SVN: r213103

re PR c/61861 (Incorrect column number for -Wdiscarded-qualifiers)

PR c/61861
* macro.c (builtin_macro): Add location parameter.  Set
location of builtin macro to the expansion point.
(enter_macro_context): Pass location to builtin_macro.

* gcc.dg/pr61861.c: New test.

From-SVN: r213102

Add moxiebox target.

From-SVN: r213098

re PR libobjc/61920 (libobjc has undefined symbols on powerpc*-linux-gnu)

2014-07-27  Alan Modra  <amodra@gmail.com>
            Matthias Klose  <doko@ubuntu.com>

        PR libobjc/61920

        * encoding.c: Define rs6000_special_adjust_field_align_p.

Co-Authored-By: Matthias Klose <doko@ubuntu.com>
From-SVN: r213097

avx512f-vbroadcastf64x4-2.c: Fix the uninitialized variable problem.

gcc/testsuite/
* gcc.target/i386/avx512f-vbroadcastf64x4-2.c: Fix the uninitialized
variable problem.

From-SVN: r213096

re PR rtl-optimization/61926 (const-anchor optimisation is sensitive to ordering)

gcc/testsuite/
PR rtl-optimization/61926
* gcc.target/mips/const-anchor-1.c, gcc.target/mips/const-anchor-2.c:
Reverse argument order.
* gcc.target/mips/const-anchor-3.c, gcc.target/mips/const-anchor-4.c:
New XFAILed tests that match the original order.

From-SVN: r213093

libgcov-driver.c (struct gcov_filename_aux): Rename ...

* libgcov-driver.c (struct gcov_filename_aux): Rename ...
(struct gcov_filename): ... here.  Include buffer and max length
fields.
(gcov_max_filename): Remove.
(gi_filename): Remove.
(gcov_exit_compute_summary): Compute max filename here.
(gcov_exit_merge_gcda): Add filename parm, adjust.
(gcov_exit_merge_summary): Likewise.
(gcov_exit_dump_gcov): Adjust for struct gcov_filename changes.
(gcov_exit): Likewise.
(__gcov_init): Don't calculate max length here.
* libgcov_util.c (max_filename_len): Remove.
(read_gcda_file): Don't calculate max length here.
(gcov_read_profile_dir): Don't propagate here.
* libgcov-driver-system.c (alloc_filename_struct): Adjust for
struct gcov_filename changes.
(gcov_exit_open_gcda_file): Likewise.

From-SVN: r213092

aarch64.md (*extr_insv_lower_reg<mode>): Remove + from the read only register.

2014-07-26  Andrew Pinski  <apinski@cavium.com>

        * config/aarch64/aarch64.md (*extr_insv_lower_reg<mode>): Remove +
        from the read only register.

From-SVN: r213090

Daily bump.

From-SVN: r213089

Follow up to r213079.

2014-07-26  Tobias Burnus  <burnus@net-b.de>

        PR fortran/61881
        PR fortran/61888
        PR fortran/57305
        * intrinsic.texi (SIZEOF): Document changed behavior
        for polymorphic arrays.

From-SVN: r213085

re PR c/61077 (_Atomic in the return type or argument types of main not diagnosed)

PR c/61077
* gcc.dg/pr61077.c: Use \[^\n\]* instead of .* in the regexp.

From-SVN: r213084

re PR target/47230 (gcc fails to bootstrap on alpha in stage2 with "relocation truncated to fit: GPREL16 against ...")

PR target/47230
* configure.ac (alpha*-*-linux*): Use mh-alpha-linux.
* configure: Regenerate.

config/ChangeLog:

PR target/47230
* mh-alpha-linux: New file.

From-SVN: r213081

Minor tweaks.

From-SVN: r213080

check.c (gfc_check_sizeof): Permit for assumed type if and only if it has an array descriptor.

2014-07-26  Tobias Burnus  <burnus@net-b.de>

        * check.c (gfc_check_sizeof): Permit for assumed type if and
        only if it has an array descriptor.
        * intrinsic.c (do_ts29113_check): Permit SIZEOF.
        (add_functions): SIZEOF is an Inquiry function.
        * intrinsic.texi (SIZEOF): Add note that only contiguous
        arrays are permitted.
        * trans-expr.c (gfc_conv_intrinsic_to_class): Handle assumed
        rank.
        * trans-intrinsic.c (gfc_conv_intrinsic_sizeof): Handle
        assumed type + array descriptor, CLASS and assumed rank.
        (gfc_conv_intrinsic_storage_size): Handle class arrays.

2014-07-26  Tobias Burnus  <burnus@net-b.de>

        * gfortran.dg/sizeof_2.f90: Change dg-error.
        * gfortran.dg/sizeof_4.f90: New.
        * gfortran.dg/storage_size_1.f08: Correct expected
        value.

From-SVN: r213079

ira-costs.c (find_costs_and_classes): For -O0...

gcc/
* ira-costs.c (find_costs_and_classes): For -O0, use the best class
as the allocation class if it isn't likely to be spilled.

From-SVN: r213078

rtl.h (tls_referenced_p): Declare.

gcc/
* rtl.h (tls_referenced_p): Declare.
* rtlanal.c (tls_referenced_p_1, tls_referenced_p): New functions.
* config/mips/mips.c (mips_tls_symbol_ref_1): Delete.
(mips_cannot_force_const_mem): Use tls_referenced_p.
* config/pa/pa-protos.h (pa_tls_referenced_p): Delete.
* config/pa/pa.h (CONSTANT_ADDRESS_P): Use tls_referenced_p
instead of pa_tls_referenced_p.
* config/pa/pa.c (hppa_legitimize_address, pa_cannot_force_const_mem)
(pa_emit_move_sequence, pa_emit_move_sequence): Likewise.
(pa_legitimate_constant_p): Likewise.
(pa_tls_symbol_ref_1, pa_tls_referenced_p): Delete.
* config/rs6000/rs6000.c (rs6000_tls_referenced_p): Delete.
(rs6000_cannot_force_const_mem, rs6000_emit_move)
(rs6000_address_for_altivec): Use tls_referenced_p instead of
rs6000_tls_referenced_p.
(rs6000_tls_symbol_ref_1): Delete.

From-SVN: r213077

re PR target/44551 ([missed optimization] AVX vextractf128 after vinsertf128)

2014-07-26  Marc Glisse  <marc.glisse@inria.fr>

PR target/44551
gcc/
* simplify-rtx.c (simplify_binary_operation_1) <VEC_SELECT>:
Optimize inverse of a VEC_CONCAT.
gcc/testsuite/
* gcc.target/i386/pr44551-1.c: New file.

From-SVN: r213076

Daily bump.

From-SVN: r213071

Make FDO more tolerant to source changes

From-SVN: r213068

Implement N4051 - Allow typename in a template template parameter

cp/

2014-07-25  Edward Smith-Rowland  <3dw4rd@verizon.net>

Implement N4051 - Allow typename in a template template parameter
* parser.c (cp_parser_type_parameter_key): New funtion;
(cp_parser_token_is_type_parameter_key): Ditto;
(cp_parser_type_parameter): Look for type-parameter-key for all versions
but pedwarn for less than cxx1z.

testsuite/

2014-07-25  Edward Smith-Rowland  <3dw4rd@verizon.net>

Implement N4051 - Allow typename in a template template parameter
* lib/target-supports.exp (check_effective_target_c++1y): Now
means C++1y and up.
(check_effective_target_c++1y_down): New.
(check_effective_target_c++1z_only): New.
(check_effective_target_c++1z): New.
* g++.dg/cpp1z/typename-tmpl-tmpl-parm.C: New.
* g++.dg/cpp1z/typename-tmpl-tmpl-parm-neg.C: New.
* g++.dg/cpp1z/typename-tmpl-tmpl-parm-.C: New.

From-SVN: r213065

simplify.c (gfc_simplify_storage_size): Use proper integer kind for the returned value.

2014-07-25  Tobias Burnus  <burnus@net-b.de>

        * simplify.c (gfc_simplify_storage_size): Use proper
        integer kind for the returned value.

2014-07-25  Tobias Burnus  <burnus@net-b.de>

        * gfortran.dg/storage_size_5.f90: New.

From-SVN: r213060

lto-streamer-out.c (struct sccs): Turn to ...

* lto-streamer-out.c (struct sccs): Turn to ...
(class DFS): ... this one; refactor the DFS walk so it can
be re-done on per-SCC basis.
(DFS::DFS): New constructor.
(DFS::~DFS): New destructor.
(hash_tree): Add new MAP argument holding in-SCC hash values;
remove POINTER_TYPE hashing hack.
(scc_entry_compare): Rename to ...
(DFS::scc_entry_compare): ... this one.
(hash_scc): Rename to ...
(DFS::hash_scc): ... this one; pass output_block instead
of streamer_cache; work harder to get unique and stable SCC
hashes.
(DFS_write_tree): Rename to ...
(DFS::DFS_write_tree): ... this one; add SINGLE_P parameter.
(lto_output_tree): Update.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r213059

libgcov-driver.c (set_gcov_dump_complete, [...]): Remove global functions polluting user's namespace.

* libgcov-driver.c (set_gcov_dump_complete,
reset_gcov_dump_complete, get_gcov_dump_complete): Remove global
functions polluting user's namespace.
(gcov_exit): Set variable directly.
(gcov_clear): Reset variable directly.
* libgcov-interface.c (get_gcov_dymp_complete,
reset_gov_dump_complete): Remove declarations.
(__gcov_reset, __gcov_dump): Don't call them.

From-SVN: r213058

Convert lto streamer out hashing to inchash

No substantial changes, although the hash values will be slightly
different.

v2: add_flag moved to inchash. Some minor changes.

gcc/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* lto-streamer-out.c (hash_tree): Convert to inchash.

From-SVN: r213057

Convert the tree.c type hashing over to inchash

v2: Use commutative interface. Be much nearer to the old
code.

gcc/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* tree.c (build_type_attribute_qual_variant): Use inchash.
(type_hash_list): Dito.
(attribute_hash_list): Dito
(iterative_hstate_expr): Dito.
(iterative_hash_expr): Dito.
(build_range_type_1): Dito.
(build_array_type_1): Dito.
(build_function_type): Dito.
(build_method_type_directly): Dito.
(build_offset_type): Dito.
(build_complex_type): Dito.
(make_vector_type): Dito.
* tree.h (iterative_hash_expr): Add compat wrapper.
(iterative_hstate_expr): Add.

gcc/lto/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* lto.c (hash_canonical_type): Call iterative_hstate_expr.

From-SVN: r213056

Convert LTO type hashing to the new inchash interface

Should not really change any behavior, it's just a more abstract
interface, but uses the same underlying hash functions.

gcc/lto/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* lto.c (hash_canonical_type): Convert to inchash.
(iterative_hash_canonical_type): Dito.

From-SVN: r213055

Add an abstract incremental hash data type

Some files in gcc, like lto or tree, do large scale incremential hashing.
The current jhash implementation of this could be likely improved
by using an incremential hash that does not do a full rehashing
for every new value added.

This patch adds a new "inchash" class that abstracts the internal
state of the hash. This makes it easier to plug in new hashes
and also cleans up the code a bit.

Right now it is just implemented in the same way as the old
iterative hash in tree.c. The previous iterative hash code
from tree.c moved into a new separate file. Also I fixed up all
users to include the new header.

It should not really significantly change any hashing by itself,
it's mostly a cleanup at this point.

v2: Remove begin. Add commutative interface.
Add merge hash interface.  Add add_flag.

gcc/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* Makefile.in (OBJS): Add inchash.o.
(PLUGIN_HEADERS): Add inchash.h.
* ipa-devirt.c: Include inchash.h.
* lto-streamer-out.c: Dito.
* tree-ssa-dom.c: Dito.
* tree-ssa-pre.c: Dito.
* tree-ssa-sccvn.c: Dito.
* tree-ssa-tail-merge.c: Dito.
* asan.c: Dito.
* tree.c (iterative_hash_hashval_t): Move to ...
(iterative_hash_host_wide_int): Move to ...
* inchash.c: Here. New file.
* tree.h (iterative_hash_hashval_t): Move to ...
(iterative_hash_host_wide_int): Move to ...
* inchash.h: Here. New file.

gcc/lto/:

2014-07-25  Andi Kleen  <ak@linux.intel.com>

* lto.c: Include inchash.h

From-SVN: r213054

ffi.c: Do not include stdlib.h.

* src/alpha/ffi.c: Do not include stdlib.h.
(ffi_closure_osf_inner) <default>: Use FFI_ASSERT instead of abort.

From-SVN: r213049

re PR middle-end/61762 (failure to optimize memcpy from constant string)

2014-07-25  Richard Biener  <rguenther@suse.de>

PR middle-end/61762
PR middle-end/61894
* fold-const.c (native_encode_int): Add and handle offset
parameter to do partial encodings of expr.
(native_encode_fixed): Likewise.
(native_encode_real): Likewise.
(native_encode_complex): Likewise.
(native_encode_vector): Likewise.
(native_encode_string): Likewise.
(native_encode_expr): Likewise.
* fold-const.c (native_encode_expr): Add offset parameter
defaulting to -1.
* gimple-fold.c (fold_string_cst_ctor_reference): Remove.
(fold_ctor_reference): Handle all reads from tcc_constant
ctors.

* gcc.dg/pr61762.c: New testcase.
* gcc.dg/fold-cstring.c: Likewise.
* gcc.dg/fold-cvect.c: Likewise.

From-SVN: r213045

tree-inline.c (estimate_move_cost): Mark speed_p argument as possibly unused.

2014-07-25  Richard Biener  <rguenther@suse.de>

* tree-inline.c (estimate_move_cost): Mark speed_p argument
as possibly unused.

From-SVN: r213043

baseline_symbols.txt: Update.

* config/abi/post/alpha-linux-gnu/baseline_symbols.txt: Update.

From-SVN: r213040

ChangeLog entry moved from config/ChangeLog to gcc/ChangeLog.

From-SVN: r213039

string_view: Make the literal operators constexpr like the ctors they call.

2014-07-25  Ed Smith-Rowland  <3dw4rd@verizon.net>

* include/experimental/string_view: Make the literal operators
constexpr like the ctors they call.

From-SVN: r213037

libitm: Improve aarch64 _ITM_beginTransaction

        * config/aarch64/sjlj.S (_ITM_beginTransaction): Use post-inc
        addressing mode in epilogue.

From-SVN: r213035

Daily bump.

From-SVN: r213034

avr-c.c (avr_cpu_cpp_builtins): Add __AVR_DEVICE_NAME__.

* config/avr/avr-c.c (avr_cpu_cpp_builtins): Add __AVR_DEVICE_NAME__.

From-SVN: r213031

AArch64: emit .note.GNU-stack

2014-07-24  Kyle McMartin  <kyle@redhat.com>
* config/aarch64/aarch64-linux.h (TARGET_ASM_FILE_END): Define.

From-SVN: r213028

intrinsic.texi (Intrinsic Procedures): Move to correct menu position to match sectioning.

* intrinsic.texi (Intrinsic Procedures) <ATOMIC_DEFINE>: Move to
correct menu position to match sectioning.

From-SVN: r213027

go-gcc.cc (Gcc_backend::global_variable_set_init): Rename symtab_get_node to symtab_node::get.

* go-gcc.cc (Gcc_backend::global_variable_set_init): Rename
symtab_get_node to symtab_node::get.

From-SVN: r213026

rs6000-protos.h (rs6000_special_adjust_field_align_p): Add prototype.

gcc/

* config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align_p):
Add prototype.
* config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p): New
function.
* config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Call it.
* config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
* config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.

gcc/testsuite/

* gcc.target/powerpc/ppc64-abi-warn-3.c: New test.

* gcc.c-torture/execute/20050316-1.x: Add -Wno-psabi.
* gcc.c-torture/execute/20050604-1.x: Add -Wno-psabi.
* gcc.c-torture/execute/20050316-3.x: New file.  Add -Wno-psabi.
* gcc.c-torture/execute/pr23135.x: Likewise.

From-SVN: r213017

rs6000.c (rs6000_function_arg_boundary): In the AIX and ELFv2 ABI...

gcc/

* config/rs6000/rs6000.c (rs6000_function_arg_boundary): In the AIX
and ELFv2 ABI, do not use the "mode == BLKmode" check to test for
aggregate types.  Instead, *all* aggregate types, except for single-
element or homogeneous float/vector aggregates, are quadword-aligned
if required by their type alignment.  Issue -Wpsabi note when a type
is now treated differently than before.

gcc/testsuite/

* gcc.target/powerpc/ppc64-abi-warn-2.c: New test.

From-SVN: r213016

rs6000.c (rs6000_function_arg): If a float argument does not fit fully into floating-point registers...

gcc/

* config/rs6000/rs6000.c (rs6000_function_arg): If a float argument
does not fit fully into floating-point registers, and there is still
space in the register parameter area, use GPRs to pass those parts
of the argument.  Issue -Wpsabi note if any parameter is now treated
differently than before.
(rs6000_arg_partial_bytes): Update.

gcc/testsuite/

* gcc.target/powerpc/ppc64-abi-warn-1.c: New test.

From-SVN: r213015

struct-layout-1.exp: Load g++-dg.exp.

gcc/testsuite/

* g++.dg/compat/struct-layout-1.exp: Load g++-dg.exp.

From-SVN: r213014

elf.h: Define TARGET_UNWIND_TABLES_DEFAULT.

* config/alpha/elf.h: Define TARGET_UNWIND_TABLES_DEFAULT.

From-SVN: r213010

cygming-crtbegin.c (deregister_frame_fn): Newly public.

* config/i386/cygming-crtbegin.c (deregister_frame_fn): Newly public.
(__gcc_deregister_frame): Move logic to detect deregister function to...
(__gcc_register_frame): here, so it's consistent with the register logic.

From-SVN: r213009

ADA fixed for symtab_node issues.

        * gcc-interface/trans.c (finalize_nrv): Correct cgraph_node (symtab_node)
function used.
* gcc-interface/utils.c (rest_of_subprog_body_compilation): Likewise.
(gnat_write_global_declarations): Likewise.

From-SVN: r213007

rtl.h (target_rtl): Remove lang_dependent_initialized.

gcc/
* rtl.h (target_rtl): Remove lang_dependent_initialized.
* toplev.c (initialize_rtl): Don't use it.  Move previously
"language-dependent" calls to...
(backend_init): ...here.
(lang_dependent_init_target): Don't set lang_dependent_initialized.
Assert that RTL initialization hasn't happend yet.

From-SVN: r213004

re PR rtl-optimization/61629 (FAIL: gcc.dg/20020312-2.c (internal compiler error))

gcc/
PR rtl-optimization/61629
* reginfo.c (reinit_regs): Only call ira_init and recog_init if
they have already been initialized.

From-SVN: r213003

re PR middle-end/61268 (ICE in vt_expand_var_loc_chain, at var-tracking.c:8262)

gcc/
PR middle-end/61268
* function.c (assign_parm_setup_reg): Prevent invalid sharing of
DECL_INCOMING_RTL and entry_parm.
(get_arg_pointer_save_area): Likewise arg_pointer_save_area.
* calls.c (load_register_parameters): Likewise argument values.
(emit_library_call_value_1, store_one_arg): Likewise argument
save areas.
* config/i386/i386.c (assign_386_stack_local): Likewise the local
stack slot.
* explow.c (validize_mem): Modify the argument in-place.

From-SVN: r213002

[AArch64] Optimize epilogue when there is no frame pointer.

From-SVN: r213000

[AArch64] Optimize prologue when there is no frame pointer.

From-SVN: r212999

[AArch64] Infrastructure to allow optional use of write back.

From-SVN: r212997

[AArch64] Extend frame state to track WB candidates.

From-SVN: r212996

Another MIPS/R6000 IPA C++ refactoring fix.

* mips.c (mips_start_unique_function): Correct cgraph_node function
used.
* rs6000.c (call_ABI_of_interest): Likewise.
(rs6000_code_end): Likewise.

From-SVN: r212995

[gcc/]

* graphite-isl-ast-to-gimple.c:
(graphite_create_new_loop): Add calling of isl_id_free to properly
decrement reference counts.

[gcc/testsuite]

* gcc.dg/graphite/isl-ast-gen-blocks-4.c: New testcase.

From-SVN: r212994

IPA C++ refactoring fix.

* cgraphunit.c (compile): Correct function used.
        * rs6000.c (rs6000_xcoff_declare_function_name): Correct symtab_node funtion used.
        (rs6000_xcoff_declare_object_name): Likewise.

From-SVN: r212993

lto-streamer-out.c (tree_is_indexable): Consider IMPORTED_DECL as non-indexable.

* lto-streamer-out.c (tree_is_indexable): Consider IMPORTED_DECL
as non-indexable.

From-SVN: r212992

re PR lto/61802 (AArch64 execute.exp failures with LTO after r212467)

PR lto/61802
* varasm.c (bss_initializer_p): Handle offlined ctors.
(align_variable, get_variable_align): Likewise.
(make_decl_one_only): Likewise.
(default_binds_local_p_1): Likewise.
(decl_binds_to_current_def_p): Likewise.
(get_variable_section): Get constructor if it is offlined.
(assemble_variable_contents): Sanity check that the caller
streamed in the ctor in LTO.

From-SVN: r212991

[gcc/]

* graphite-isl-ast-to-gimple.c:
(binary_op_to_tree): Add calling of translate_isl_ast_node_block.
(gcc_expression_from_isl_expr_op): Move isl_ast_op_pdiv_q,
isl_ast_op_pdiv_r to the different case.

[gcc/testsuite]

* gcc.dg/graphite/isl-ast-gen-blocks-3.c: New testcase.

From-SVN: r212990

PR 61876: Do not convert cast + __builtin_round into __builtin_lround unless -fno-math-errno is used.

PR middle-end/61876
* convert.c (convert_to_integer): Do not convert BUILT_IN_ROUND and cast
when flag_errno_math is on.

From-SVN: r212989